Optimization & adoption

Once your AI spend is attributed, the same graph powers two more decisions: where can we spend less without losing quality? and which teams have genuinely adopted AI, and which need enablement? Optimization recommendations in the Value scorecard turn attributed spend into reviewable savings opportunities. Adoption intelligence reads consumption as cohort-level adoption signals: always privacy-safe, never surveillance. This page covers both.

Optimization

Optimization recommendations live in the Value scorecard (Cost / Value scorecard), next to the realized savings verified for the recommendations you applied. The recommendations form a queue of specific, evidence-backed opportunities to reduce AI cost without degrading the workload that depends on it. Each recommendation is a reviewable object you act on deliberately; Venturi never changes your systems on its own.

Recommendations are advisory by default

Venturi is read-only on your environment (see trust & security for how that boundary is enforced). An optimization recommendation is a proposal; Venturi never reaches into your AI configuration to apply a change. A recommendation takes effect only when a human approves it and enacts it in their own systems. For a named, operationally trusted workload that you have moved to Active (Gate), Venturi can also emit a signed enforcement recommendation (deny or reroute) to your own enforcement point, which decides whether to apply it. Venturi still applies nothing itself, and the decision-time interceptor stays fail-open.

What a recommendation contains

Each recommendation names the opportunity, the spend it touches, the estimated savings, and the workload-specific evidence on whether the change is quality-equivalent for that workload, which you validate before applying it. Every recommendation includes:

• Expected savings or risk reduction;
• The affected team, service, and model;
• The supporting evidence and its confidence;
• Operational risk notes;
• Any approval and rollback requirements;
• Its savings-eligibility state, computed from the same confidence object that gates chargeback.

You can open the evidence drawer behind any recommendation to see exactly which invocations it is based on and why the proposed change is expected to be safe for that workload.

Recommendations fall into the same five opportunity classes described in optimization & governance:

Recommendation	What it finds
Model right-sizing	An over-provisioned workload and a cheaper model candidate, backed by workload-specific evidence from your own shadow data; the change requires your validation before you apply it.
Price-reversal avoidance	Cases where a “cheaper” reasoning model actually costs more because it expands its thinking tokens; surfaced because Venturi costs on the served model with measured tokens, never the requested model or a list price.
Idle or redundant usage	Duplicate inference paths, idle deployments, and spend with no owner.
Provider or tier mix	Equivalent provider, region, and service-tier options within the workload’s constraints.
Commitment opportunities	Stable usage that may qualify for a lower effective rate under a provider commitment.

Reviewing and approving a recommendation

Each recommendation is an intervention that moves through the canonical intervention lifecycle (PROPOSED, APPROVED, ACTIVE, COMPLETED, or REJECTED, plus REVERTED when you undo an applied change), with collaboration at each step:

From	Action	To
PROPOSED	Approve after review.	APPROVED
PROPOSED	Reject with a recorded reason code.	REJECTED
APPROVED	You apply the change in your systems.	ACTIVE
ACTIVE	Record completion.	COMPLETED
ACTIVE	You undo the change and record the reversal.	REVERTED

• Comment and assign on a recommendation like any workflow object: route it to the team that owns the workload.
• Approve a recommendation to mark it for enactment, or reject it with a reason code if it is not right. Approval is an explicit Admin action, the proposer can never be its sole approver, and high-impact recommendations can require a second approver by configurable threshold.
• Apply the change in your own systems, then mark it applied; the intervention becomes ACTIVE, and later COMPLETED, with realized savings measured against the reconciled bill as described under verified savings. If you later revert an applied change in your own systems, mark it reverted: the intervention moves from ACTIVE to REVERTED, and the revert is recorded in the same trail.

Every decision (approve, reject, apply, revert) writes a governance trail so finance and engineering can reconstruct the basis later.

Verified savings

Savings count as verified and realized only after comparison with a frozen, usage-normalized counterfactual baseline. Organic usage growth and unrealized recommendations do not count as realized savings. Receipts preserve the evidence for your review. Venturi’s annual license does not charge a share of those savings. See the billing basis and the full measurement method in optimization & governance.

Measure AI Returns

A useful AI return answers a specific decision: whether to expand, change or stop a defined workflow. Venturi applies a shared measurement contract across teams. Customer-specific acceptance rules, labor bases and financial assumptions remain versioned inputs. These definitions describe the measurement contract; the release-state guide governs tenant availability and verification.

Keep Five Results Separate

Result	What It Supports
Accepted-work economics	Cost of delivering a unique, accepted business outcome within a declared boundary
Gross work equivalence	Baseline human effort needed to produce the accepted task mix
Net equivalent capacity	Work equivalence after current human effort and disjoint overhead, expressed in hours and E-FTE
AI-attributable operational change	An identified effect on output, effort, quality or redeployment
Realized financial return	Non-overlapping, evidenced financial benefit compared with complete incremental investment

A team can produce more accepted work without reducing payroll. A faster workflow can create useful capacity without proving cash savings. Show those results separately so finance and workflow owners can use the same evidence.

Read Existing Business Evidence

Venturi can read customer logs, telemetry and systems of record across the workflow. Broad read access supports reconstruction of work, acceptance and cost. It does not turn an event timestamp into active labor or establish what would have happened without AI.

Evidence	Read-Only Sources and Interpretation
AI execution and paid attempts	Existing provider, gateway, application and agent traces; preserve logical operations, attempts, retries and billable leaf events separately
Business acceptance and quality	Ticket, issue, deployment, transaction, approval, evaluation and incident histories; map native status semantics and later revocations
Human effort and capacity	Existing compatible worklogs, workforce schedules and team-period labor records; session-derived effort remains estimated
Money and investment	Billing, subscriptions, procurement, expense, payroll cost policies, revenue and general-ledger records; retain currency, period and cost scope
Comparable baseline	Existing customer experiments, rollout histories and comparable workload cohorts; preserve the pre-treatment task mix and reference condition
Identity and lineage	Dated organization, service, workflow and business-object identifiers; distinguish exact joins, inferred ownership and shared allocation

Declare a stable business work unit before joining the evidence. A case, accepted change, fulfilled order or approved analysis can span several systems and many model calls. Use deterministic identifiers or declared mappings; temporal proximity alone cannot establish an outcome link.

Existing status histories can establish acceptance through a versioned mapping. No new customer-written event, employee timer, survey or Venturi-assigned experiment is a prerequisite. Customer systems remain read-only. Venturi’s own configuration records the mapping and assumptions. Cohort-level measurement can proceed without request-to-task joins when its team, work and labor boundaries are supported.

Count the Complete Workflow

For accepted-work economics, divide complete included workflow cost by unique accepted outcomes. Keep AI-only cost visible when full workflow cost is unavailable. Include failed, rejected, abandoned and retried attempts in their proper cohort, together with review, rework, tools and allocated support. Count a business outcome once.

Align costs and outcomes to the same production cohort and maturity window. Opening work in progress, closing backlog and delayed acceptance need explicit treatment. Late events, corrections, reopens and revoked acceptance produce versioned revisions; they cannot silently inflate an earlier denominator.

Reconcile billed cash separately from amortized economic cost. Direct cost, shared allocation and the explicit unallocated residual must conserve the source total. Inclusive parent costs and their child charges cannot both be summed.

Match the Labor Basis

For a dedicated team, the aggregate path starts from accepted output per labor hour in a comparable baseline. Where stratum labor is supported, divide each stratum’s current accepted output by its own baseline output-per-labor-hour rate. When only total team hours exist, use a stable-mix team-period comparison, evidenced workload weights or a validated whole-basket model. Never reuse total team hours as the labor denominator for each stratum. Sum the supported baseline-equivalent hours, subtract current labor hours and disjoint overhead, then divide by compatible reference hours for the period.

For a task-effort path, estimate baseline labor for the accepted task mix. Subtract all current human effort, including review, failed attempts, rework, manual fallback and disjoint setup or maintenance. Divide the resulting net hours by the same declared reference-hours basis.

Both paths require finite, positive baseline and reference denominators. Unsupported strata stay explicit. The paths are alternatives and cannot be added together. They are comparable only when workflow, role mix, labor basis, period and overhead treatment align.

Recorded worklogs retain their source meaning. Inferred effort must be calibrated against compatible recorded effort at the available granularity. Union a person’s overlapping time across applications and devices, allocate concurrent tasks once, exclude machine activity and preserve unassigned or unobserved effort. Paid team capacity can reconcile totals but cannot validate active effort. A universal inactivity timeout is not an effort measurement.

Report role-equivalent capacity without ranking employees or treating every role as interchangeable. The minimum-cohort and access controls below apply to these views and exports. New work can have positive baseline-equivalent capacity without establishing historical hours released. Credit redeployment only when the subsequent use is evidenced.

Calculate Financial Return without Double Counting

Let B be attributable incremental financial benefit before included costs, and I be complete incremental investment and operating cost. Net benefit is B - I; financial ROI is (B - I) / I.

A known positive denominator is required for the ratio. If complete incremental cost is unknown, show supported components. Net benefit also remains unavailable unless a complete already-net change is independently evidenced. Even with that already-net numerator, ROI requires the complete positive investment denominator.

Keep realized cash reduction, incremental contribution profit, valued capacity, future cost avoidance and modeled risk separate. Revenue alone is not profit. Payback uses cash timing; unit economics can use a stated amortization policy. Do not add an upfront payment and its amortized expense to the same basis.

Cache, routing, batching, retry and procurement improvements can explain the same cost reduction. Allocate each financial benefit once. The same hour cannot fund both extra-output value and payroll savings unless separate uses are evidenced. Negative results and uncertainty crossing zero remain visible.

Establish What AI Changed

Freeze baseline eligibility, acceptance, maturity, task mix, cost basis and estimation rules before comparing results. Retain the entire eligible cohort, including failed work. Selecting only accepted AI tasks can bias the comparison.

Existing customer-run randomized experiments provide one comparison path. Existing phased rollouts or other natural variation can support a justified quasi-experiment. Assess pre-trends, selection, demand changes, spillovers and other contemporaneous changes. A historical or peer comparison remains descriptive when causal assumptions are unsupported. Matching alone does not establish causality.

State whether the reference is manual work or a prior AI workflow. Shared model uncertainty does not disappear with more events. External benchmarks and hierarchical estimates are explicitly modeled until validated for the customer. An uncalibrated sensitivity range is not a confidence interval or a guaranteed bound.

Metadata repricing supports cost scenarios. It cannot establish semantic equivalence or quality without readable, existing evaluation or outcome evidence. Ownership confidence, coverage and experimentally estimated effect uncertainty are separate quantities.

Check Ownership, Discovery and Attribution Reliability

Show exact, inferred and unresolved ownership with its provenance and freshness. Shared-cost allocation is a separate policy dimension: it preserves the resolved owner and cannot replace that owner. Independently labeled samples support correctness estimates within their tested population. Discovery coverage needs an independently supported eligible denominator; an unknown denominator leaves coverage unknown.

A high exact-join share measures linkage coverage. Invoice conservation checks the accounting total. Neither establishes ownership correctness or an AI-caused business benefit. Missing evidence withholds only the dependent calculation; supported costs, usage and outcomes remain visible.

Apply the Contract across Teams

Team	Example Accepted Work and Evidence
Engineering	Reviewed, deployed change after its defect window; repository, deployment, issue and incident histories
Product and digital experience	Fulfilled customer task or transaction; product events, experiments and billing
Customer support	Durable resolution after the reopen window; case, workforce and feedback records
Sales	Qualified stage completion or recognized sale; CRM history, fulfillment and finance records
Marketing	Validated campaign outcome; campaign costs, experiments, conversions and contribution records
Finance	Reconciled close, collectible recovery or accepted forecast; ledgers, payments and forecast vintages
Procurement	Accepted purchase or supplier decision; contracts, orders, invoices and delivery records
Legal	Approved matter or document; matter histories, approvals, worklogs and billed service cost
People operations	Accepted service case or demonstrated proficiency; HR service and learning records with cohort safeguards
IT and service operations	Restored service or fulfilled request; service-management, incident and work records
Security	Resolved incident or accepted control evidence; incident, exposure, remediation and cost records
Data and analytics	Accepted analysis or decision support; query, workflow, approval and downstream decision records
Research and development	Independently validated experiment or milestone; experiment, evaluation and subsequent outcome records
Operations and supply chain	Accepted order, plan or resource decision; process, inventory, quality and finance records

Native statuses require their documented semantics. A closed ticket can later reopen; a merged change can fail after deployment; a CRM stage is not recognized revenue. Choose a maturity window and retain those downstream consequences.

Metric Definitions

These definitions are generated from the Corpus revision c0fbdf62aa93b8937a26acac1356de625a50c6eb (current/product/domain-metrics/DOMAIN_METRIC_CATALOG_V1.json). Each named output keeps its own inputs, denominator and evidence class. The ordering supports economic decisions; it does not establish tenant availability.

01. Fully loaded cost per accepted outcome

METRIC-ROI-01. Total cost of every paid attempt, review, rework and allocated support cost in a comparable cohort / unique accepted business outcomes. Show AI-only and full-workflow cost separately; compare with the accepted-task baseline.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Outcome + cost; Accepted work, complete relevant cost components and their allocations

Evidence limits: Reconcile costs across all attempts and divide by mature accepted units; publish AI-only cost separately when the full workflow cost is incomplete; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Full workflow cohort cost

Contract: METRIC-ROI-01.full_workflow_cohort_cost.

Complete economic cost of every paid attempt and included human, tool, infrastructure and allocated support component for the evaluated work cohort.

Property	Contract
Calculation	sum(billable_leaf_costs) + sum(human_review_rework_costs) + sum(disjoint_support_allocations), after unique billable-leaf deduplication and cost-boundary reconciliation
Unit	currency
Required evidence	billable_leaf_costs; human_review_rework_costs; disjoint_support_allocations; full_cost_boundary; cost_component_completeness
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: billable_leaf_costs, human_review_rework_costs, disjoint_support_allocations, full_cost_boundary, cost_component_completeness. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show full cost boundary, allocation policy and monetary basis. A partial component total must not be labeled fully loaded.

Ai cohort cost

Contract: METRIC-ROI-01.ai_cohort_cost.

AI-only cost of all paid attempts in the evaluated cohort, independently available when other workflow costs are missing.

Property	Contract
Calculation	sum(reconciled_unique_ai_attempt_costs)
Unit	currency
Required evidence	reconciled_unique_ai_attempt_costs; ai_cost_boundary
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: reconciled_unique_ai_attempt_costs, ai_cost_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show ai cohort cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Full workflow cost per accepted outcome

Contract: METRIC-ROI-01.full_workflow_cost_per_accepted_outcome.

Complete cohort delivery cost divided by unique mature accepted business outcomes.

Property	Contract
Calculation	full_workflow_cohort_cost / unique_mature_accepted_outcome_count
Unit	currency/accepted_outcome
Required evidence	full_workflow_cohort_cost; unique_mature_accepted_outcome_count; cost_component_completeness; cost_allocation_basis; cohort_acceptance_and_maturity_contract
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: full_workflow_cohort_cost, unique_mature_accepted_outcome_count, cost_component_completeness, cost_allocation_basis, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Label fully loaded only when every relevant cost component is evidenced. Include failed, rejected, abandoned and retried work; preserve AI-only output independently.

Ai cost per accepted outcome

Contract: METRIC-ROI-01.ai_cost_per_accepted_outcome.

AI-only paid-attempt cost per unique mature accepted business outcome.

Property	Contract
Calculation	ai_cohort_cost / unique_mature_accepted_outcome_count
Unit	currency/accepted_outcome
Required evidence	ai_cohort_cost; unique_mature_accepted_outcome_count; ai_cost_boundary; cohort_acceptance_and_maturity_contract
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: ai_cohort_cost, unique_mature_accepted_outcome_count, ai_cost_boundary, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Explicitly label AI-only. This is not full workflow unit cost.

Baseline cost per accepted outcome

Contract: METRIC-ROI-01.baseline_cost_per_accepted_outcome.

Unit cost of the named comparable alternative at the same accepted-task and cost boundary.

Property	Contract
Calculation	baseline_complete_cohort_cost / baseline_unique_mature_accepted_outcome_count
Unit	currency/accepted_outcome
Required evidence	baseline_complete_cohort_cost; baseline_unique_mature_accepted_outcome_count; comparison_contract
Denominator	reference: baseline_unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_complete_cohort_cost, baseline_unique_mature_accepted_outcome_count, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show baseline cost per accepted outcome with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Accepted unit cost improvement

Contract: METRIC-ROI-01.accepted_unit_cost_improvement.

Signed economic unit-cost reduction against a comparable accepted-task alternative.

Property	Contract
Calculation	baseline_cost_per_accepted_outcome - full_workflow_cost_per_accepted_outcome
Unit	currency/accepted_outcome
Required evidence	baseline_cost_per_accepted_outcome; full_workflow_cost_per_accepted_outcome; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_cost_per_accepted_outcome, full_workflow_cost_per_accepted_outcome, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Positive is lower current economic unit cost; negative is deterioration. A cost-efficiency estimate is not cash savings.

02. Incremental net benefit and financial ROI

METRIC-ROI-02. Incremental attributable benefit before included costs minus complete incremental cost; divide that net benefit by the same incremental cost for ROI. Report net dollars alongside the ratio.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Finance + comparison; Finance transactions, declared investment scope, comparable histories and benefit dates

Evidence limits: Reconcile non-overlapping financial changes, net benefit and cash payback; preserve operational value when a monetary consequence is unsupported; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Incremental financial benefit before cost

Contract: METRIC-ROI-02.incremental_financial_benefit_before_cost.

Attributable non-overlapping financial benefit before costs included in the investment denominator.

Property	Contract
Calculation	sum(entry.amount_before_included_cost for unique, realized entries in realized_benefit_entries under benefit_cost_inclusion_map)
Unit	currency
Required evidence	realized_benefit_entries; benefit_cost_inclusion_map; financial_realization_evidence; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: realized_benefit_entries, benefit_cost_inclusion_map, financial_realization_evidence, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show incremental financial benefit before cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Complete incremental investment

Contract: METRIC-ROI-02.complete_incremental_investment.

Complete incremental cost of the declared investment, including implementation and operating changes without duplicate amortization.

Property	Contract
Calculation	sum(unique incrementally incurred amounts in incremental_cost_components under cost_inclusion_map)
Unit	currency
Required evidence	investment_scope; incremental_cost_components; cost_completeness_evidence; cost_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Name the alternative and cost changes relative to it. Separate cash timing from amortized economics.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: investment_scope, incremental_cost_components, cost_completeness_evidence, cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Publish signed complete incremental cost. A nonpositive investment leaves the conventional ROI ratio not applicable; missing cost is not zero.

Incremental net benefit

Contract: METRIC-ROI-02.incremental_net_benefit.

Complete signed incremental financial benefit after included costs, with an explicitly reconciled already-net route.

Property	Contract
Calculation	if net_benefit_evidence_mode = gross: incremental_financial_benefit_before_cost - complete_incremental_investment; if net_benefit_evidence_mode = independently_complete_already_net: evidenced_already_net_change
Unit	currency
Required evidence	net_benefit_evidence_mode; comparison_contract; benefit_cost_inclusion_map
Alternative evidence	mode: gross; required inputs: incremental_financial_benefit_before_cost; complete_incremental_investment; condition: Both benefit before included costs and the complete incremental investment are independently supported under the same declared scope. Subtract included costs once.; mode: independently_complete_already_net; required inputs: evidenced_already_net_change; independent_net_completeness_evidence; condition: The independently evidenced change already includes every cost required by the declared net scope. Do not subtract those costs again; this supports net dollars even when a separate complete ROI investment denominator is unavailable.
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Require the common comparison and inclusion evidence plus only the selected input_alternatives branch. Unknown complete investment withholds the gross branch; a complete independently evidenced already-net branch can still publish net dollars. Missing evidence never becomes zero, and ROI separately requires its full positive investment denominator.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Preserve negative net dollars. Unknown investment leaves net benefit unavailable unless an independently complete already-net change is evidenced. Never subtract included costs twice.

Financial roi

Contract: METRIC-ROI-02.financial_roi.

Signed realized net financial benefit divided by the same complete positive incremental investment.

Property	Contract
Calculation	incremental_net_benefit / complete_incremental_investment
Unit	ratio
Required evidence	incremental_net_benefit; complete_incremental_investment; investment_scope; benefit_cost_inclusion_map
Denominator	reference: complete_incremental_investment; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incremental_net_benefit, complete_incremental_investment, investment_scope, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Display as ratio or percent with the complete investment scope and net dollars. Preserve negative ROI. For known nonpositive investment show not applicable; for unknown investment show unavailable. Do not alias narrower AI-cost ROI.

03. Realized cash expense reduction

METRIC-ROI-03. Comparable baseline cash expenditure minus actual cash expenditure, net of implementation and operating changes. Require evidence of reduced invoices, overtime, contractor bills or eliminated spend.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Finance + comparison; Finance transactions, declared investment scope, comparable histories and benefit dates

Evidence limits: Reconcile non-overlapping financial changes, net benefit and cash payback; preserve operational value when a monetary consequence is unsupported; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Baseline cash expenditure

Contract: METRIC-ROI-03.baseline_cash_expenditure.

Comparable baseline cash expenditure for the accepted scope.

Property	Contract
Calculation	sum(baseline_dated_cash_expenditure_entries)
Unit	currency
Required evidence	baseline_cash_entries; cash_scope; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_cash_entries, cash_scope, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show baseline cash expenditure with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Actual cash expenditure

Contract: METRIC-ROI-03.actual_cash_expenditure.

Actual cash paid within the declared service or expense boundary.

Property	Contract
Calculation	sum(actual_dated_cash_expenditure_entries)
Unit	currency
Required evidence	actual_cash_entries; cash_scope
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: actual_cash_entries, cash_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show actual cash expenditure with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Gross cash expense reduction

Contract: METRIC-ROI-03.gross_cash_expense_reduction.

Attributable cash expense reduction before disjoint implementation and operating changes.

Property	Contract
Calculation	baseline_cash_expenditure - actual_cash_expenditure
Unit	currency
Required evidence	baseline_cash_expenditure; actual_cash_expenditure; reduced_invoice_or_overtime_or_contractor_evidence; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_cash_expenditure, actual_cash_expenditure, reduced_invoice_or_overtime_or_contractor_evidence, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show gross cash expense reduction with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Net realized cash expense reduction

Contract: METRIC-ROI-03.net_realized_cash_expense_reduction.

Realized cash expense reduction after implementation and operating changes not already included.

Property	Contract
Calculation	gross_cash_expense_reduction - disjoint_incremental_implementation_and_operating_cash_cost
Unit	currency
Required evidence	gross_cash_expense_reduction; disjoint_incremental_implementation_and_operating_cash_cost; benefit_cost_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: gross_cash_expense_reduction, disjoint_incremental_implementation_and_operating_cash_cost, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	A signed net cash measure. Unchanged payroll or hours-equivalence alone does not support this output; retain every already-included cost in the investment denominator if used for ROI.

04. Net labor-equivalent capacity, released effort and redeployment

METRIC-ROI-04. Estimate baseline labor required for the same accepted task mix using comparable team output per labor hour or task effort; subtract all current human effort and disjoint overhead. Report role-equivalent hours/E-FTE, separately identified released hours and evidenced redeployment. Section 8 defines the common contract and prevents conversion into unsupported payroll savings.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Outcome + team labor capacity or recorded/inferred task effort + baseline; Accepted output and existing team labor capacity, recorded effort or calibrated activity inference; comparable baseline

Evidence limits: Apply section 8.3; distinguish output equivalence, released labor, causal effect and redeployment; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Baseline accepted output per labor hour

Contract: METRIC-ROI-04.baseline_accepted_output_per_labor_hour.

Accepted baseline production rate within a supported homogeneous task and role stratum.

Property	Contract
Calculation	baseline_mature_accepted_output / baseline_labor_hours
Unit	accepted_outcome/labor_hour
Required evidence	baseline_mature_accepted_output; baseline_labor_hours; labor_basis; stratum_contract; comparison_contract
Denominator	reference: baseline_labor_hours; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Baseline accepted output and labor hours must both be finite and strictly positive within the same supported stratum and compatible labor basis.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_mature_accepted_output, baseline_labor_hours, labor_basis, stratum_contract, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute rates within compatible strata. Never invent stratum hours from total team hours or average unlike rates.
Display	Show baseline vintage, labor basis and task stratum. A zero-output baseline or missing labor hours is unsupported.

Baseline equivalent hours

Contract: METRIC-ROI-04.baseline_equivalent_hours.

Human labor that the comparable accepted output basket would require under the declared baseline and role basis.

Property	Contract
Calculation	if calculation_path = aggregate_output_labor: sum_s(current_mature_accepted_output_by_supported_stratum[s] / (baseline_mature_accepted_output_by_supported_stratum[s] / baseline_labor_hours_by_supported_stratum[s])); if calculation_path = accepted_task_effort: sum(supported_expected_baseline_hours_per_accepted_item_and_role); if calculation_path = aggregate_validated_production_model: validated_historical_production_model.predict_required_labor_hours(current_accepted_output_basket, current_role_labor_mix)
Unit	labor_hour
Required evidence	calculation_path; current_mature_accepted_work; role_and_labor_basis; lifecycle_or_validated_lag_alignment; baseline_version_and_reference_condition
Alternative evidence	mode: aggregate_output_labor; required inputs: current_mature_accepted_output_by_supported_stratum; baseline_mature_accepted_output_by_supported_stratum; baseline_labor_hours_by_supported_stratum; condition: Each baseline accepted-output count, labor-hours value and derived rate is finite and strictly positive; strata and labor bases are compatible. Total team hours alone require stable task mix, existing validated weights or a separately validated whole-basket production model, never invented stratum hours.; mode: accepted_task_effort; required inputs: accepted_work_item_ids_and_roles; supported_expected_baseline_hours_per_accepted_item_and_role; condition: Expected baseline labor is supported for the same task scope, complexity, role and acceptance standard, including ordinary failed work, review and support in the cost-to-deliver basis.; mode: aggregate_validated_production_model; required inputs: current_accepted_output_basket; current_role_labor_mix; validated_historical_production_model; model_prediction_uncertainty; condition: Existing total team labor supports a validated whole-output-basket model with comparable pre-treatment task mix and compatible role/labor basis. The output is a baseline estimate with retained prediction uncertainty; no invented task-category labor allocations.
Calculation conditions	aggregate output labor: Every included baseline accepted-output count, baseline labor-hours value and derived baseline_accepted_output_per_labor_hour_s must be finite and strictly positive.; Each current output stratum must have a supported compatible baseline. Publish explicitly scoped partial estimates with unsupported strata visible; never silently omit or fill them.; Current accepted output must be finite and nonnegative; complete lifecycle or validated lag/inventory alignment is required.; aggregate validated production model: Model validation supports the intended team/workflow/time population and outperforms or justifies its use against a simpler benchmark.; The prediction is finite, nonnegative and retains shared baseline/model uncertainty; unsupported current output strata remain explicit.; accepted task effort: Expected baseline hours per accepted task and role must be finite, nonnegative, supported and compatible with the current work scope.; Retain every eligible unsuccessful attempt in current human hours; newly enabled demand is work equivalence, not released labor.
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Use role- and period-specific labor/reference-hour bases, complete unsuccessful work and review, disjoint overhead and a supported task-mix/lag adjustment; no universal reference-hour default.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Require the common work, labor-basis, timing and baseline evidence plus only the selected input_alternatives branch. Missing, zero or nonfinite baseline rate blocks only affected aggregate strata. Preserve unsupported strata and disclose any scoped partial result. Do not require task effort for a supported aggregate estimate, or aggregate stratum hours for a supported task-effort estimate.
Aggregation	Sum compatible role hours across disjoint work scopes; preserve unsupported strata explicitly. Path A and Path B are alternatives, never additive.
Display	Human work equivalence includes newly enabled valuable work and does not prove historical hours were released. Human-infeasible work has no default finite human equivalent. Path A requires positive finite baseline output, labor hours and rate for every included stratum; unsupported strata remain visible.

Recorded current human hours

Contract: METRIC-ROI-04.recorded_current_human_hours.

Recorded human labor consumed by every evaluated attempt, including review, failures, rework, supervision and supported shared work.

Property	Contract
Calculation	sum(nonoverlapping_recorded_person_hours_in_workflow_boundary)
Unit	labor_hour
Required evidence	recorded_labor_records; person_time_overlap_resolution; labor_basis; workflow_labor_boundary; unassigned_and_shared_effort
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: recorded_labor_records, person_time_overlap_resolution, labor_basis, workflow_labor_boundary, unassigned_and_shared_effort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Union overlapping intervals per person, sum across people and compatible roles, preserve unallocated time and include all evaluated work.
Display	Show the source’s recorded labor basis. Paid, scheduled, worked and active effort hours are distinct; worklogs are recorded claims within their source meaning.

Inferred current human hours

Contract: METRIC-ROI-04.inferred_current_human_hours.

Calibrated estimate of workflow human effort from existing activity records.

Property	Contract
Calculation	sum_person(integral_over_supported_intervals(workflow_allocation_weight_person(t) dt)); sum_workflows(weight_person(t)) <= 1
Unit	labor_hour
Required evidence	activity_intervals; actor_classification; calibrated_allocation_model; held_out_effort_validation; unallocated_effort; workflow_labor_boundary
Denominator	Not applicable
Evidence class	inferred_effort
Baseline	Calibrate against existing recorded effort with compatible semantics; staffed totals alone cannot validate active-effort estimates.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: activity_intervals, actor_classification, calibrated_allocation_model, held_out_effort_validation, unallocated_effort, workflow_labor_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Union per-person intervals across apps/devices; allocate at most one person-hour per elapsed hour, retain unallocated remainder and propagate shared model uncertainty.
Display	Publish inferred effort with calibration and uncertainty. Missing events do not imply idleness; a sensitivity range is not a confidence interval.

Net labor equivalent hours

Contract: METRIC-ROI-04.net_labor_equivalent_hours.

Baseline-equivalent work delivered beyond all included current human effort and disjoint overhead.

Property	Contract
Calculation	baseline_equivalent_hours - all_current_human_hours - disjoint_setup_maintenance_and_other_overhead_hours
Unit	labor_hour
Required evidence	baseline_equivalent_hours; all_current_human_hours; current_effort_evidence_basis; disjoint_setup_maintenance_and_other_overhead_hours; role_and_labor_basis; overhead_inclusion_map
Input definitions	all current human hours: Complete current workflow human labor on the declared compatible role/basis, assembled from recorded and/or calibrated inferred effort with disjoint coverage; never add overlapping recorded and inferred estimates. Includes unsuccessful attempts, review, rework, supervision, shared and unassigned effort within the boundary.; disjoint setup maintenance and other overhead hours: Only hours not already included in all_current_human_hours; allocation version and covered period are explicit.
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Use role- and period-specific labor/reference-hour bases, complete unsuccessful work and review, disjoint overhead and a supported task-mix/lag adjustment; no universal reference-hour default.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_equivalent_hours, all_current_human_hours, current_effort_evidence_basis, disjoint_setup_maintenance_and_other_overhead_hours, role_and_labor_basis, overhead_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Sum compatible signed role-hour components. Do not add aggregate and task-reconstruction estimates or deduct overhead twice.
Display	Show negative capacity as well as positive. Preserve recorded/inferred effort provenance, task/role composition and comparison condition; this is not payroll saving.

Efte

Contract: METRIC-ROI-04.efte.

Equivalent full-time capacity in the stated customer period and compatible role/labor basis.

Property	Contract
Calculation	net_labor_equivalent_hours_for_role / customer_period_reference_hours_for_role
Unit	FTE_equivalent
Required evidence	net_labor_equivalent_hours_for_role; customer_period_reference_hours_for_role; reference_hours_provenance; role_and_labor_basis; reporting_period
Input definitions	net labor equivalent hours for role: The role-scoped value of net_labor_equivalent_hours, not an unweighted mixture of incompatible role or labor bases.; customer period reference hours for role: Customer-specific period/role reference hours with explicit worked, paid, scheduled or active-effort basis and provenance; finite and strictly positive.
Denominator	reference: customer_period_reference_hours_for_role; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Finite positive customer period and role reference hours must use the same paid/worked/active labor basis as the numerator.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Use role- and period-specific labor/reference-hour bases, complete unsuccessful work and review, disjoint overhead and a supported task-mix/lag adjustment; no universal reference-hour default.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: net_labor_equivalent_hours_for_role, customer_period_reference_hours_for_role, reference_hours_provenance, role_and_labor_basis, reporting_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible hours before division by the declared common basis. Never sum incompatible role ratios or add Path A and Path B.
Display	Display period, role and labor basis next to E-FTE. No universal 160-hour or 2080-hour default; no claim of employees replaced or cash saved.

Scenario efte

Contract: METRIC-ROI-04.scenario_efte.

Explicit scenario for labor-equivalent capacity when baseline, effort or reference hours rely on unvalidated assumptions.

Property	Contract
Calculation	(assumed_baseline_equivalent_hours - assumed_current_human_hours - assumed_disjoint_overhead_hours) / assumed_period_reference_hours
Unit	FTE_equivalent
Required evidence	scenario_baseline_hours; scenario_current_hours; scenario_overhead_hours; assumed_period_reference_hours; scenario_assumptions_and_ranges
Denominator	reference: assumed_period_reference_hours; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Finite positive declared scenario reference hours must match the scenario labor basis, role and period.; applies when: The scenario uses the declared period and role labor basis.
Evidence class	modeled_scenario
Baseline	Explicit assumptions for a named alternative; do not merge scenario values with calibrated estimates.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: scenario_baseline_hours, scenario_current_hours, scenario_overhead_hours, assumed_period_reference_hours, scenario_assumptions_and_ranges. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Combine only scenarios with a common declared basis and propagate shared parameter assumptions.
Display	Always label modeled scenario and show assumed hours/reference. Scenario ranges are not guaranteed actual-labor bounds.

Ai attributable released hours

Contract: METRIC-ROI-04.ai_attributable_released_hours.

Reduction in labor consumed for comparable existing demand attributable to AI under a supported identification design.

Property	Contract
Calculation	identified_expected_human_hours_without_intervention - identified_expected_human_hours_with_intervention_for_existing_demand
Unit	labor_hour
Required evidence	existing_demand_evidence; complete_cohort_effort; identified_comparison; task_mix_and_labor_basis; effect_uncertainty
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: existing_demand_evidence, complete_cohort_effort, identified_comparison, task_mix_and_labor_basis, effect_uncertainty. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Keep released labor separate from output-equivalent capacity and newly enabled work. Do not infer release solely from higher output at unchanged staffing.

Evidenced redeployed hours

Contract: METRIC-ROI-04.evidenced_redeployed_hours.

Supported labor that moved from the measured workflow to separately evidenced accepted work.

Property	Contract
Calculation	sum(nonoverlapping_role_hours_linked_to_verified_redeployment_assignments_and_accepted_work)
Unit	labor_hour
Required evidence	redeployment_assignments; role_effort_records; destination_accepted_work; source_and_destination_overlap_ledger
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: redeployment_assignments, role_effort_records, destination_accepted_work, source_and_destination_overlap_ledger. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show where capacity moved. Monetary value requires a separate evidenced economic consequence and cannot duplicate the same released hour’s value.

05. Incremental accepted throughput

METRIC-ROI-05. Additional accepted outcomes per paid team-hour or fixed resource budget versus the comparator, with quality and demand held comparable. Monetize only fulfilled demand or verified avoided expense.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Outcome + comparison; Accepted output and existing team labor capacity, recorded effort or calibrated activity inference; comparable baseline

Evidence limits: Apply section 8.3; distinguish output equivalence, released labor, causal effect and redeployment; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Accepted throughput per paid team hour

Contract: METRIC-ROI-05.accepted_throughput_per_paid_team_hour.

Mature accepted output delivered per paid team hour within an evidenced workflow boundary.

Property	Contract
Calculation	unique_mature_accepted_outcome_count / paid_team_hours
Unit	accepted_outcome/paid_team_hour
Required evidence	unique_mature_accepted_outcome_count; paid_team_hours; workflow_labor_boundary; cohort_acceptance_and_maturity_contract
Denominator	reference: paid_team_hours; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: unique_mature_accepted_outcome_count, paid_team_hours, workflow_labor_boundary, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Explicitly label paid-hours basis and task mix; paid-hours productivity is not active-effort productivity.

Incremental accepted throughput per paid hour

Contract: METRIC-ROI-05.incremental_accepted_throughput_per_paid_hour.

Signed change in comparable accepted output per paid team hour.

Property	Contract
Calculation	current_accepted_throughput_per_paid_hour - baseline_accepted_throughput_per_paid_hour
Unit	accepted_outcome/paid_team_hour
Required evidence	current_accepted_throughput_per_paid_hour; baseline_accepted_throughput_per_paid_hour; comparison_contract; demand_and_quality_comparability
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_accepted_throughput_per_paid_hour, baseline_accepted_throughput_per_paid_hour, comparison_contract, demand_and_quality_comparability. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show incremental accepted throughput per paid hour with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Incremental accepted output at fixed budget

Contract: METRIC-ROI-05.incremental_accepted_output_at_fixed_budget.

Additional accepted output under the same declared resource budget and comparable quality/demand.

Property	Contract
Calculation	current_accepted_output_at_budget - baseline_accepted_output_at_same_budget
Unit	accepted_outcome
Required evidence	current_accepted_output_at_budget; baseline_accepted_output_at_same_budget; fixed_resource_budget; comparison_contract; demand_and_quality_comparability
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_accepted_output_at_budget, baseline_accepted_output_at_same_budget, fixed_resource_budget, comparison_contract, demand_and_quality_comparability. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show incremental accepted output at fixed budget with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Incremental fulfilled demand contribution

Contract: METRIC-ROI-05.incremental_fulfilled_demand_contribution.

Realized incremental contribution from fulfilled demand enabled by added throughput, after associated delivery costs.

Property	Contract
Calculation	incremental_recognized_revenue_from_fulfilled_demand - associated_incremental_delivery_costs
Unit	currency
Required evidence	fulfilled_demand_records; recognized_revenue; associated_delivery_costs; identified_comparison; nonoverlapping_benefit_ledger
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: fulfilled_demand_records, recognized_revenue, associated_delivery_costs, identified_comparison, nonoverlapping_benefit_ledger. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Do not monetize mere spare capacity or unfulfilled demand. This benefit may overlap rank 06 and must have one shared benefit ID.

Verified avoided throughput expense

Contract: METRIC-ROI-05.verified_avoided_throughput_expense.

Avoided expense from the additional throughput only after the spending counterfactual and period are supported.

Property	Contract
Calculation	supported_counterfactual_expense - actual_comparable_expense - disjoint_intervention_cost
Unit	currency
Required evidence	mature_expense_counterfactual; actual_expense; disjoint_intervention_cost; nonoverlapping_benefit_ledger
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: mature_expense_counterfactual, actual_expense, disjoint_intervention_cost, nonoverlapping_benefit_ledger. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Dated future plans remain a forecast until their period and counterfactual mature. Unchanged headcount alone proves no avoided hire.

06. Incremental contribution profit

METRIC-ROI-06. Change in revenue less the associated delivery costs, including AI. Test conversion, retention, expansion, acquisition and new-feature monetization separately; join transactions to the intervention.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Finance + comparison; Finance transactions, declared investment scope, comparable histories and benefit dates

Evidence limits: Reconcile non-overlapping financial changes, net benefit and cash payback; preserve operational value when a monetary consequence is unsupported; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Incremental recognized revenue

Contract: METRIC-ROI-06.incremental_recognized_revenue.

Attributable change in recognized revenue for a separately evaluated conversion, retention, expansion, acquisition or new-feature effect.

Property	Contract
Calculation	current_recognized_revenue - comparable_counterfactual_recognized_revenue
Unit	currency
Required evidence	transaction_and_fulfillment_records; recognized_revenue; commercial_effect_type; identified_comparison
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: transaction_and_fulfillment_records, recognized_revenue, commercial_effect_type, identified_comparison. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show the commercial effect type. Pipeline, bookings, cash receipts and revenue are distinct; revenue alone is not profit.

Associated incremental delivery cost

Contract: METRIC-ROI-06.associated_incremental_delivery_cost.

All delivery costs associated with the incremental commercial outcome, including AI and disjoint implementation/operations.

Property	Contract
Calculation	sum(unique_associated_incremental_delivery_cost_components)
Unit	currency
Required evidence	delivery_cost_components; cost_completeness; commercial_effect_type; benefit_cost_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: delivery_cost_components, cost_completeness, commercial_effect_type, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show associated incremental delivery cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Incremental contribution profit

Contract: METRIC-ROI-06.incremental_contribution_profit.

Change in contribution after the delivery costs associated with the attributable commercial effect.

Property	Contract
Calculation	incremental_recognized_revenue - associated_incremental_delivery_cost
Unit	currency
Required evidence	incremental_recognized_revenue; associated_incremental_delivery_cost; commercial_effect_type; nonoverlapping_benefit_ledger
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incremental_recognized_revenue, associated_incremental_delivery_cost, commercial_effect_type, nonoverlapping_benefit_ledger. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Report conversion, retention, expansion, acquisition and new-feature effects separately; prevent overlapping transaction benefits. This is already net of mapped delivery costs.

07. Avoided rework, error and remediation cost

METRIC-ROI-07. Difference in quality-adjusted rework hours, refunds, defects, repeat contacts and other realized losses, less the intervention cost. Include downstream failures after initial acceptance.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Outcome + finance; Defects/reopens, review/intervention, failed attempts and linked terminal outcomes

Evidence limits: Reconstruct accepted automation, rework and rescue; no-human-intervention claims require coverage of the intervention path; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Rework hours reduction

Contract: METRIC-ROI-07.rework_hours_reduction.

Comparable reduction in complete downstream rework and remediation person-hours, retaining recorded or inferred effort provenance.

Property	Contract
Calculation	baseline_rework_person_hours - current_rework_person_hours
Unit	labor_hour
Required evidence	baseline_rework_person_hours; current_rework_person_hours; effort_evidence_basis; mature_defect_and_reopen_records; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_rework_person_hours, current_rework_person_hours, effort_evidence_basis, mature_defect_and_reopen_records, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show rework hours reduction with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Refund loss reduction

Contract: METRIC-ROI-07.refund_loss_reduction.

Attributable reduction in realized refunds for comparable accepted scope and maturity.

Property	Contract
Calculation	baseline_realized_refunds - current_realized_refunds
Unit	currency
Required evidence	baseline_realized_refunds; current_realized_refunds; refund_work_linkage; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_realized_refunds, current_realized_refunds, refund_work_linkage, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show refund loss reduction with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Defect rate change

Contract: METRIC-ROI-07.defect_rate_change.

Signed change in mature defect incidence for the same accepted-work definition.

Property	Contract
Calculation	current_defect_count / current_mature_eligible_outcomes - baseline_defect_count / baseline_mature_eligible_outcomes
Unit	ratio
Required evidence	current_defect_count; current_mature_eligible_outcomes; baseline_defect_count; baseline_mature_eligible_outcomes; comparison_contract
Denominator	reference: current_mature_eligible_outcomes and baseline_mature_eligible_outcomes; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_defect_count, current_mature_eligible_outcomes, baseline_defect_count, baseline_mature_eligible_outcomes, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Negative change is fewer defects. Follow defects and revocations after initial acceptance; do not ignore failed outcomes.

Repeat contact rate change

Contract: METRIC-ROI-07.repeat_contact_rate_change.

Signed change in repeat contacts for comparable cases over the declared follow-up window.

Property	Contract
Calculation	current_cases_with_repeat_contact / current_mature_cases - baseline_cases_with_repeat_contact / baseline_mature_cases
Unit	ratio
Required evidence	current_cases_with_repeat_contact; current_mature_cases; baseline_cases_with_repeat_contact; baseline_mature_cases; follow_up_window; comparison_contract
Denominator	reference: current_mature_cases and baseline_mature_cases; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_cases_with_repeat_contact, current_mature_cases, baseline_cases_with_repeat_contact, baseline_mature_cases, follow_up_window, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show repeat contact rate change with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Other realized error loss reduction

Contract: METRIC-ROI-07.other_realized_error_loss_reduction.

Attributable reduction in separately evidenced error losses not already counted as refunds or rework expense.

Property	Contract
Calculation	baseline_other_realized_error_losses - current_other_realized_error_losses
Unit	currency
Required evidence	baseline_other_realized_error_losses; current_other_realized_error_losses; loss_category_and_exclusion_map; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_other_realized_error_losses, current_other_realized_error_losses, loss_category_and_exclusion_map, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show other realized error loss reduction with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Net rework error remediation benefit

Contract: METRIC-ROI-07.net_rework_error_remediation_benefit.

Complete realized loss reduction and reduced rework expense less disjoint intervention cost.

Property	Contract
Calculation	sum(nonoverlapping_realized_loss_and_rework_expense_reductions) - disjoint_intervention_cost
Unit	currency
Required evidence	nonoverlapping_realized_loss_and_rework_expense_reductions; disjoint_intervention_cost; benefit_cost_inclusion_map; mature_downstream_failure_evidence
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: nonoverlapping_realized_loss_and_rework_expense_reductions, disjoint_intervention_cost, benefit_cost_inclusion_map, mature_downstream_failure_evidence. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Hours-equivalence is not realized expense reduction. Retain downstream failures and negative benefit; never subtract mapped intervention costs twice.

08. End-to-end cycle-time improvement

METRIC-ROI-08. Change in request-to-accepted-delivery time at P50/P95, holding task class and quality comparable. Price only an evidenced consequence such as avoided delay or earlier revenue.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Workflow boundaries + outcome; Cross-system work start/end, deadlines, handoffs and downstream outcomes; recorded or calibrated person-time for effort and labor-wait claims

Evidence limits: Separate elapsed time, queue time and person-time; price only evidenced business consequences; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Cycle time p50

Contract: METRIC-ROI-08.cycle_time_p50.

Median elapsed time from a supported work request to mature accepted delivery.

Property	Contract
Calculation	quantile(request_to_accepted_delivery_durations, 0.50)
Unit	hour
Required evidence	request_to_accepted_delivery_durations; cohort_acceptance_and_maturity_contract; lifecycle_boundary
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: request_to_accepted_delivery_durations, cohort_acceptance_and_maturity_contract, lifecycle_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the declared quantile from compatible underlying observations; never average cohort percentiles.
Display	Elapsed time includes supported waits and processing; it is not human labor time. Report eligible, pending and censored work separately.

Cycle time p95

Contract: METRIC-ROI-08.cycle_time_p95.

Ninety-fifth percentile elapsed time from request to mature accepted delivery.

Property	Contract
Calculation	quantile(request_to_accepted_delivery_durations, 0.95)
Unit	hour
Required evidence	request_to_accepted_delivery_durations; cohort_acceptance_and_maturity_contract; lifecycle_boundary
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: request_to_accepted_delivery_durations, cohort_acceptance_and_maturity_contract, lifecycle_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the declared quantile from compatible underlying observations; never average cohort percentiles.
Display	Show cycle time p95 with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cycle time p50 improvement

Contract: METRIC-ROI-08.cycle_time_p50_improvement.

Comparable decrease in median end-to-end accepted-delivery time.

Property	Contract
Calculation	baseline_cycle_time_p50 - current_cycle_time_p50
Unit	hour
Required evidence	baseline_cycle_time_p50; current_cycle_time_p50; comparison_contract; completion_censoring_policy
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_cycle_time_p50, current_cycle_time_p50, comparison_contract, completion_censoring_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the declared quantile from compatible underlying observations; never average cohort percentiles. Compute baseline-minus-current only after recomputing both quantiles.
Display	Show cycle time p50 improvement with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cycle time p95 improvement

Contract: METRIC-ROI-08.cycle_time_p95_improvement.

Comparable decrease in tail end-to-end accepted-delivery time.

Property	Contract
Calculation	baseline_cycle_time_p95 - current_cycle_time_p95
Unit	hour
Required evidence	baseline_cycle_time_p95; current_cycle_time_p95; comparison_contract; completion_censoring_policy
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_cycle_time_p95, current_cycle_time_p95, comparison_contract, completion_censoring_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the declared quantile from compatible underlying observations; never average cohort percentiles. Compute baseline-minus-current only after recomputing both quantiles.
Display	Show cycle time p95 improvement with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Realized cycle time consequence

Contract: METRIC-ROI-08.realized_cycle_time_consequence.

Attributable financial consequence of faster delivery, such as evidenced delay expense avoided or incremental contribution from earlier fulfillment.

Property	Contract
Calculation	sum(nonoverlapping_verified_cycle_time_financial_consequences) - disjoint_intervention_cost
Unit	currency
Required evidence	dated_delivery_consequence_records; comparison_contract; nonoverlapping_benefit_ledger; disjoint_intervention_cost
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: dated_delivery_consequence_records, comparison_contract, nonoverlapping_benefit_ledger, disjoint_intervention_cost. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Price an evidenced consequence; elapsed hours times a wage do not establish this output. Keep cash timing separate from recognized contribution.

09. Accepted automation without human intervention

METRIC-ROI-09. Unique accepted outcomes with no human intervention / eligible work items; compare total cost and quality with assisted and human-only paths. Count escalations and supervision separately.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Outcome + intervention records; Defects/reopens, review/intervention, failed attempts and linked terminal outcomes

Evidence limits: Reconstruct accepted automation, rework and rescue; no-human-intervention claims require coverage of the intervention path; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Accepted no human intervention count

Contract: METRIC-ROI-09.accepted_no_human_intervention_count.

Unique mature accepted outcomes with complete evidence that the delivery path required no human intervention.

Property	Contract
Calculation	count_unique(mature_accepted_outcomes_with_complete_no_intervention_path)
Unit	accepted_outcome
Required evidence	accepted_outcome_records; complete_human_intervention_path_coverage; cohort_acceptance_and_maturity_contract
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: accepted_outcome_records, complete_human_intervention_path_coverage, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Unknown intervention coverage is not zero intervention. Supervision, review and escalation must be recorded separately.

Accepted no human intervention rate

Contract: METRIC-ROI-09.accepted_no_human_intervention_rate.

Accepted outcomes with no human intervention divided by all eligible work items.

Property	Contract
Calculation	accepted_no_human_intervention_count / eligible_work_item_count
Unit	ratio
Required evidence	accepted_no_human_intervention_count; eligible_work_item_count; eligibility_contract
Denominator	reference: eligible_work_item_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: accepted_no_human_intervention_count, eligible_work_item_count, eligibility_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Use all eligible work, including failed, rejected and abandoned paths. Do not use only successful automated cases as denominator.

Human escalation count

Contract: METRIC-ROI-09.human_escalation_count.

Unique eligible work items requiring human escalation.

Property	Contract
Calculation	count_unique(work_items_with_evidenced_human_escalation)
Unit	work_item
Required evidence	human_escalation_records; eligible_work_ids; intervention_path_coverage
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: human_escalation_records, eligible_work_ids, intervention_path_coverage. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show human escalation count with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Human supervision hours

Contract: METRIC-ROI-09.human_supervision_hours.

Recorded human supervision effort consumed by the evaluated automation cohort.

Property	Contract
Calculation	sum(nonoverlapping_recorded_supervision_person_hours)
Unit	labor_hour
Required evidence	recorded_supervision_effort; person_time_overlap_resolution; workflow_labor_boundary
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: recorded_supervision_effort, person_time_overlap_resolution, workflow_labor_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Retain supervision even when the final outcome is classified as automated. If inferred, publish an inferred-effort result rather than relabeling it recorded.

Path cost per accepted outcome

Contract: METRIC-ROI-09.path_cost_per_accepted_outcome.

Complete accepted unit cost separately for automated, assisted and human-only delivery paths.

Property	Contract
Calculation	complete_path_cost / unique_mature_accepted_outcomes_for_delivery_path
Unit	currency/accepted_outcome
Required evidence	delivery_path; complete_path_cost; unique_mature_accepted_outcomes_for_delivery_path; path_cost_completeness; cohort_acceptance_and_maturity_contract
Denominator	reference: unique_mature_accepted_outcomes_for_delivery_path; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: delivery_path, complete_path_cost, unique_mature_accepted_outcomes_for_delivery_path, path_cost_completeness, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show automated, assisted and human-only paths with the same full-cost boundary; include unsuccessful attempts and supervision.

Path quality rate

Contract: METRIC-ROI-09.path_quality_rate.

Mature accepted outcome yield for each automated, assisted or human-only path.

Property	Contract
Calculation	unique_mature_accepted_outcomes_for_delivery_path / eligible_work_items_for_delivery_path
Unit	ratio
Required evidence	delivery_path; unique_mature_accepted_outcomes_for_delivery_path; eligible_work_items_for_delivery_path; downstream_quality_maturity
Denominator	reference: eligible_work_items_for_delivery_path; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: delivery_path, unique_mature_accepted_outcomes_for_delivery_path, eligible_work_items_for_delivery_path, downstream_quality_maturity. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Separate path quality and cost. Descriptive path differences may reflect selection and do not establish causal automation benefit.

AI output override or abandonment rate

Contract: METRIC-ROI-09.ai_output_override_or_abandonment_rate.

Share of mature eligible AI output decision units with observable terminal actions that have an evidenced qualifying override or abandonment.

Property	Contract
Calculation	count_unique(eligible_mature_terminal_output_units_with_qualifying_override_or_abandonment) / count_unique(eligible_mature_terminal_output_units)
Unit	ratio
Required evidence	output_population_and_grain_contract; versioned_override_and_abandonment_policy; observation_maturity_and_source_coverage; result_source_provenance
Alternative evidence	mode: source_event_records; required inputs: eligible_ai_output_population; output_identity_and_lineage; terminal_action_records; human_action_or_revision_metadata; condition: The output lifecycle is supported by existing metadata, terminal classification coverage is complete for the declared cohort, and the union is deduplicated by tenant and output decision identity.; mode: source_native_aggregate; required inputs: source_union_numerator; source_terminal_output_denominator; aggregate_population_and_overlap_evidence; condition: Existing aggregate records establish compatible output grain, deduplication, union overlap, policy, window and classification coverage; separate category counts without overlap evidence are insufficient.
Input definitions	eligible mature terminal output unit count: Distinct eligible AI outputs first delivered during the reporting period, with observable terminal actions and mature versioned windows. Retries, receipts, refreshes and human revisions do not add output units.; qualifying override or abandonment: Union of positively evidenced human override, manual reversion, supported heavy-edit classification and explicit reconciled terminal abandonment. Acceptance after an override does not erase that action. Automated expiry alone does not establish human intent.; heavy edit evidence: Existing source-provided magnitude or explicit disposition with versioned unit, original AI revision, comparison method and threshold; no raw-content collection or substitution by text-length difference, edit count or elapsed time.; result source provenance: Source event, observation and data-as-of timestamps, source revisions and late-arrival/correction lineage for the scoped result.
Denominator	reference: eligible_mature_terminal_output_unit_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, output grain, first-delivery cohort, policy, maturity horizon and terminal-action population; classification-unknown terminal units cannot silently leave the denominator.
Evidence class	observed_operational
Baseline	No baseline for the observed level. Comparisons require compatible eligibility, action policy, maturity and source coverage.
Causal interpretation	Descriptive behavior only. The result does not establish dissatisfaction, poor quality, wasted spend or an AI-attributable economic effect.
Missing evidence	Return unavailable with a specific reason for missing required evidence, incomplete terminal classification, unknown category overlap or a zero denominator; retain immature windows and supported counts. Inactivity is not abandonment, missing events are not zero override, and external benchmarks cannot supply dispositions.
Aggregation	Union qualifying override and abandonment identities once within the mature eligible terminal cohort; deduplicate source replicas and receipts. Reaggregate compatible numerator and denominator populations, never average ratios or pool incompatible policies.
Display	Show the observed-terminal cohort scope, numerator and denominator counts, eligible/pending/unobserved-terminal/classification-unknown counts, output grain, policy version, observation horizon, timestamps, source revision and coverage. Preserve unknowns and historical as-of revisions.

Human review and correction minutes per accepted output

Contract: METRIC-ROI-09.human_review_correction_minutes_per_accepted_output.

Include human review and correction effort across all attempts in the same output cohort, including rejected and abandoned work. Union overlapping intervals separately for each person across roles and sources, then sum person-time across people; simultaneous effort by different people adds. Retain recorded versus calibrated inferred-active effort as separate bases; the inferred_effort result class never promotes calibrated time to an observation, and divide by mature accepted outputs.

Property	Contract
Calculation	complete_nonoverlapping_review_and_correction_person_minutes / mature_accepted_output_count
Unit	person_minute_per_accepted_output
Required evidence	review_and_correction_effort_evidence; complete_nonoverlapping_review_and_correction_person_minutes; mature_accepted_output_count; matched_output_cohort_and_effort_basis
Denominator	reference: mature_accepted_output_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	inferred_effort
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing complete effort or mature accepted count withholds the ratio. Retain supported review/remediation hours and observed accepted counts; never infer human active effort from elapsed workflow time alone.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

10. Decision quality and realized decision value

METRIC-ROI-10. Change in realized utility or loss from AI-assisted forecasts, pricing, procurement, fraud decisions or resource allocation. Compare decision policies using existing experiment or comparable historical records; predictive accuracy is a diagnostic until consequences are observed.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Decision/action records + outcome + comparison; Existing decision/experiment/configuration histories, outcomes and economic consequences

Evidence limits: Compare policies or budget levels under supported identification; otherwise report descriptive association or a scenario; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Realized decision utility change

Contract: METRIC-ROI-10.realized_decision_utility_change.

Change in realized decision utility under predeclared validated consequences and utility weights.

Property	Contract
Calculation	mean(current_realized_decision_utility) - mean(comparable_baseline_realized_decision_utility)
Unit	utility_unit/decision
Required evidence	decision_policy_versions; decision_action_outcome_links; predeclared_validated_utility_function; comparison_contract; mature_decision_outcomes
Denominator	reference: current_mature_decision_count and baseline_mature_decision_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: decision_policy_versions, decision_action_outcome_links, predeclared_validated_utility_function, comparison_contract, mature_decision_outcomes. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute mean realized utility from compatible decision cohorts under one validated utility function before comparison; retain the full eligible cohort.
Display	Show realized decision utility change with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Realized decision loss reduction

Contract: METRIC-ROI-10.realized_decision_loss_reduction.

Attributable reduction in actual financial losses from comparable decision policies.

Property	Contract
Calculation	baseline_realized_decision_losses - current_realized_decision_losses - disjoint_policy_intervention_cost
Unit	currency
Required evidence	realized_loss_records; decision_action_outcome_links; comparison_contract; disjoint_policy_intervention_cost
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: realized_loss_records, decision_action_outcome_links, comparison_contract, disjoint_policy_intervention_cost. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show realized decision loss reduction with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Predictive accuracy

Contract: METRIC-ROI-10.predictive_accuracy.

Diagnostic prediction accuracy against independently observed decision labels under a declared scoring rule.

Property	Contract
Calculation	declared_scoring_function(predictions, independently_observed_outcome_labels)
Unit	score
Required evidence	frozen_predictions; independent_outcome_labels; scoring_rule; prediction_horizon
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: frozen_predictions, independent_outcome_labels, scoring_rule, prediction_horizon. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the declared score over compatible labeled observations; do not combine different scores.
Display	Predictive accuracy is diagnostic until consequences are observed; no automatic conversion to realized decision value.

11. Working-capital and cash-recovery value

METRIC-ROI-11. Incremental collectible cash recovered and financing cost avoided through faster collections, fewer disputes or better inventory decisions, net of intervention cost. Separate timing, recovered principal and profit.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Ledger/payment/inventory records + baseline; Finance transactions, declared investment scope, comparable histories and benefit dates

Evidence limits: Reconcile non-overlapping financial changes, net benefit and cash payback; preserve operational value when a monetary consequence is unsupported; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Incremental collectible cash recovered

Contract: METRIC-ROI-11.incremental_collectible_cash_recovered.

Incremental principal actually collected from supported recoverable balances versus the declared comparator.

Property	Contract
Calculation	actual_collected_principal - comparable_expected_collected_principal
Unit	currency
Required evidence	receivable_or_recovery_ledger; dated_payment_records; collectibility_evidence; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: receivable_or_recovery_ledger, dated_payment_records, collectibility_evidence, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Recovered principal is cash recovery, not automatically revenue or profit. Separate collection amount from payment timing.

Cash collection time improvement

Contract: METRIC-ROI-11.cash_collection_time_improvement.

Reduction in time from the defined receivable milestone to matched cash collection.

Property	Contract
Calculation	baseline_collection_elapsed_time - current_collection_elapsed_time
Unit	day
Required evidence	receivable_milestone_dates; matched_cash_dates; comparison_contract; collection_timing_statistic
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: receivable_milestone_dates, matched_cash_dates, comparison_contract, collection_timing_statistic. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the declared collection-time statistic over comparable balances; retain censoring and unpaid balances.
Display	Show cash collection time improvement with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Financing cost avoided

Contract: METRIC-ROI-11.financing_cost_avoided.

Evidenced financing expense reduced by earlier collection or inventory release.

Property	Contract
Calculation	counterfactual_financing_expense_for_supported_balance_timing - actual_financing_expense
Unit	currency
Required evidence	balance_timing_histories; applicable_financing_contract_or_rate; actual_financing_expense; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: balance_timing_histories, applicable_financing_contract_or_rate, actual_financing_expense, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	An assumed internal cost of capital is a scenario, not realized expense. Preserve balance, days, rate and cash-flow boundaries.

Net cash recovery benefit

Contract: METRIC-ROI-11.net_cash_recovery_benefit.

Incremental recovered cash and disjoint realized financing savings less intervention cash costs, explicitly scoped as cash.

Property	Contract
Calculation	incremental_collectible_cash_recovered + nonoverlapping_financing_cost_avoided - disjoint_recovery_intervention_cash_cost
Unit	currency
Required evidence	incremental_collectible_cash_recovered; nonoverlapping_financing_cost_avoided; disjoint_recovery_intervention_cash_cost; cash_benefit_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incremental_collectible_cash_recovered, nonoverlapping_financing_cost_avoided, disjoint_recovery_intervention_cash_cost, cash_benefit_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Label cash-recovery scope. Recovered principal, timing benefit and contribution must remain separate ledger categories; do not relabel this total as profit.

Working capital financing scenario

Contract: METRIC-ROI-11.working_capital_financing_scenario.

Modeled financing value of a supported working-capital timing change under an explicit rate assumption.

Property	Contract
Calculation	sum(balance_amount * financing_days_change / declared_day_count_basis * assumed_financing_rate for nonoverlapping balances in working_capital_timing_change) - scenario_intervention_cost
Unit	currency
Required evidence	working_capital_timing_change; assumed_financing_rate; day_count_basis; scenario_intervention_cost
Denominator	reference: declared_day_count_basis; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Finite positive declared rate day-count convention; align balance currency and applicable time basis.
Evidence class	modeled_scenario
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: working_capital_timing_change, assumed_financing_rate, day_count_basis, scenario_intervention_cost. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Calculate financing value per non-overlapping balance/time/rate scope under its declared day-count convention, sum compatible currency values, and subtract each intervention cost once.
Display	Label modeled financing value and its rate; never combine it with realized financing savings for the same balance.

12. Marginal return on additional AI budget

METRIC-ROI-12. Change in attributable benefit less change in cost, per additional dollar of cost, estimated from existing controlled budget/service levels or a credible naturally varying comparison. Use this to scale, constrain or retire a workflow.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Finance + existing experiment or identified comparison; Existing decision/experiment/configuration histories, outcomes and economic consequences

Evidence limits: Compare policies or budget levels under supported identification; otherwise report descriptive association or a scenario; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Incremental budget cost change

Contract: METRIC-ROI-12.incremental_budget_cost_change.

Change in complete AI-related cost between supported budget or service levels.

Property	Contract
Calculation	cost_at_higher_budget_level - cost_at_reference_budget_level
Unit	currency
Required evidence	complete_cost_by_budget_or_service_level; level_definition; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: complete_cost_by_budget_or_service_level, level_definition, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show incremental budget cost change with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Incremental budget net benefit change

Contract: METRIC-ROI-12.incremental_budget_net_benefit_change.

Change in attributable benefit less the change in complete cost across supported budget levels.

Property	Contract
Calculation	attributable_benefit_change - incremental_budget_cost_change
Unit	currency
Required evidence	attributable_benefit_by_budget_or_service_level; incremental_budget_cost_change; identified_budget_comparison
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: attributable_benefit_by_budget_or_service_level, incremental_budget_cost_change, identified_budget_comparison. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show incremental budget net benefit change with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Marginal return on additional budget

Contract: METRIC-ROI-12.marginal_return_on_additional_budget.

Incremental net benefit per additional positive dollar of complete cost.

Property	Contract
Calculation	incremental_budget_net_benefit_change / incremental_budget_cost_change
Unit	ratio
Required evidence	incremental_budget_net_benefit_change; incremental_budget_cost_change; identified_budget_comparison
Denominator	reference: incremental_budget_cost_change; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incremental_budget_net_benefit_change, incremental_budget_cost_change, identified_budget_comparison. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Publish the evaluated budget interval, uncertainty and negative returns. Unsupported natural variation yields association or a separately labeled scenario, not a causal marginal return.

13. Quality-preserving model and routing efficiency

METRIC-ROI-13. Cost per accepted outcome for the current route minus the lowest-cost tested route meeting the same acceptance, latency, security and residency requirements.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Usage + outcome + comparison; Existing route/variant/retrieval/tool/evaluation records, full cost and acceptance

Evidence limits: Compare observed variants; unobserved routes or ablations require separately identified readable evaluation evidence for semantic claims; metadata-only cost alternatives stay price simulations; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Current route cost per accepted outcome

Contract: METRIC-ROI-13.current_route_cost_per_accepted_outcome.

Complete unit cost for the current route within the defined accepted workload.

Property	Contract
Calculation	current_route_complete_cost / current_route_mature_accepted_outcomes
Unit	currency/accepted_outcome
Required evidence	current_route_complete_cost; current_route_mature_accepted_outcomes; route_contract; cohort_acceptance_and_maturity_contract
Denominator	reference: current_route_mature_accepted_outcomes; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_route_complete_cost, current_route_mature_accepted_outcomes, route_contract, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show current route cost per accepted outcome with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Lowest tested qualified route unit cost

Contract: METRIC-ROI-13.lowest_tested_qualified_route_unit_cost.

Lowest complete unit cost among tested routes meeting the same acceptance, latency, security and residency requirements.

Property	Contract
Calculation	min(complete_cost_route / mature_accepted_outcomes_route for tested routes satisfying all declared requirements)
Unit	currency/accepted_outcome
Required evidence	readable_route_evaluation_results; complete_tested_route_costs; tested_route_mature_accepted_outcomes; acceptance_latency_security_residency_requirements; comparison_contract
Denominator	reference: tested_route_mature_accepted_outcomes for each qualifying route; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: readable_route_evaluation_results, complete_tested_route_costs, tested_route_mature_accepted_outcomes, acceptance_latency_security_residency_requirements, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute eligible route costs within the same workload; select only among tested qualifying routes, never average minima.
Display	Identify tested route set and evaluation provenance. Metadata-only repricing does not establish a qualified route.

Quality preserving routing cost improvement

Contract: METRIC-ROI-13.quality_preserving_routing_cost_improvement.

Signed cost reduction available from the lowest-cost tested route meeting unchanged requirements.

Property	Contract
Calculation	current_route_cost_per_accepted_outcome - lowest_tested_qualified_route_unit_cost
Unit	currency/accepted_outcome
Required evidence	current_route_cost_per_accepted_outcome; lowest_tested_qualified_route_unit_cost; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_route_cost_per_accepted_outcome, lowest_tested_qualified_route_unit_cost, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	A tested alternative’s unit-cost difference is an efficiency comparison. Realized savings require an observed adopted route and financial evidence.

Route price only scenario

Contract: METRIC-ROI-13.route_price_only_scenario.

Hypothetical price change from repricing observed usage under an explicitly selected alternative rate contract.

Property	Contract
Calculation	observed_usage_cost_at_current_rates - modeled_same_usage_cost_at_alternative_rates
Unit	currency
Required evidence	compatible_usage_components; current_rate_version; alternative_rate_version; scenario_usage_assumptions
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Hold modeled usage explicitly fixed; expose incompatibility between provider/model token or billing definitions.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: compatible_usage_components, current_rate_version, alternative_rate_version, scenario_usage_assumptions. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Label price-only simulation. It provides no evidence of alternative quality, latency, security, residency or semantic equivalence.

14. Failure, retry, abandonment and duplicate-work cost

METRIC-ROI-14. Cost of terminally failed, rejected, abandoned, duplicated and unnecessary retry chains, with disjoint classifications. A failed call that leads to a useful fallback is not automatically waste.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Attempt chains + terminal outcomes; Paid-attempt chains and terminal outcomes; cache eligibility/retention plus read/write/storage charges; batch/service-tier usage, deadlines and acceptance; applicable rate and contract versions

Evidence limits: Reconcile actual costs and comparable price/cost alternatives; hypothetical route/latency/quality effects stay modeled; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Terminal failed work cost

Contract: METRIC-ROI-14.terminal_failed_work_cost.

Paid cost uniquely assigned to terminally failed work chains.

Property	Contract
Calculation	sum(unique_paid_costs_classified_terminal_failed)
Unit	currency
Required evidence	billable_attempt_lineage; terminal_outcomes; disjoint_waste_classification
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: billable_attempt_lineage, terminal_outcomes, disjoint_waste_classification. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	A failed call followed by an accepted fallback is not automatically terminal waste. Preserve useful rescue cost separately.

Rejected work cost

Contract: METRIC-ROI-14.rejected_work_cost.

Paid cost uniquely assigned to terminally rejected work under the acceptance contract.

Property	Contract
Calculation	sum(unique_paid_costs_classified_rejected)
Unit	currency
Required evidence	billable_attempt_lineage; terminal_rejection_evidence; disjoint_waste_classification
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: billable_attempt_lineage, terminal_rejection_evidence, disjoint_waste_classification. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show rejected work cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Abandoned work cost

Contract: METRIC-ROI-14.abandoned_work_cost.

Paid cost uniquely assigned to abandoned work after the declared maturity rule.

Property	Contract
Calculation	sum(unique_paid_costs_classified_abandoned)
Unit	currency
Required evidence	billable_attempt_lineage; abandonment_and_maturity_evidence; disjoint_waste_classification
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: billable_attempt_lineage, abandonment_and_maturity_evidence, disjoint_waste_classification. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show abandoned work cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Duplicate work cost

Contract: METRIC-ROI-14.duplicate_work_cost.

Cost of evidenced duplicate business work under a fixed outcome and lineage definition.

Property	Contract
Calculation	sum(unique_paid_costs_classified_duplicate)
Unit	currency
Required evidence	billable_attempt_lineage; duplicate_business_work_evidence; disjoint_waste_classification
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: billable_attempt_lineage, duplicate_business_work_evidence, disjoint_waste_classification. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Repeated model calls are not automatically duplicate business work. Charge every billable leaf once and preserve necessary retries.

Unnecessary retry cost

Contract: METRIC-ROI-14.unnecessary_retry_cost.

Cost of retries independently classified as unnecessary within an evaluated attempt chain.

Property	Contract
Calculation	sum(unique_paid_costs_classified_unnecessary_retry)
Unit	currency
Required evidence	billable_attempt_lineage; retry_necessity_evidence; disjoint_waste_classification
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: billable_attempt_lineage, retry_necessity_evidence, disjoint_waste_classification. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show unnecessary retry cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Classified unproductive work cost

Contract: METRIC-ROI-14.classified_unproductive_work_cost.

Union of disjoint terminal-failure, rejection, abandonment, duplicate and unnecessary-retry costs.

Property	Contract
Calculation	sum(unique_costs_in_disjoint_unproductive_work_categories)
Unit	currency
Required evidence	terminal_failed_work_cost; rejected_work_cost; abandoned_work_cost; duplicate_work_cost; unnecessary_retry_cost; category_exclusivity_evidence
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: terminal_failed_work_cost, rejected_work_cost, abandoned_work_cost, duplicate_work_cost, unnecessary_retry_cost, category_exclusivity_evidence. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Do not require unavailable categories for separately supported category outputs. A full union requires complete classification; retain unresolved work separately.

Classified unproductive spend share

Contract: METRIC-ROI-14.classified_unproductive_spend_share.

Disjoint classified unproductive work cost divided by complete comparable paid work cost.

Property	Contract
Calculation	classified_unproductive_work_cost / complete_cohort_paid_work_cost
Unit	ratio
Required evidence	classified_unproductive_work_cost; complete_cohort_paid_work_cost; category_exclusivity_evidence
Denominator	reference: complete_cohort_paid_work_cost; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: classified_unproductive_work_cost, complete_cohort_paid_work_cost, category_exclusivity_evidence. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show classified unproductive spend share with its coverage classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Avoidable workflow cost scenario

Contract: METRIC-ROI-14.avoidable_workflow_cost_scenario.

Estimate the signed cost difference for an evidenced prevention scenario that preserves comparable accepted work. Retain paid retries, failures and duplicate classifications once; classified waste is not automatically avoidable.

Property	Contract
Calculation	complete_observed_workflow_cost - supported_prevention_scenario_cost - disjoint_prevention_cost
Unit	USD
Required evidence	complete_observed_workflow_cost; supported_prevention_scenario_cost; disjoint_prevention_cost; comparable_accepted_work_evidence; versioned_prevention_scenario
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Absent prevention or acceptance-preservation evidence withholds this scenario, retaining supported failed, abandoned, duplicate and retry cost observations.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

15. Net prompt-caching benefit

METRIC-ROI-15. Total comparable uncached execution cost minus total actual execution cost with caching, including residual uncached input, output, cache reads/writes, storage, refresh and overhead. Match model, prompt eligibility, time-to-live and quality; do not add cache tokens twice.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Provider cache usage + effective prices; Paid-attempt chains and terminal outcomes; cache eligibility/retention plus read/write/storage charges; batch/service-tier usage, deadlines and acceptance; applicable rate and contract versions

Evidence limits: Reconcile actual costs and comparable price/cost alternatives; hypothetical route/latency/quality effects stay modeled; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Comparable uncached execution cost

Contract: METRIC-ROI-15.comparable_uncached_execution_cost.

Complete counterfactual execution cost for the same cache-eligible workload without caching.

Property	Contract
Calculation	sum(comparable_uncached_input_cost + comparable_output_cost + applicable_uncached_execution_overhead)
Unit	currency
Required evidence	cache_eligible_workload; comparable_uncached_usage; effective_rate_versions; model_prompt_ttl_quality_comparability; uncached_overhead
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Match model, prompt eligibility, retention period and quality.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cache_eligible_workload, comparable_uncached_usage, effective_rate_versions, model_prompt_ttl_quality_comparability, uncached_overhead. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show comparable uncached execution cost with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cache read cost

Contract: METRIC-ROI-15.cache_read_cost.

Observed cache-read charges within the declared workload.

Property	Contract
Calculation	sum(unique_cache_read_charges)
Unit	currency
Required evidence	cache_read_usage; effective_cache_read_rates; billing_reconciliation
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cache_read_usage, effective_cache_read_rates, billing_reconciliation. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show cache read cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cache write cost

Contract: METRIC-ROI-15.cache_write_cost.

Observed cache-write charges within the declared workload.

Property	Contract
Calculation	sum(unique_cache_write_charges)
Unit	currency
Required evidence	cache_write_usage; effective_cache_write_rates; billing_reconciliation
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cache_write_usage, effective_cache_write_rates, billing_reconciliation. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show cache write cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cache storage cost

Contract: METRIC-ROI-15.cache_storage_cost.

Cache storage expense allocated under the declared boundary.

Property	Contract
Calculation	sum(cache_storage_charges_allocated_once)
Unit	currency
Required evidence	cache_storage_charges; storage_allocation_basis
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cache_storage_charges, storage_allocation_basis. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show cache storage cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cache refresh cost

Contract: METRIC-ROI-15.cache_refresh_cost.

Cache refresh costs not already included in writes or generation.

Property	Contract
Calculation	sum(disjoint_cache_refresh_charges)
Unit	currency
Required evidence	refresh_event_costs; refresh_write_overlap_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: refresh_event_costs, refresh_write_overlap_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show cache refresh cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Residual uncached input cost

Contract: METRIC-ROI-15.residual_uncached_input_cost.

Cost of input usage not billed as cache reads or writes in cached execution.

Property	Contract
Calculation	sum(unique_residual_uncached_input_charges)
Unit	currency
Required evidence	residual_uncached_input_usage; effective_input_rates; cache_billing_component_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: residual_uncached_input_usage, effective_input_rates, cache_billing_component_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show residual uncached input cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cached execution output cost

Contract: METRIC-ROI-15.cached_execution_output_cost.

Output generation charges incurred while using caching.

Property	Contract
Calculation	sum(unique_output_generation_charges_for_cached_execution)
Unit	currency
Required evidence	output_generation_usage; effective_output_rates; cached_execution_scope
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: output_generation_usage, effective_output_rates, cached_execution_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show cached execution output cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Caching overhead cost

Contract: METRIC-ROI-15.caching_overhead_cost.

Disjoint operating and support costs of caching not captured in other billing components.

Property	Contract
Calculation	sum(disjoint_caching_operating_and_support_costs)
Unit	currency
Required evidence	caching_overhead_costs; component_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: caching_overhead_costs, component_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show caching overhead cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Actual execution cost with caching

Contract: METRIC-ROI-15.actual_execution_cost_with_caching.

Complete actual execution cost including every disjoint cache and residual generation component.

Property	Contract
Calculation	cache_read_cost + cache_write_cost + cache_storage_cost + cache_refresh_cost + residual_uncached_input_cost + cached_execution_output_cost + caching_overhead_cost
Unit	currency
Required evidence	cache_read_cost; cache_write_cost; cache_storage_cost; cache_refresh_cost; residual_uncached_input_cost; cached_execution_output_cost; caching_overhead_cost; cost_component_completeness
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cache_read_cost, cache_write_cost, cache_storage_cost, cache_refresh_cost, residual_uncached_input_cost, cached_execution_output_cost, caching_overhead_cost, cost_component_completeness. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	No cache token or refresh cost may be counted twice. A missing component leaves this total unavailable without suppressing observed components.

Net prompt caching benefit

Contract: METRIC-ROI-15.net_prompt_caching_benefit.

Signed complete cost difference between comparable uncached execution and actual cached execution.

Property	Contract
Calculation	comparable_uncached_execution_cost - actual_execution_cost_with_caching
Unit	currency
Required evidence	comparable_uncached_execution_cost; actual_execution_cost_with_caching; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Match model, cache eligibility, retention and quality.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: comparable_uncached_execution_cost, actual_execution_cost_with_caching, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Negative values mean caching increased total cost. Do not add this explanation again to a portfolio cost saving already including it.

16. Agent orchestration efficiency

METRIC-ROI-16. Accepted task cost and acceptance yield versus model calls, tool calls, reasoning usage, loops and handoffs per task. Identify redundant steps using existing variant/ablation evidence or readable results from an identified evaluation source; label counterfactual estimates separately. Call count or metadata-only repricing cannot establish semantic redundancy.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Linked run/attempt/tool events + outcomes; Existing route/variant/retrieval/tool/evaluation records, full cost and acceptance

Evidence limits: Compare observed variants; unobserved routes or ablations require separately identified readable evaluation evidence for semantic claims; metadata-only cost alternatives stay price simulations; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Orchestrated task cost per accepted outcome

Contract: METRIC-ROI-16.orchestrated_task_cost_per_accepted_outcome.

Complete orchestration cohort cost per unique accepted task.

Property	Contract
Calculation	complete_orchestration_cohort_cost / unique_mature_accepted_task_count
Unit	currency/accepted_outcome
Required evidence	complete_orchestration_cohort_cost; unique_mature_accepted_task_count; parent_child_cost_reconciliation; cohort_acceptance_and_maturity_contract
Denominator	reference: unique_mature_accepted_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: complete_orchestration_cohort_cost, unique_mature_accepted_task_count, parent_child_cost_reconciliation, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show orchestrated task cost per accepted outcome with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Orchestrated acceptance yield

Contract: METRIC-ROI-16.orchestrated_acceptance_yield.

Mature accepted business tasks divided by all eligible orchestrated tasks.

Property	Contract
Calculation	unique_mature_accepted_task_count / eligible_orchestrated_task_count
Unit	ratio
Required evidence	unique_mature_accepted_task_count; eligible_orchestrated_task_count; cohort_acceptance_and_maturity_contract
Denominator	reference: eligible_orchestrated_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: unique_mature_accepted_task_count, eligible_orchestrated_task_count, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show orchestrated acceptance yield with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Model calls per task

Contract: METRIC-ROI-16.model_calls_per_task.

Diagnostic orchestration resource consumption per eligible task, including unsuccessful task paths.

Property	Contract
Calculation	model_call_count / eligible_orchestrated_task_count
Unit	call/task
Required evidence	model_call_count; eligible_orchestrated_task_count; linked_attempt_tool_run_lineage; usage_unit_definition
Denominator	reference: eligible_orchestrated_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: model_call_count, eligible_orchestrated_task_count, linked_attempt_tool_run_lineage, usage_unit_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Compare only compatible model/workload and usage definitions. Resource counts alone do not establish semantic redundancy or business value.

Tool calls per task

Contract: METRIC-ROI-16.tool_calls_per_task.

Diagnostic orchestration resource consumption per eligible task, including unsuccessful task paths.

Property	Contract
Calculation	tool_call_count / eligible_orchestrated_task_count
Unit	call/task
Required evidence	tool_call_count; eligible_orchestrated_task_count; linked_attempt_tool_run_lineage; usage_unit_definition
Denominator	reference: eligible_orchestrated_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: tool_call_count, eligible_orchestrated_task_count, linked_attempt_tool_run_lineage, usage_unit_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Compare only compatible model/workload and usage definitions. Resource counts alone do not establish semantic redundancy or business value.

Reasoning usage per task

Contract: METRIC-ROI-16.reasoning_usage_per_task.

Diagnostic orchestration resource consumption per eligible task, including unsuccessful task paths.

Property	Contract
Calculation	reasoning_usage_quantity / eligible_orchestrated_task_count
Unit	declared_reasoning_unit/task
Required evidence	reasoning_usage_quantity; eligible_orchestrated_task_count; linked_attempt_tool_run_lineage; usage_unit_definition
Denominator	reference: eligible_orchestrated_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: reasoning_usage_quantity, eligible_orchestrated_task_count, linked_attempt_tool_run_lineage, usage_unit_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Compare only compatible model/workload and usage definitions. Resource counts alone do not establish semantic redundancy or business value.

Loops per task

Contract: METRIC-ROI-16.loops_per_task.

Diagnostic orchestration resource consumption per eligible task, including unsuccessful task paths.

Property	Contract
Calculation	orchestration_loop_count / eligible_orchestrated_task_count
Unit	loop/task
Required evidence	orchestration_loop_count; eligible_orchestrated_task_count; linked_attempt_tool_run_lineage; usage_unit_definition
Denominator	reference: eligible_orchestrated_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: orchestration_loop_count, eligible_orchestrated_task_count, linked_attempt_tool_run_lineage, usage_unit_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Compare only compatible model/workload and usage definitions. Resource counts alone do not establish semantic redundancy or business value.

Handoffs per task

Contract: METRIC-ROI-16.handoffs_per_task.

Diagnostic orchestration resource consumption per eligible task, including unsuccessful task paths.

Property	Contract
Calculation	orchestration_handoff_count / eligible_orchestrated_task_count
Unit	handoff/task
Required evidence	orchestration_handoff_count; eligible_orchestrated_task_count; linked_attempt_tool_run_lineage; usage_unit_definition
Denominator	reference: eligible_orchestrated_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: orchestration_handoff_count, eligible_orchestrated_task_count, linked_attempt_tool_run_lineage, usage_unit_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Compare only compatible model/workload and usage definitions. Resource counts alone do not establish semantic redundancy or business value.

Evaluated redundant step cost

Contract: METRIC-ROI-16.evaluated_redundant_step_cost.

Estimated avoidable cost of steps supported as redundant by existing readable variant or ablation evaluation results.

Property	Contract
Calculation	complete_cost_of_evaluated_original_path - complete_cost_of_evaluated_qualified_reduced_path
Unit	currency
Required evidence	readable_variant_or_ablation_results; original_and_reduced_path_complete_costs; acceptance_quality_and_constraint_parity; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: readable_variant_or_ablation_results, original_and_reduced_path_complete_costs, acceptance_quality_and_constraint_parity, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Require semantic evaluation evidence for redundancy. Metadata-only repricing remains a price scenario; never infer successful ablation from call count.

Nominal-path call amplification

Contract: METRIC-ROI-16.nominal_path_call_amplification.

Count all model and tool call occurrences, including retries, fallbacks and loop iterations, once under the same declared run. Compare with the versioned nominal path for the same work class; nested tracing spans do not create calls.

Property	Contract
Calculation	observed_model_and_tool_call_count / declared_nominal_path_call_count
Unit	factor
Required evidence	complete_run_attempt_call_metadata; observed_model_and_tool_call_count; declared_nominal_path_call_count; versioned_workflow_and_call_grain
Denominator	reference: declared_nominal_path_call_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	An undeclared, nonpositive or incompatible nominal path withholds amplification while observed calls and supported paid costs remain visible. Values below one remain signed evidence of a shorter path, not automatic success.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Observed amplification call cost

Contract: METRIC-ROI-16.observed_amplification_call_cost.

Sum supported cost of actual calls positively aligned as additional to the declared nominal path, deduplicating billable leaves and overlapping retry, fallback and loop classifications. Preserve useful rescue calls and paid failures. This observed amount is not a counterfactual saving, an ablation difference or automatically avoidable cost.

Property	Contract
Calculation	sum_unique(supported_billable_leaf_cost_of_calls_aligned_as_additional_to_nominal_path)
Unit	USD
Required evidence	actual_to_nominal_call_path_mapping; versioned_nominal_path; unique_billable_leaf_cost_records; compatible_currency_cost_basis_and_run_scope; retry_fallback_loop_classification
Denominator	Not applicable
Evidence class	observed_financial
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing actual-to-nominal alignment, complete supported extra-call cost, or compatible run scope withholds the full amount. Preserve supported total workflow cost and amplification factor independently; an unavailable alignment is not zero extra cost.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

17. Context and generation efficiency

METRIC-ROI-17. Accepted outcome cost before and after context selection, compression, output-length or reasoning changes. Report token quantities as diagnostics within the same model and workload.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Usage + outcome + comparison; Existing route/variant/retrieval/tool/evaluation records, full cost and acceptance

Evidence limits: Compare observed variants; unobserved routes or ablations require separately identified readable evaluation evidence for semantic claims; metadata-only cost alternatives stay price simulations; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Accepted cost change after context generation change

Contract: METRIC-ROI-17.accepted_cost_change_after_context_generation_change.

Signed accepted-unit cost change after context selection, compression, output-length or reasoning changes under comparable quality.

Property	Contract
Calculation	current_complete_cost_per_accepted_outcome - baseline_complete_cost_per_accepted_outcome
Unit	currency/accepted_outcome
Required evidence	current_complete_cost_per_accepted_outcome; baseline_complete_cost_per_accepted_outcome; readable_variant_acceptance_results; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_complete_cost_per_accepted_outcome, baseline_complete_cost_per_accepted_outcome, readable_variant_acceptance_results, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Negative means lower cost. Match workload/model and acceptance; no semantic quality inference from token quantities alone.

Input context tokens per task

Contract: METRIC-ROI-17.input_context_tokens_per_task.

Token-use diagnostic for a declared compatible model and workload.

Property	Contract
Calculation	input_context_token_count / eligible_task_count
Unit	token/task
Required evidence	input_context_token_count; eligible_task_count; model_version; token_definition
Denominator	reference: eligible_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: input_context_token_count, eligible_task_count, model_version, token_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Supporting diagnostic within the same model and workload. Token count is not accepted output, value or a universal cross-model workload unit.

Output tokens per task

Contract: METRIC-ROI-17.output_tokens_per_task.

Token-use diagnostic for a declared compatible model and workload.

Property	Contract
Calculation	output_token_count / eligible_task_count
Unit	token/task
Required evidence	output_token_count; eligible_task_count; model_version; token_definition
Denominator	reference: eligible_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: output_token_count, eligible_task_count, model_version, token_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Supporting diagnostic within the same model and workload. Token count is not accepted output, value or a universal cross-model workload unit.

Reasoning tokens per task

Contract: METRIC-ROI-17.reasoning_tokens_per_task.

Token-use diagnostic for a declared compatible model and workload.

Property	Contract
Calculation	reasoning_token_count / eligible_task_count
Unit	token/task
Required evidence	reasoning_token_count; eligible_task_count; model_version; token_definition
Denominator	reference: eligible_task_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: reasoning_token_count, eligible_task_count, model_version, token_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Supporting diagnostic within the same model and workload. Token count is not accepted output, value or a universal cross-model workload unit.

18. Retrieval, tool and evaluation economics

METRIC-ROI-18. Net accepted-outcome gain and cost change from retrieval, reranking, tools, validation or evaluators, including ingestion, indexing, storage, tool fees and false acceptance/rejection costs.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Tool/evaluation evidence + complete costs; Existing route/variant/retrieval/tool/evaluation records, full cost and acceptance

Evidence limits: Compare observed variants; unobserved routes or ablations require separately identified readable evaluation evidence for semantic claims; metadata-only cost alternatives stay price simulations; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Accepted outcome gain from retrieval tools evaluation

Contract: METRIC-ROI-18.accepted_outcome_gain_from_retrieval_tools_evaluation.

Signed gain in comparable accepted business outcomes from a supported retrieval, tool or evaluator variant.

Property	Contract
Calculation	current_mature_accepted_outcomes - comparable_baseline_mature_accepted_outcomes
Unit	accepted_outcome
Required evidence	readable_variant_results; current_mature_accepted_outcomes; comparable_baseline_mature_accepted_outcomes; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: readable_variant_results, current_mature_accepted_outcomes, comparable_baseline_mature_accepted_outcomes, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show accepted outcome gain from retrieval tools evaluation with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Complete retrieval tool evaluation cost

Contract: METRIC-ROI-18.complete_retrieval_tool_evaluation_cost.

Complete disjoint cost of retrieval, reranking, tools and evaluation within the delivery boundary.

Property	Contract
Calculation	sum(unique_ingestion_indexing_storage_reranking_tool_evaluator_and_operating_costs)
Unit	currency
Required evidence	ingestion_costs; indexing_costs; storage_costs; reranking_costs; tool_fees; evaluator_costs; operating_costs; component_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: ingestion_costs, indexing_costs, storage_costs, reranking_costs, tool_fees, evaluator_costs, operating_costs, component_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show complete retrieval tool evaluation cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Retrieval tool evaluation cost change

Contract: METRIC-ROI-18.retrieval_tool_evaluation_cost_change.

Signed complete cost change across comparable variants.

Property	Contract
Calculation	current_complete_retrieval_tool_evaluation_cost - baseline_complete_retrieval_tool_evaluation_cost
Unit	currency
Required evidence	current_complete_retrieval_tool_evaluation_cost; baseline_complete_retrieval_tool_evaluation_cost; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: current_complete_retrieval_tool_evaluation_cost, baseline_complete_retrieval_tool_evaluation_cost, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show retrieval tool evaluation cost change with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

False acceptance loss

Contract: METRIC-ROI-18.false_acceptance_loss.

Observed downstream losses from evaluator decisions independently determined to be false acceptances.

Property	Contract
Calculation	sum(unique_realized_losses_linked_to_validated_false_acceptances)
Unit	currency
Required evidence	validated_false_acceptance_records; mature_downstream_loss_records; loss_linkage_and_deduplication
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: validated_false_acceptance_records, mature_downstream_loss_records, loss_linkage_and_deduplication. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show false acceptance loss with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

False rejection cost

Contract: METRIC-ROI-18.false_rejection_cost.

Evidenced extra expense or lost contribution from independently validated false rejections.

Property	Contract
Calculation	sum(nonoverlapping_evidenced_false_rejection_expense_and_contribution_loss)
Unit	currency
Required evidence	validated_false_rejection_records; expense_or_fulfillment_consequence_evidence; nonoverlapping_loss_ledger
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: validated_false_rejection_records, expense_or_fulfillment_consequence_evidence, nonoverlapping_loss_ledger. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	A rejected candidate is not automatically valuable lost work. Unobserved counterfactual value remains modeled.

Net retrieval tool evaluation benefit

Contract: METRIC-ROI-18.net_retrieval_tool_evaluation_benefit.

Attributable financial gain after complete added retrieval/tool/evaluation cost and disjoint decision-error losses.

Property	Contract
Calculation	attributable_financial_outcome_gain - incremental_complete_component_cost - disjoint_false_acceptance_loss - disjoint_false_rejection_cost
Unit	currency
Required evidence	attributable_financial_outcome_gain; incremental_complete_component_cost; disjoint_false_acceptance_loss; disjoint_false_rejection_cost; benefit_cost_inclusion_map; identified_comparison
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: attributable_financial_outcome_gain, incremental_complete_component_cost, disjoint_false_acceptance_loss, disjoint_false_rejection_cost, benefit_cost_inclusion_map, identified_comparison. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Publish operational outcome gain independently when monetary consequences are unsupported. Avoid counting evaluator losses twice in downstream outcome value.

19. Knowledge reuse and shared-capability value

METRIC-ROI-19. Reduction in repeated research, asset creation, evaluation or integration effort per accepted result through reused knowledge/workflows. Count shared development once; separate reuse value from cache and staffing benefits already credited.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Asset lineage + reuse events + time/outcome comparison; Asset/license/seat inventories, reuse/subscription histories, accepted work and applicable effort or invoices

Evidence limits: Identify reuse and cost consolidation without counting the same saved effort or canceled license twice; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Reuse effort reduction per accepted result

Contract: METRIC-ROI-19.reuse_effort_reduction_per_accepted_result.

Reduction in repeated research, asset creation, evaluation or integration labor for comparable accepted results using shared assets.

Property	Contract
Calculation	baseline_labor_per_accepted_result - reuse_path_labor_per_accepted_result
Unit	labor_hour/accepted_outcome
Required evidence	asset_lineage; reuse_events; baseline_labor_per_accepted_result; reuse_path_labor_per_accepted_result; effort_evidence_basis; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: asset_lineage, reuse_events, baseline_labor_per_accepted_result, reuse_path_labor_per_accepted_result, effort_evidence_basis, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Keep recorded/inferred effort and task class visible. Do not count the same effort as reuse, cache benefit and staffing benefit.

Shared capability development cost

Contract: METRIC-ROI-19.shared_capability_development_cost.

Shared asset development and maintenance cost counted once and allocated by a declared basis.

Property	Contract
Calculation	sum(unique_shared_development_and_maintenance_costs)
Unit	currency
Required evidence	shared_asset_cost_ledger; stable_shared_cost_ids; allocation_policy
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: shared_asset_cost_ledger, stable_shared_cost_ids, allocation_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show shared capability development cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Net shared capability cost improvement

Contract: METRIC-ROI-19.net_shared_capability_cost_improvement.

Comparable complete cost reduction from reuse after disjoint shared development and maintenance allocations.

Property	Contract
Calculation	baseline_complete_delivery_cost - current_reuse_delivery_cost - disjoint_shared_development_maintenance_allocation
Unit	currency
Required evidence	asset_lineage; reuse_events; baseline_complete_delivery_cost; current_reuse_delivery_cost; disjoint_shared_development_maintenance_allocation; comparison_contract; benefit_overlap_map
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: asset_lineage, reuse_events, baseline_complete_delivery_cost, current_reuse_delivery_cost, disjoint_shared_development_maintenance_allocation, comparison_contract, benefit_overlap_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Label economic cost efficiency. Realized cash benefit needs separate spending evidence and one benefit ID across consuming teams.

20. Seat-license and tool-consolidation value

METRIC-ROI-20. Paid license cost / active outcome-producing seat; realized savings from removed or resized subscriptions after migration and overlap costs. Seat activity alone does not prove business value.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; License invoices + roster + outcomes; Asset/license/seat inventories, reuse/subscription histories, accepted work and applicable effort or invoices

Evidence limits: Identify reuse and cost consolidation without counting the same saved effort or canceled license twice; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Paid license cost

Contract: METRIC-ROI-20.paid_license_cost.

Observed subscription and seat-license expense in the declared period.

Property	Contract
Calculation	sum(unique_subscription_and_seat_license_charges)
Unit	currency
Required evidence	license_invoices; contract_and_seat_inventory; expense_period
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: license_invoices, contract_and_seat_inventory, expense_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show paid license cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Active outcome producing seat count

Contract: METRIC-ROI-20.active_outcome_producing_seat_count.

Licensed seats with supported accepted business output in the period.

Property	Contract
Calculation	count_unique(licensed_seats_linked_to_mature_accepted_work)
Unit	seat
Required evidence	license_seat_roster; seat_actor_work_linkage; accepted_work; cohort_acceptance_and_maturity_contract
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: license_seat_roster, seat_actor_work_linkage, accepted_work, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show active outcome producing seat count with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

License cost per outcome producing seat

Contract: METRIC-ROI-20.license_cost_per_outcome_producing_seat.

Paid license cost per active seat producing accepted outcomes.

Property	Contract
Calculation	paid_license_cost / active_outcome_producing_seat_count
Unit	currency/seat
Required evidence	paid_license_cost; active_outcome_producing_seat_count; seat_and_invoice_scope
Denominator	reference: active_outcome_producing_seat_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: paid_license_cost, active_outcome_producing_seat_count, seat_and_invoice_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Active login or license assignment alone is not productive output. Retain paid, active and productive seat counts separately.

Gross subscription expense reduction

Contract: METRIC-ROI-20.gross_subscription_expense_reduction.

Realized reduction in removed or resized subscriptions under a valid comparable contract boundary.

Property	Contract
Calculation	comparable_baseline_subscription_expense - actual_post_change_subscription_expense
Unit	currency
Required evidence	baseline_contract_invoices; post_change_contract_invoices; subscription_removal_or_resize_evidence; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_contract_invoices, post_change_contract_invoices, subscription_removal_or_resize_evidence, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show gross subscription expense reduction with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Net tool consolidation benefit

Contract: METRIC-ROI-20.net_tool_consolidation_benefit.

Subscription expense reduction after migration, overlap and disjoint operating costs.

Property	Contract
Calculation	gross_subscription_expense_reduction - migration_cost - overlap_subscription_cost - disjoint_added_operating_cost
Unit	currency
Required evidence	gross_subscription_expense_reduction; migration_cost; overlap_subscription_cost; disjoint_added_operating_cost; benefit_cost_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: gross_subscription_expense_reduction, migration_cost, overlap_subscription_cost, disjoint_added_operating_cost, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Count each canceled or resized subscription once. Preserve negative transition periods and distinguish cash from amortized economic cost.

Outcome-producing license utilization rate

Contract: METRIC-ROI-20.outcome_producing_license_utilization_rate.

Resolve licensed seat or licensed workflow identities to mature accepted work during the same license period. Evaluate each entity type separately; an employee, account, invocation or enabled license alone does not establish productive use.

Property	Contract
Calculation	distinct_eligible_licensed_entities_with_mature_accepted_work / eligible_licensed_entity_count
Unit	ratio
Required evidence	license_roster_and_effective_terms; licensed_entity_to_work_evidence; mature_accepted_work_records; eligible_licensed_entity_count; licensed_entity_type
Denominator	reference: eligible_licensed_entity_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing roster completeness or accepted-work linkage withholds the full-scope rate. Preserve supported active outcome-producing seat counts and unknown linkage.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

21. Batching and service-tier economics

METRIC-ROI-21. Comparable workload cost difference for synchronous, asynchronous, batch or priority execution, net of delay and operational costs and subject to the same deadline and acceptance rules.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Provider billing + workflow deadlines; Paid-attempt chains and terminal outcomes; cache eligibility/retention plus read/write/storage charges; batch/service-tier usage, deadlines and acceptance; applicable rate and contract versions

Evidence limits: Reconcile actual costs and comparable price/cost alternatives; hypothetical route/latency/quality effects stay modeled; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Execution tier cost per accepted outcome

Contract: METRIC-ROI-21.execution_tier_cost_per_accepted_outcome.

Complete cost per accepted outcome by synchronous, asynchronous, batch or priority execution tier.

Property	Contract
Calculation	complete_tier_workload_cost / tier_mature_accepted_outcomes
Unit	currency/accepted_outcome
Required evidence	execution_tier; complete_tier_workload_cost; tier_mature_accepted_outcomes; deadline_and_acceptance_contract
Denominator	reference: tier_mature_accepted_outcomes; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: execution_tier, complete_tier_workload_cost, tier_mature_accepted_outcomes, deadline_and_acceptance_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show execution tier cost per accepted outcome with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Tier delay and operating cost

Contract: METRIC-ROI-21.tier_delay_and_operating_cost.

Evidenced disjoint cost consequences of execution-tier delay and operations.

Property	Contract
Calculation	sum(dated_delay_consequence_costs + disjoint_tier_operating_costs)
Unit	currency
Required evidence	delay_consequence_records; tier_operating_costs; component_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: delay_consequence_records, tier_operating_costs, component_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Unpriced delay remains an elapsed-time observation. Hypothetical latency or quality consequences stay modeled.

Net batch service tier cost improvement

Contract: METRIC-ROI-21.net_batch_service_tier_cost_improvement.

Signed comparable cost improvement from execution-tier choice after delay and operating consequences.

Property	Contract
Calculation	baseline_complete_workload_cost - alternative_complete_workload_cost - disjoint_tier_delay_and_operating_cost
Unit	currency
Required evidence	baseline_complete_workload_cost; alternative_complete_workload_cost; disjoint_tier_delay_and_operating_cost; observed_or_evaluated_deadline_and_acceptance_parity; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_complete_workload_cost, alternative_complete_workload_cost, disjoint_tier_delay_and_operating_cost, observed_or_evaluated_deadline_and_acceptance_parity, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Require identical business deadlines and acceptance rules. Lower provider price alone does not establish a valid operational alternative.

22. Commitment and provisioned-capacity economics

METRIC-ROI-22. Amortized used and unused commitment cost, utilization and net cost versus the applicable alternative. Separate unavoidable commitments from cancellable future capacity.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Contracts + billing + demand; Commitments, provisioned resources, demand, serving/queue records, linked accepted outcomes, deadlines and service-target results

Evidence limits: Allocate effective cost and useful capacity; provider API workloads do not imply access to provider-owned hardware measurements; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Amortized used commitment cost

Contract: METRIC-ROI-22.amortized_used_commitment_cost.

Declared-period amortized commitment cost allocated to consumed capacity.

Property	Contract
Calculation	amortized_commitment_cost * supported_used_capacity_share
Unit	currency
Required evidence	commitment_contract; amortization_schedule; supported_used_capacity_share; allocation_policy
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: commitment_contract, amortization_schedule, supported_used_capacity_share, allocation_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Allocation is economic cost, not exact per-request telemetry. Reconcile used and unused amounts to the same commitment pool.

Amortized unused commitment cost

Contract: METRIC-ROI-22.amortized_unused_commitment_cost.

Amortized commitment cost allocated to unused capacity in the declared period.

Property	Contract
Calculation	amortized_commitment_cost - amortized_used_commitment_cost
Unit	currency
Required evidence	amortized_commitment_cost; amortized_used_commitment_cost; allocation_policy
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: amortized_commitment_cost, amortized_used_commitment_cost, allocation_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show amortized unused commitment cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Commitment capacity utilization

Contract: METRIC-ROI-22.commitment_capacity_utilization.

Consumed eligible capacity relative to contracted provisioned capacity under the contract’s unit and time scope.

Property	Contract
Calculation	eligible_consumed_capacity / contracted_provisioned_capacity
Unit	ratio
Required evidence	eligible_consumed_capacity; contracted_provisioned_capacity; contract_capacity_unit_and_window
Denominator	reference: contracted_provisioned_capacity; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: eligible_consumed_capacity, contracted_provisioned_capacity, contract_capacity_unit_and_window. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show commitment capacity utilization with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Net commitment cost improvement

Contract: METRIC-ROI-22.net_commitment_cost_improvement.

Net economic cost reduction against the applicable comparable noncommitted or alternative contract.

Property	Contract
Calculation	complete_comparable_alternative_cost - complete_commitment_cost_including_unused_capacity
Unit	currency
Required evidence	complete_comparable_alternative_cost; complete_commitment_cost_including_unused_capacity; contract_version; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Use an applicable valid contract/rate comparator.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: complete_comparable_alternative_cost, complete_commitment_cost_including_unused_capacity, contract_version, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show net commitment cost improvement with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Unavoidable committed cost

Contract: METRIC-ROI-22.unavoidable_committed_cost.

Remaining cost contractually unavoidable over the declared horizon.

Property	Contract
Calculation	sum(non_cancellable_remaining_contractual_payment_obligations)
Unit	currency
Required evidence	contract_obligations; cancellation_terms; as_of_date; horizon
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: contract_obligations, cancellation_terms, as_of_date, horizon. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Distinguish unavoidable cash obligations from amortized expense and future alternatives.

Cancellable future capacity cost

Contract: METRIC-ROI-22.cancellable_future_capacity_cost.

Remaining future spend that can be canceled under evidenced terms, before applicable exit penalties.

Property	Contract
Calculation	sum(cancellable_future_capacity_charges) - applicable_cancellation_penalties
Unit	currency
Required evidence	future_capacity_schedule; cancellation_terms; cancellation_penalties; as_of_date
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Current contract terms and a declared cancellation date; no claim that cancellation has occurred.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: future_capacity_schedule, cancellation_terms, cancellation_penalties, as_of_date. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Opportunity under contract terms, not realized savings. Keep cancellation eligibility and timing visible.

23. Provider price, contract and mix efficiency

METRIC-ROI-23. Separate spend change into workload volume, model/provider mix, effective rates, cache/batch effects and currency. Credit negotiated savings only against a valid contracted comparator.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Billing + versioned rates + mix; Paid-attempt chains and terminal outcomes; cache eligibility/retention plus read/write/storage charges; batch/service-tier usage, deadlines and acceptance; applicable rate and contract versions

Evidence limits: Reconcile actual costs and comparable price/cost alternatives; hypothetical route/latency/quality effects stay modeled; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Spend change from volume

Contract: METRIC-ROI-23.spend_change_from_volume.

Signed component of total spend change under a frozen, order-defined or symmetric decomposition that conserves the observed change.

Property	Contract
Calculation	declared_decomposition_component(volume, baseline_usage_and_rates, current_usage_and_rates)
Unit	currency
Required evidence	baseline_usage_and_rates; current_usage_and_rates; frozen_decomposition_method; compatible_billing_units; component_interaction_allocation
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_usage_and_rates, current_usage_and_rates, frozen_decomposition_method, compatible_billing_units, component_interaction_allocation. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Apply one frozen decomposition to the combined compatible scope. Sum all components plus explicitly unexplained residual to observed spend change.
Display	Show method, interactions and sign. Components explain one spend change and are not additive benefits beyond that change.

Spend change from model provider mix

Contract: METRIC-ROI-23.spend_change_from_model_provider_mix.

Signed component of total spend change under a frozen, order-defined or symmetric decomposition that conserves the observed change.

Property	Contract
Calculation	declared_decomposition_component(model_provider_mix, baseline_usage_and_rates, current_usage_and_rates)
Unit	currency
Required evidence	baseline_usage_and_rates; current_usage_and_rates; frozen_decomposition_method; compatible_billing_units; component_interaction_allocation
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_usage_and_rates, current_usage_and_rates, frozen_decomposition_method, compatible_billing_units, component_interaction_allocation. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Apply one frozen decomposition to the combined compatible scope. Sum all components plus explicitly unexplained residual to observed spend change.
Display	Show method, interactions and sign. Components explain one spend change and are not additive benefits beyond that change.

Spend change from effective rates

Contract: METRIC-ROI-23.spend_change_from_effective_rates.

Signed component of total spend change under a frozen, order-defined or symmetric decomposition that conserves the observed change.

Property	Contract
Calculation	declared_decomposition_component(effective_rates, baseline_usage_and_rates, current_usage_and_rates)
Unit	currency
Required evidence	baseline_usage_and_rates; current_usage_and_rates; frozen_decomposition_method; compatible_billing_units; component_interaction_allocation
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_usage_and_rates, current_usage_and_rates, frozen_decomposition_method, compatible_billing_units, component_interaction_allocation. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Apply one frozen decomposition to the combined compatible scope. Sum all components plus explicitly unexplained residual to observed spend change.
Display	Show method, interactions and sign. Components explain one spend change and are not additive benefits beyond that change.

Spend change from cache batch effects

Contract: METRIC-ROI-23.spend_change_from_cache_batch_effects.

Signed component of total spend change under a frozen, order-defined or symmetric decomposition that conserves the observed change.

Property	Contract
Calculation	declared_decomposition_component(cache_batch, baseline_usage_and_rates, current_usage_and_rates)
Unit	currency
Required evidence	baseline_usage_and_rates; current_usage_and_rates; frozen_decomposition_method; compatible_billing_units; component_interaction_allocation
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_usage_and_rates, current_usage_and_rates, frozen_decomposition_method, compatible_billing_units, component_interaction_allocation. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Apply one frozen decomposition to the combined compatible scope. Sum all components plus explicitly unexplained residual to observed spend change.
Display	Show method, interactions and sign. Components explain one spend change and are not additive benefits beyond that change.

Spend change from currency

Contract: METRIC-ROI-23.spend_change_from_currency.

Signed component of total spend change under a frozen, order-defined or symmetric decomposition that conserves the observed change.

Property	Contract
Calculation	declared_decomposition_component(currency, baseline_usage_and_rates, current_usage_and_rates)
Unit	currency
Required evidence	baseline_usage_and_rates; current_usage_and_rates; frozen_decomposition_method; compatible_billing_units; component_interaction_allocation
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_usage_and_rates, current_usage_and_rates, frozen_decomposition_method, compatible_billing_units, component_interaction_allocation. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Apply one frozen decomposition to the combined compatible scope. Sum all components plus explicitly unexplained residual to observed spend change.
Display	Show method, interactions and sign. Components explain one spend change and are not additive benefits beyond that change.

Negotiated contract cost improvement

Contract: METRIC-ROI-23.negotiated_contract_cost_improvement.

Supported price benefit against the applicable contracted comparator at comparable usage and terms.

Property	Contract
Calculation	comparable_usage_cost_at_valid_baseline_contract - comparable_usage_cost_at_current_contract
Unit	currency
Required evidence	valid_baseline_contract; current_contract; comparable_usage; effective_dates; included_fees_credits_and_terms
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	A valid dated contracted comparator with comparable volume, commitments and terms.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: valid_baseline_contract, current_contract, comparable_usage, effective_dates, included_fees_credits_and_terms. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Do not credit negotiated savings against an inapplicable public list price. Label modeled usage repricing and observed invoice changes distinctly.

Spend decomposition residual

Contract: METRIC-ROI-23.spend_decomposition_residual.

Observed spend change not explained by the declared volume, mix, rate, cache/batch and currency decomposition.

Property	Contract
Calculation	observed_spend_change - sum(declared_decomposition_components)
Unit	currency
Required evidence	observed_spend_change; declared_decomposition_components
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: observed_spend_change, declared_decomposition_components. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Preserve signed unexplained residual rather than forcing complete attribution.

Gross supported cost per invocation

Contract: METRIC-ROI-23.gross_cost_per_invocation.

Use model invocation grain, including paid failures and retries once. A conversation turn or accepted business result is not an invocation. Keep inference-only and fuller declared cost boundaries separate.

Property	Contract
Calculation	complete_supported_invocation_cost / eligible_invocation_count
Unit	USD_per_invocation
Required evidence	invocation_identity_and_lifecycle; complete_supported_invocation_cost; eligible_invocation_count; currency_and_cost_basis
Denominator	reference: eligible_invocation_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_financial
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing rates or incomplete paid-attempt cost withholds this unit cost while observed invocation counts remain usable.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

24. Latency and deadline value

METRIC-ROI-24. Accepted outcomes within the business deadline / eligible work, plus controlled impact of latency on abandonment, conversion, SLA credits or labor wait. Request latency alone is a diagnostic.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Lifecycle events + business consequence; Cross-system work start/end, deadlines, handoffs and downstream outcomes; recorded or calibrated person-time for effort and labor-wait claims

Evidence limits: Separate elapsed time, queue time and person-time; price only evidenced business consequences; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Accepted deadline attainment rate

Contract: METRIC-ROI-24.accepted_deadline_attainment_rate.

Eligible business work delivered as accepted within its declared deadline.

Property	Contract
Calculation	unique_accepted_outcomes_within_business_deadline / eligible_work_item_count
Unit	ratio
Required evidence	accepted_outcome_records; work_start_and_acceptance_dates; business_deadlines; eligible_work_item_count; cohort_acceptance_and_maturity_contract
Denominator	reference: eligible_work_item_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: accepted_outcome_records, work_start_and_acceptance_dates, business_deadlines, eligible_work_item_count, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Use business deadlines and eligible work, including missed, rejected and abandoned work. Fast request responses alone do not satisfy this measure.

Latency effect on abandonment

Contract: METRIC-ROI-24.latency_effect_on_abandonment.

Identified change in work abandonment under comparable latency conditions.

Property	Contract
Calculation	identified_abandonment_probability_current_latency - identified_abandonment_probability_reference_latency
Unit	probability_point
Required evidence	lifecycle_abandonment_records; latency_condition_history; identified_comparison; eligible_work_cohort
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: lifecycle_abandonment_records, latency_condition_history, identified_comparison, eligible_work_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Estimate the complete-cohort probability difference under one comparison design; do not sum probabilities.
Display	Show latency effect on abandonment with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Latency effect on conversion

Contract: METRIC-ROI-24.latency_effect_on_conversion.

Identified change in realized conversion probability associated with a supported latency intervention.

Property	Contract
Calculation	identified_conversion_probability_current_latency - identified_conversion_probability_reference_latency
Unit	probability_point
Required evidence	fulfilled_conversion_records; latency_condition_history; identified_comparison; eligible_work_cohort
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: fulfilled_conversion_records, latency_condition_history, identified_comparison, eligible_work_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Estimate comparable complete-cohort conversion probabilities; aggregate components before calculating differences.
Display	Show latency effect on conversion with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Latency sla credit change

Contract: METRIC-ROI-24.latency_sla_credit_change.

Attributable change in actual SLA credit expense under the relevant contract.

Property	Contract
Calculation	baseline_actual_or_supported_counterfactual_sla_credits - current_actual_sla_credits
Unit	currency
Required evidence	sla_contract; dated_credit_entries; latency_and_incident_records; identified_comparison
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: sla_contract, dated_credit_entries, latency_and_incident_records, identified_comparison. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show latency sla credit change with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Latency labor wait change

Contract: METRIC-ROI-24.latency_labor_wait_change.

Signed reduction in person-time actually spent waiting under supported workflow evidence.

Property	Contract
Calculation	baseline_supported_person_wait_hours - current_supported_person_wait_hours
Unit	labor_hour
Required evidence	recorded_or_calibrated_person_wait_records; human_machine_wait_classification; comparison_contract; effort_evidence_basis
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: recorded_or_calibrated_person_wait_records, human_machine_wait_classification, comparison_contract, effort_evidence_basis. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Elapsed queue delay is not automatically human waiting. Preserve recorded/inferred distinction and concurrent productive activity.

Request latency

Contract: METRIC-ROI-24.request_latency.

Diagnostic elapsed request duration under a declared measurement and statistic.

Property	Contract
Calculation	declared_statistic(request_end_timestamp - request_start_timestamp)
Unit	millisecond
Required evidence	request_boundary_events; request_latency_statistic; measurement_boundary
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: request_boundary_events, request_latency_statistic, measurement_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the named statistic over compatible underlying request durations; never average percentiles.
Display	Request latency is diagnostic, distinct from business deadlines, workflow cycle time and person-hours.

End-to-end workflow duration

Contract: METRIC-ROI-24.workflow_elapsed_duration.

Use the declared whole-workflow run boundary, including supported queueing and retries. Preserve terminal type and open/immature runs; request latency is a different grain.

Property	Contract
Calculation	workflow_terminal_at - workflow_started_at
Unit	second
Required evidence	matched_workflow_run_identity; workflow_started_at; workflow_terminal_at; clock_context
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing endpoints or incompatible clocks makes the duration unavailable. Negative durations remain clock-inconsistent rather than clipped. Complete root timing can remain available when stage traces are incomplete.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Critical-path stage duration

Contract: METRIC-ROI-24.critical_path_stage_duration.

Partition the observed workflow critical path into disjoint stage intervals by a deterministic tie rule. Queue intervals require actual boundaries; unclassified critical-path time remains explicit and stage totals reconcile to elapsed duration.

Property	Contract
Calculation	sum(nonoverlapping_stage_intervals_on_declared_workflow_critical_path)
Unit	second
Required evidence	complete_workflow_dependency_dag; matched_stage_interval_metadata; versioned_critical_path_and_partition_rule; clock_context
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Incomplete dependency or stage timing evidence withholds the decomposition while supported root elapsed duration remains available. Never sum parallel or nested spans as elapsed time.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Workflow span work duration

Contract: METRIC-ROI-24.workflow_span_work_duration.

Report unique observed span work separately from elapsed workflow time, with population labels for parallel or nested spans. State whether parent-inclusive or leaf-only accounting is used; these populations cannot be combined or billed again as disjoint time.

Property	Contract
Calculation	sum(duration_of_unique_spans_in_declared_span_population)
Unit	span_second
Required evidence	span_identity_and_parent_links; matched_span_start_and_end_times; declared_span_population_and_nesting_policy; clock_context
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing span boundaries or identity makes the affected span population unavailable. Preserve supported observations and trace coverage; incomplete traces cannot represent full-workflow work.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

25. Reliability, fallback and continuity value

METRIC-ROI-25. Accepted outcomes rescued after failure, net rescue cost, and observed lost-work reduction. Provider concentration and failover readiness support the interpretation; avoided outage losses remain modeled unless identified.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Failure chains + accepted outcomes + incident evidence; Defects/reopens, review/intervention, failed attempts and linked terminal outcomes

Evidence limits: Reconstruct accepted automation, rework and rescue; no-human-intervention claims require coverage of the intervention path; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Accepted rescued outcome count

Contract: METRIC-ROI-25.accepted_rescued_outcome_count.

Unique mature accepted business outcomes reached through a linked successful rescue after failure.

Property	Contract
Calculation	count_unique(accepted_outcomes_with_evidenced_failed_primary_and_successful_rescue_chain)
Unit	accepted_outcome
Required evidence	primary_failure_and_fallback_lineage; terminal_accepted_outcomes; cohort_acceptance_and_maturity_contract
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: primary_failure_and_fallback_lineage, terminal_accepted_outcomes, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show accepted rescued outcome count with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Accepted rescue rate

Contract: METRIC-ROI-25.accepted_rescue_rate.

Accepted rescues divided by eligible failed primary work with sufficiently observed rescue paths.

Property	Contract
Calculation	accepted_rescued_outcome_count / eligible_failed_primary_work_count
Unit	ratio
Required evidence	accepted_rescued_outcome_count; eligible_failed_primary_work_count; rescue_path_coverage
Denominator	reference: eligible_failed_primary_work_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: accepted_rescued_outcome_count, eligible_failed_primary_work_count, rescue_path_coverage. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show accepted rescue rate with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Net rescue cost

Contract: METRIC-ROI-25.net_rescue_cost.

Incremental complete cost of rescue relative to the named alternative, including failed primary attempts exactly once.

Property	Contract
Calculation	complete_rescued_chain_cost - supported_alternative_chain_cost
Unit	currency
Required evidence	complete_rescued_chain_cost; supported_alternative_chain_cost; parent_child_cost_reconciliation; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: complete_rescued_chain_cost, supported_alternative_chain_cost, parent_child_cost_reconciliation, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Positive is extra rescue cost; negative is lower cost. Preserve the full chain cost and distinguish a no-rescue alternative from a successful primary alternative.

Observed lost work reduction

Contract: METRIC-ROI-25.observed_lost_work_reduction.

Reduction in lost accepted-work opportunity during comparable incidents, supported by mature incident/work outcomes.

Property	Contract
Calculation	baseline_lost_work_count - current_lost_work_count
Unit	work_item
Required evidence	incident_exposure_records; lost_work_and_terminal_outcomes; comparison_contract; maturity_window
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incident_exposure_records, lost_work_and_terminal_outcomes, comparison_contract, maturity_window. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show observed lost work reduction with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Provider concentration

Contract: METRIC-ROI-25.provider_concentration.

Concentration of the declared workload or cost across providers as a supporting resilience diagnostic.

Property	Contract
Calculation	sum(provider_share_in_declared_basis ** 2)
Unit	HHI
Required evidence	provider_workload_or_cost_totals; declared_concentration_basis; complete_scope_total
Denominator	reference: complete_scope_total; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: provider_workload_or_cost_totals, declared_concentration_basis, complete_scope_total. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute provider shares and HHI from combined compatible scope; never add concentration indices.
Display	Identify whether concentration is cost-, request- or outcome-weighted. Concentration alone does not price outage risk.

Verified failover readiness share

Contract: METRIC-ROI-25.verified_failover_readiness_share.

Share of eligible failover scenarios with current supported successful evidence under declared service and acceptance requirements.

Property	Contract
Calculation	verified_ready_failover_scenarios / eligible_failover_scenarios
Unit	ratio
Required evidence	existing_failover_evaluation_evidence; failover_scenario_inventory; required_acceptance_and_service_targets; evidence_freshness
Denominator	reference: eligible_failover_scenarios; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: existing_failover_evaluation_evidence, failover_scenario_inventory, required_acceptance_and_service_targets, evidence_freshness. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Read existing failover evidence. Readiness is not proof of avoided outage losses.

Avoided outage loss scenario

Contract: METRIC-ROI-25.avoided_outage_loss_scenario.

Modeled change in outage loss based on explicit probabilities, exposure and severity assumptions.

Property	Contract
Calculation	expected_outage_loss_without_continuity_controls - expected_outage_loss_with_controls - disjoint_control_cost
Unit	currency
Required evidence	outage_probability_scenarios; exposure_and_loss_severity_model; control_cost; scenario_horizon
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Explicit outage scenario alternatives, calibrated where possible and bounded to the declared horizon.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: outage_probability_scenarios, exposure_and_loss_severity_model, control_cost, scenario_horizon. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Always modeled unless a separately identified realized loss effect exists; never add expected loss to realized cash ROI.

Forwarding continuity during attribution degradation

Contract: METRIC-ROI-25.degraded_forwarding_continuity_rate.

Match receipt and forwarding records independently of the degraded attribution component. Count each eligible request once within declared degradation intervals; a forwarded request does not imply provider success or accepted business work.

Property	Contract
Calculation	eligible_requests_with_evidenced_forwarding_during_degradation / eligible_requests_during_degradation
Unit	ratio
Required evidence	independent_request_receipt_and_forwarding_metadata; attribution_degradation_intervals; eligible_requests_during_degradation; versioned_fail_open_rule
Denominator	reference: eligible_requests_during_degradation; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown independent request denominator or forwarding receipt withholds the full-scope rate. A disabled collector cannot declare its own telemetry complete, and no new inline dependency is introduced.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Request latency difference during attribution degradation

Contract: METRIC-ROI-25.degraded_request_latency_difference.

Compare mean request durations under the same workload, request boundary, source sampling and clock basis. Retain sample counts and uncertainty; raw latency change is descriptive unless a separately justified causal design identifies attribution-induced latency.

Property	Contract
Calculation	mean_eligible_degraded_request_duration - mean_compatible_reference_request_duration
Unit	second
Required evidence	independent_request_timing_metadata; attribution_degradation_intervals; compatible_reference_request_timing; declared_workload_and_clock_comparison
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing comparable reference timing withholds the difference while supported degraded duration observations remain visible. Do not subtract percentiles or pool incompatible workloads.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

26. Productive AI adoption

METRIC-ROI-26. Eligible team members or workflow instances producing accepted AI-assisted work / the relevant eligible population. Pair with cost and outcome gain; keep licensed, active and productive adoption distinct.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Roster/work inventory + usage + outcomes; Eligible work/rosters, longitudinal exposure, accepted outcomes and supported benefit measures

Evidence limits: Calculate productive adoption, distributions, retention and coverage; uncovered opportunity is a forecast; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Licensed adoption rate

Contract: METRIC-ROI-26.licensed_adoption_rate.

Eligible team members holding the relevant license or entitlement.

Property	Contract
Calculation	eligible_licensed_member_count / eligible_team_member_count
Unit	ratio
Required evidence	effective_roster; license_or_entitlement_history; eligible_team_member_count
Denominator	reference: eligible_team_member_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: effective_roster, license_or_entitlement_history, eligible_team_member_count. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show licensed adoption rate with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Active adoption rate

Contract: METRIC-ROI-26.active_adoption_rate.

Eligible team members with supported AI usage in the declared window.

Property	Contract
Calculation	eligible_active_ai_member_count / eligible_team_member_count
Unit	ratio
Required evidence	effective_roster; dated_ai_usage; eligible_team_member_count; active_use_definition
Denominator	reference: eligible_team_member_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: effective_roster, dated_ai_usage, eligible_team_member_count, active_use_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Active use is separate from licensed access and accepted productive work.

Productive member count

Contract: METRIC-ROI-26.productive_member_count.

Eligible team members associated with mature accepted AI-assisted work under supported links.

Property	Contract
Calculation	count_unique(eligible_members_with_supported_accepted_ai_assisted_work)
Unit	member
Required evidence	effective_roster; supported_ai_exposure_and_work_links; accepted_work; cohort_acceptance_and_maturity_contract
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: effective_roster, supported_ai_exposure_and_work_links, accepted_work, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show productive member count with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Productive member adoption rate

Contract: METRIC-ROI-26.productive_member_adoption_rate.

Eligible team members producing accepted AI-assisted work divided by the eligible roster.

Property	Contract
Calculation	productive_member_count / eligible_team_member_count
Unit	ratio
Required evidence	productive_member_count; eligible_team_member_count; effective_roster_scope
Denominator	reference: eligible_team_member_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: productive_member_count, eligible_team_member_count, effective_roster_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Pair with cost and outcome gain. Team-period exposure must not imply individual request/task linkage that is unsupported.

Productive workflow adoption rate

Contract: METRIC-ROI-26.productive_workflow_adoption_rate.

Eligible workflow instances producing accepted AI-assisted outcomes divided by eligible workflow volume.

Property	Contract
Calculation	accepted_ai_assisted_workflow_instance_count / eligible_workflow_instance_count
Unit	ratio
Required evidence	accepted_ai_assisted_workflow_instance_count; eligible_workflow_instance_count; workflow_eligibility_contract; supported_exposure_scope
Denominator	reference: eligible_workflow_instance_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: accepted_ai_assisted_workflow_instance_count, eligible_workflow_instance_count, workflow_eligibility_contract, supported_exposure_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show productive workflow adoption rate with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Workflow stage count

Contract: METRIC-ROI-26.workflow_stage_count.

Count distinct approved workflow definitions in noticed, trialed, repeated, embedded or accepted stages from existing evidence. Declare exclusive current-state versus cumulative-funnel policy; missing stages remain unknown.

Property	Contract
Calculation	count_unique(eligible_approved_workflow_definitions_in_declared_stage)
Unit	workflow
Required evidence	approved_workflow_registry; stage_evidence; versioned_stage_definition; cohort_window_and_stage_policy
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing stage evidence is unknown, not the lowest stage or zero. A complete known cohort with no qualifying workflows is zero.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Workflow stage conversion rate

Contract: METRIC-ROI-26.workflow_stage_conversion_rate.

Freeze the source-stage cohort and observation horizon before examining transitions. Count workflows reaching the declared target stage once; matching tenant, stage policy and cohort maturity are required.

Property	Contract
Calculation	unique_mature_source_stage_workflows_reaching_target_stage / eligible_mature_source_stage_workflow_count
Unit	ratio
Required evidence	source_stage_cohort_ids; target_stage_transition_evidence; eligible_mature_source_stage_workflow_count; versioned_stage_and_maturity_rules
Denominator	reference: eligible_mature_source_stage_workflow_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown or immature transitions retain their states. Missing or nonpositive mature source-stage denominator withholds the ratio while supported stage counts remain available.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Accepted AI-assisted workflow instance count

Contract: METRIC-ROI-26.accepted_ai_assisted_workflow_instance_count.

Count distinct accepted workflow instances with supported AI assistance, using versioned work-unit and acceptance rules. Group only by authorized aggregate business-unit, role-group and geography dimensions.

Property	Contract
Calculation	count_unique(eligible_mature_accepted_ai_assisted_workflow_instances)
Unit	workflow_instance
Required evidence	workflow_instance_identity; supported_ai_assistance_evidence; mature_unrevoked_acceptance; declared_reporting_cohort
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown exposure or acceptance withholds affected classifications; absent prices or total eligible-universe denominator does not withhold supported accepted-instance counts.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

27. Distribution of value across teams and cohorts

METRIC-ROI-27. Within-comparable-task changes in accepted cost, quality and throughput by team, experience, workflow and accessibility/language cohort. Show distributions and negative effects, not an employee leaderboard.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Cohort metadata + comparable outcomes; Eligible work/rosters, longitudinal exposure, accepted outcomes and supported benefit measures

Evidence limits: Calculate productive adoption, distributions, retention and coverage; uncovered opportunity is a forecast; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Cohort accepted unit cost change

Contract: METRIC-ROI-27.cohort_accepted_unit_cost_change.

Comparable accepted-unit cost change by team, experience, workflow or declared accessibility/language cohort.

Property	Contract
Calculation	current_cohort_cost_per_accepted_outcome - baseline_cohort_cost_per_accepted_outcome
Unit	currency/accepted_outcome
Required evidence	cohort_metadata; current_cohort_cost_per_accepted_outcome; baseline_cohort_cost_per_accepted_outcome; pre_treatment_comparable_task_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cohort_metadata, current_cohort_cost_per_accepted_outcome, baseline_cohort_cost_per_accepted_outcome, pre_treatment_comparable_task_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Publish distributions, uncertainty and negative cohorts; do not build employee performance rankings.

Cohort quality change

Contract: METRIC-ROI-27.cohort_quality_change.

Comparable mature quality-rate change by declared cohort.

Property	Contract
Calculation	current_cohort_quality_rate - baseline_cohort_quality_rate
Unit	probability_point
Required evidence	cohort_metadata; mature_cohort_quality_records; validated_quality_definition; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cohort_metadata, mature_cohort_quality_records, validated_quality_definition, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible quality numerators and denominators within each cohort before differencing; retain cohort distribution.
Display	Show cohort quality change with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cohort accepted throughput change

Contract: METRIC-ROI-27.cohort_accepted_throughput_change.

Comparable accepted-output rate change by declared cohort.

Property	Contract
Calculation	current_cohort_accepted_output_rate - baseline_cohort_accepted_output_rate
Unit	accepted_outcome/declared_resource_unit
Required evidence	cohort_metadata; current_cohort_accepted_output_rate; baseline_cohort_accepted_output_rate; resource_basis; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cohort_metadata, current_cohort_accepted_output_rate, baseline_cohort_accepted_output_rate, resource_basis, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show the resource denominator and heterogeneous effects. Cohort membership or adoption does not independently establish causality.

28. Retention of beneficial workflows

METRIC-ROI-28. Share of mature workflow/team cohorts that sustain accepted use and a positive measured benefit over the selected horizon. Repeated usage without benefit is a separate retention measure.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Longitudinal outcomes + benefit records; Eligible work/rosters, longitudinal exposure, accepted outcomes and supported benefit measures

Evidence limits: Calculate productive adoption, distributions, retention and coverage; uncovered opportunity is a forecast; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Beneficial workflow retention rate

Contract: METRIC-ROI-28.beneficial_workflow_retention_rate.

Mature workflow or team cohorts that sustain accepted use and a positive supported benefit over the horizon.

Property	Contract
Calculation	mature_cohorts_with_sustained_accepted_use_and_positive_benefit / eligible_mature_cohorts
Unit	ratio
Required evidence	longitudinal_cohort_usage; mature_accepted_outcomes; supported_nonduplicated_benefit; retention_horizon; eligible_mature_cohorts
Denominator	reference: eligible_mature_cohorts; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: longitudinal_cohort_usage, mature_accepted_outcomes, supported_nonduplicated_benefit, retention_horizon, eligible_mature_cohorts. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show benefit type and evidence basis. Immature cohorts and unpriced usage are not beneficial-retention successes.

Usage retention rate

Contract: METRIC-ROI-28.usage_retention_rate.

Mature eligible cohorts with sustained use, independently of whether benefit is supported.

Property	Contract
Calculation	mature_cohorts_with_sustained_use / eligible_mature_cohorts
Unit	ratio
Required evidence	longitudinal_cohort_usage; retention_horizon; eligible_mature_cohorts; retained_use_definition
Denominator	reference: eligible_mature_cohorts; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: longitudinal_cohort_usage, retention_horizon, eligible_mature_cohorts, retained_use_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Repeated usage without demonstrated benefit remains usage retention, not beneficial-workflow retention.

29. Cross-team coordination and handoff value

METRIC-ROI-29. Change in total effort, queue time and accepted completion across the entire multi-team process. Record added review or remediation downstream so local savings cannot hide a larger transferred burden.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Shared workflow IDs + handoff/time/outcome records; Cross-system work start/end, deadlines, handoffs and downstream outcomes; recorded or calibrated person-time for effort and labor-wait claims

Evidence limits: Separate elapsed time, queue time and person-time; price only evidenced business consequences; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Cross team total effort change

Contract: METRIC-ROI-29.cross_team_total_effort_change.

Signed change in complete person-hours across every participating team for comparable accepted process scope.

Property	Contract
Calculation	current_all_team_person_hours - baseline_all_team_person_hours
Unit	labor_hour
Required evidence	shared_workflow_ids; recorded_or_calibrated_all_team_effort; effort_evidence_basis; shared_support_and_unassigned_effort; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: shared_workflow_ids, recorded_or_calibrated_all_team_effort, effort_evidence_basis, shared_support_and_unassigned_effort, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Negative is less total effort. Preserve downstream review, remediation and transferred work; local savings alone are insufficient.

Cross team queue time change

Contract: METRIC-ROI-29.cross_team_queue_time_change.

Signed change in elapsed queue/wait duration across supported process handoffs.

Property	Contract
Calculation	current_total_process_queue_time - baseline_total_process_queue_time
Unit	hour
Required evidence	cross_team_handoff_events; queue_state_boundaries; comparison_contract; queue_statistic
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: cross_team_handoff_events, queue_state_boundaries, comparison_contract, queue_statistic. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute queue-time statistic from compatible whole-process paths, avoiding overlap double counting; compare like statistics.
Display	Elapsed queue time is distinct from person-time and machine execution.

Cross team accepted completion change

Contract: METRIC-ROI-29.cross_team_accepted_completion_change.

Change in mature accepted completions across the complete multi-team process.

Property	Contract
Calculation	current_whole_process_accepted_count - baseline_comparable_accepted_count
Unit	accepted_outcome
Required evidence	shared_workflow_ids; whole_process_terminal_outcomes; comparison_contract; cohort_acceptance_and_maturity_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: shared_workflow_ids, whole_process_terminal_outcomes, comparison_contract, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show cross team accepted completion change with its baseline_estimate classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Downstream review hours

Contract: METRIC-ROI-29.downstream_review_hours.

Recorded downstream review labor within the end-to-end process boundary.

Property	Contract
Calculation	sum(nonoverlapping_recorded_downstream_review_person_hours)
Unit	labor_hour
Required evidence	downstream_review_records; recorded_person_time; workflow_handoff_lineage
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: downstream_review_records, recorded_person_time, workflow_handoff_lineage. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Include in total effort once. Inferred review time must retain its inferred-effort label.

Downstream remediation hours

Contract: METRIC-ROI-29.downstream_remediation_hours.

Recorded downstream remediation labor associated with delivered or reopened work.

Property	Contract
Calculation	sum(nonoverlapping_recorded_downstream_remediation_person_hours)
Unit	labor_hour
Required evidence	downstream_defect_reopen_records; recorded_remediation_person_time; workflow_handoff_lineage
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: downstream_defect_reopen_records, recorded_remediation_person_time, workflow_handoff_lineage. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Preserve late failures and include these hours once in complete workflow effort.

30. Opportunity coverage and expansion value

METRIC-ROI-30. Share of eligible workflow volume already measured; estimated net value of uncovered use cases = addressable accepted work times tested unit benefit minus rollout costs. Label the opportunity forecast.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Workflow inventory + tested unit economics; Eligible work/rosters, longitudinal exposure, accepted outcomes and supported benefit measures

Evidence limits: Calculate productive adoption, distributions, retention and coverage; uncovered opportunity is a forecast; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Measured eligible workflow volume share

Contract: METRIC-ROI-30.measured_eligible_workflow_volume_share.

Share of independently enumerated eligible workflow volume with the declared measurement contract.

Property	Contract
Calculation	measured_eligible_workflow_volume / total_eligible_workflow_volume
Unit	ratio
Required evidence	independent_workflow_inventory; measured_workflow_ids; total_eligible_workflow_volume; volume_unit_and_period
Denominator	reference: total_eligible_workflow_volume; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: independent_workflow_inventory, measured_workflow_ids, total_eligible_workflow_volume, volume_unit_and_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	The denominator must be independent of observed usage; missing inventory means unknown coverage.

Addressable uncovered accepted work

Contract: METRIC-ROI-30.addressable_uncovered_accepted_work.

Forecast volume of uncovered work expected to satisfy the tested acceptance and demand scope.

Property	Contract
Calculation	sum(eligible_uncovered_work_volume * supported_acceptance_and_addressability_factors)
Unit	accepted_outcome
Required evidence	uncovered_work_inventory; eligible_demand_evidence; tested_acceptance_scope; forecast_assumptions
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Use tested comparable task evidence and explicit transfer/acceptance assumptions.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: uncovered_work_inventory, eligible_demand_evidence, tested_acceptance_scope, forecast_assumptions. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Label opportunity forecast, not observed output. New objects alone do not establish addressable demand.

Uncovered opportunity net value

Contract: METRIC-ROI-30.uncovered_opportunity_net_value.

Forecast net value of uncovered eligible use cases after rollout costs.

Property	Contract
Calculation	addressable_uncovered_accepted_work * tested_unit_net_benefit - disjoint_rollout_costs
Unit	currency
Required evidence	addressable_uncovered_accepted_work; tested_unit_net_benefit; disjoint_rollout_costs; transfer_and_demand_assumptions; benefit_cost_inclusion_map
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Explicitly assess transfer from tested workflows to uncovered work.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: addressable_uncovered_accepted_work, tested_unit_net_benefit, disjoint_rollout_costs, transfer_and_demand_assumptions, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show forecast horizon, demand, unit-benefit evidence and uncertainty. This is not realized financial ROI.

Recurring accepted program target attainment

Contract: METRIC-ROI-30.recurring_accepted_program_target_attainment.

Compare supported recurring accepted workflow instances with a target fixed before the review period. Match target and actual unit, population and period; a mandate target is not observed adoption or a causal baseline.

Property	Contract
Calculation	verified_recurring_accepted_workflow_volume / frozen_declared_program_target_volume
Unit	ratio
Required evidence	verified_recurring_accepted_workflow_volume; frozen_declared_program_target_volume; versioned_production_and_recurrence_rules; matched_program_work_unit_and_period
Denominator	reference: frozen_declared_program_target_volume; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown, changed-without-history or nonpositive targets withhold attainment. Preserve supported production volume and target revisions separately; attainment may exceed one.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

31. Learning, onboarding and skill effects

METRIC-ROI-31. Time and cost to independently accepted proficiency, later error/rework rates and retained ability with and without assistance. Capture benefits and skill degradation over a declared follow-up period.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Learning/process records + outcomes + comparison; Existing learning/assessment or experiment registries and later accepted outcomes

Evidence limits: Measure proficiency/validated milestones; retained skill, counterfactual learning and option value require their own evidence or remain unknown/modeled; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Time to independently accepted proficiency

Contract: METRIC-ROI-31.time_to_independently_accepted_proficiency.

Elapsed time from the declared learning/onboarding start to independent accepted proficiency.

Property	Contract
Calculation	independent_proficiency_acceptance_timestamp - learning_start_timestamp
Unit	day
Required evidence	learning_start_records; independent_proficiency_assessment; accepted_proficiency_definition; follow_up_contract
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: learning_start_records, independent_proficiency_assessment, accepted_proficiency_definition, follow_up_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the named duration statistic from compatible learners; preserve unachieved and censored proficiency.
Display	Course completion or assisted task completion alone does not establish independent proficiency.

Cost to independently accepted proficiency

Contract: METRIC-ROI-31.cost_to_independently_accepted_proficiency.

Complete learning and support cost per independently accepted proficiency outcome.

Property	Contract
Calculation	complete_learning_onboarding_support_cost / independently_accepted_proficiency_count
Unit	currency/proficiency_outcome
Required evidence	complete_learning_onboarding_support_cost; independently_accepted_proficiency_count; cost_scope; accepted_proficiency_definition
Denominator	reference: independently_accepted_proficiency_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: complete_learning_onboarding_support_cost, independently_accepted_proficiency_count, cost_scope, accepted_proficiency_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show cost to independently accepted proficiency with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Later error rate

Contract: METRIC-ROI-31.later_error_rate.

Observed error incidence over the declared mature follow-up work after proficiency.

Property	Contract
Calculation	later_error_work_count / later_eligible_work_count
Unit	ratio
Required evidence	later_error_records; later_eligible_work_count; follow_up_period; mature_quality_definition
Denominator	reference: later_eligible_work_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: later_error_records, later_eligible_work_count, follow_up_period, mature_quality_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show later error rate with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Later rework hours

Contract: METRIC-ROI-31.later_rework_hours.

Recorded rework effort during the declared post-proficiency follow-up period.

Property	Contract
Calculation	sum(nonoverlapping_recorded_follow_up_rework_hours)
Unit	labor_hour
Required evidence	follow_up_rework_records; recorded_person_time; follow_up_period
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: follow_up_rework_records, recorded_person_time, follow_up_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show later rework hours with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Retained skill with assistance

Contract: METRIC-ROI-31.retained_skill_with_assistance.

Retained task capability under a declared assisted follow-up assessment.

Property	Contract
Calculation	declared_validated_skill_score(assisted_follow_up_assessment_results)
Unit	score
Required evidence	assisted_follow_up_assessment; validated_scoring_rule; task_scope; follow_up_period
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: assisted_follow_up_assessment, validated_scoring_rule, task_scope, follow_up_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the same validated score from compatible assessments; do not combine different skill scales.
Display	Keep assistance condition explicit. This does not establish unassisted retained capability.

Retained skill without assistance

Contract: METRIC-ROI-31.retained_skill_without_assistance.

Retained task capability under an independently observed unassisted follow-up assessment.

Property	Contract
Calculation	declared_validated_skill_score(unassisted_follow_up_assessment_results)
Unit	score
Required evidence	unassisted_follow_up_assessment; validated_scoring_rule; task_scope; follow_up_period
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: unassisted_follow_up_assessment, validated_scoring_rule, task_scope, follow_up_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the same validated score from compatible assessments; do not combine different skill scales.
Display	Show retained skill without assistance with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Learning or skill effect

Contract: METRIC-ROI-31.learning_or_skill_effect.

Signed identified change in proficiency, later error/rework or retained skill under a named validated outcome scale.

Property	Contract
Calculation	identified_current_learning_outcome - identified_comparable_reference_learning_outcome
Unit	declared_learning_outcome_unit
Required evidence	learning_outcome_type; validated_learning_measure; identified_comparison; follow_up_period
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: learning_outcome_type, validated_learning_measure, identified_comparison, follow_up_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Use one named outcome scale and follow-up horizon per estimate; retain heterogeneous and negative effects.
Display	Report gains and skill degradation. Counterfactual learning requires its own identification evidence.

32. Innovation and experiment option value

METRIC-ROI-32. Cost per independently validated experiment or decision-changing milestone; realized commercial value when linked later. Model option value separately with probability, horizon and follow-on cost assumptions.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Experiment/decision registry + validation + later outcomes; Existing learning/assessment or experiment registries and later accepted outcomes

Evidence limits: Measure proficiency/validated milestones; retained skill, counterfactual learning and option value require their own evidence or remain unknown/modeled; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Cost per independently validated experiment

Contract: METRIC-ROI-32.cost_per_independently_validated_experiment.

Complete experimental cost per unique independently validated experiment.

Property	Contract
Calculation	complete_experiment_cohort_cost / independently_validated_experiment_count
Unit	currency/experiment
Required evidence	experiment_registry; complete_experiment_cohort_cost; independent_validation_records; validated_experiment_definition
Denominator	reference: independently_validated_experiment_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: experiment_registry, complete_experiment_cohort_cost, independent_validation_records, validated_experiment_definition. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Include failed and invalid experiments in cohort costs. Generated variants are not independently validated experiments.

Cost per decision changing milestone

Contract: METRIC-ROI-32.cost_per_decision_changing_milestone.

Complete experiment/program cost per unique evidenced milestone that changed a decision.

Property	Contract
Calculation	complete_program_cohort_cost / independently_validated_decision_changing_milestone_count
Unit	currency/milestone
Required evidence	decision_registry; complete_program_cohort_cost; milestone_validation; decision_change_evidence
Denominator	reference: independently_validated_decision_changing_milestone_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: decision_registry, complete_program_cohort_cost, milestone_validation, decision_change_evidence. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show cost per decision changing milestone with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Realized innovation commercial value

Contract: METRIC-ROI-32.realized_innovation_commercial_value.

Attributable later commercial contribution linked to the validated experiment or milestone, after disjoint commercial costs.

Property	Contract
Calculation	linked_incremental_recognized_revenue - linked_incremental_delivery_costs - disjoint_follow_on_costs
Unit	currency
Required evidence	experiment_to_commercial_lineage; realized_financial_records; identified_comparison; follow_on_costs; benefit_cost_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: experiment_to_commercial_lineage, realized_financial_records, identified_comparison, follow_on_costs, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Validation alone is not commercial value. Keep later realized value separate from speculative option value.

Innovation option value scenario

Contract: METRIC-ROI-32.innovation_option_value_scenario.

Modeled value of future options under explicit probability, timing and follow-on cost assumptions.

Property	Contract
Calculation	sum(probability_scenario * discounted_future_net_payoff_scenario) - disjoint_current_option_cost
Unit	currency
Required evidence	option_scenario_probabilities; payoff_horizons; follow_on_costs; discounting_policy; disjoint_current_option_cost
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Explicit alternative, scenario probability model and horizon; probabilities and payoffs require provenance.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: option_scenario_probabilities, payoff_horizons, follow_on_costs, discounting_policy, disjoint_current_option_cost. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Publish assumptions, ranges and residual uncertainty. Modeled option value is excluded from realized financial ROI.

33. Time to realized value and payback

METRIC-ROI-33. Elapsed time to the first verified benefit; payback occurs when cumulative incremental cash benefits cover cumulative cash costs. Keep projected payback and annualized run rates distinct from realized results.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Implementation dates + cash benefit/cost ledger; Finance transactions, declared investment scope, comparable histories and benefit dates

Evidence limits: Reconcile non-overlapping financial changes, net benefit and cash payback; preserve operational value when a monetary consequence is unsupported; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Time to first verified benefit

Contract: METRIC-ROI-33.time_to_first_verified_benefit.

Elapsed time from the declared implementation start to the first independently supported benefit.

Property	Contract
Calculation	first_verified_benefit_timestamp - implementation_start_timestamp
Unit	day
Required evidence	implementation_start_timestamp; verified_benefit_record; benefit_realization_and_type
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: implementation_start_timestamp, verified_benefit_record, benefit_realization_and_type. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the chosen duration distribution for compatible implementation cohorts; keep no-benefit cohorts right-censored.
Display	Name the benefit type. Dashboard activation, first usage or an unpriced forecast is not automatically realized value.

Cumulative incremental cash net benefit

Contract: METRIC-ROI-33.cumulative_incremental_cash_net_benefit.

Cumulative dated incremental cash benefits less cumulative actual incremental cash disbursements.

Property	Contract
Calculation	cumulative_nonoverlapping_incremental_cash_benefits - cumulative_incremental_cash_disbursements
Unit	currency
Required evidence	dated_cash_benefit_ledger; dated_incremental_cash_disbursements; cash_scope; benefit_cost_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: dated_cash_benefit_ledger, dated_incremental_cash_disbursements, cash_scope, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Use actual cash timing; never count both upfront payment and its amortized expense in cash payback.

Realized cash payback time

Contract: METRIC-ROI-33.realized_cash_payback_time.

First elapsed time at which cumulative incremental cash benefits cover cumulative actual incremental cash costs.

Property	Contract
Calculation	earliest_t_with(cumulative_cash_benefits(t) >= cumulative_cash_costs(t)) - investment_start_time, after a positive cash outlay
Unit	day
Required evidence	dated_cash_benefit_ledger; dated_cash_cost_ledger; positive_initial_cash_outlay; investment_start_time; cash_scope
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: dated_cash_benefit_ledger, dated_cash_cost_ledger, positive_initial_cash_outlay, investment_start_time, cash_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute payback from the combined compatible dated cash ledger; never average payback periods.
Display	If costs have not been recovered, show not yet reached and the observation horizon. No positive outlay means conventional payback is not applicable.

Projected cash payback time

Contract: METRIC-ROI-33.projected_cash_payback_time.

Scenario date when projected cumulative incremental cash benefits cover projected cumulative cash costs.

Property	Contract
Calculation	earliest_projected_t_with(projected_cumulative_cash_benefits(t) >= projected_cumulative_cash_costs(t)) - investment_start_time
Unit	day
Required evidence	dated_observed_cash_flows; projected_cash_flow_assumptions; positive_investment_scope; forecast_horizon
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Explicit forecast of the named investment and alternative with cash timing.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: dated_observed_cash_flows, projected_cash_flow_assumptions, positive_investment_scope, forecast_horizon. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute projected crossover from compatible dated cash-flow scenarios.
Display	Label projected, including no crossover within horizon; never present it as realized payback.

Annualized benefit run rate

Contract: METRIC-ROI-33.annualized_benefit_run_rate.

Scenario annualization of a supported period benefit under an explicit repeatability assumption.

Property	Contract
Calculation	supported_period_benefit * declared_annualization_factor
Unit	currency/year
Required evidence	supported_period_benefit; observed_period_length; declared_annualization_factor; seasonality_and_repeatability_assumptions
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Supported period result and explicit stationarity, seasonality and demand assumptions.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: supported_period_benefit, observed_period_length, declared_annualization_factor, seasonality_and_repeatability_assumptions. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Annualized run rate is a projection, not annual realized cash benefit.

34. Budget and demand-planning value

METRIC-ROI-34. Budget variance and out-of-sample forecast error by team/workflow; monetize only evidenced reductions in emergency buying, unused capacity or planning effort.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Budgets + frozen forecasts + actuals; Frozen budget/demand forecasts, their issue dates and realized cost/demand

Evidence limits: Test forecast error and evidenced planning consequences; later revised forecasts cannot replace original vintages; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Budget variance

Contract: METRIC-ROI-34.budget_variance.

Actual spend minus the budget frozen for the corresponding period and scope.

Property	Contract
Calculation	actual_scope_cost - frozen_budget
Unit	currency
Required evidence	actual_scope_cost; frozen_budget; budget_vintage; compatible_period_and_cost_basis
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: actual_scope_cost, frozen_budget, budget_vintage, compatible_period_and_cost_basis. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show budget variance with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Relative budget variance

Contract: METRIC-ROI-34.relative_budget_variance.

Signed budget variance divided by the positive compatible frozen budget.

Property	Contract
Calculation	budget_variance / frozen_budget
Unit	ratio
Required evidence	budget_variance; frozen_budget; budget_vintage
Denominator	reference: frozen_budget; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: budget_variance, frozen_budget, budget_vintage. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show relative budget variance with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Frozen cost forecast error

Contract: METRIC-ROI-34.frozen_cost_forecast_error.

Actual cost minus the forecast frozen before the evaluation period.

Property	Contract
Calculation	actual_cost - issued_frozen_cost_forecast
Unit	currency
Required evidence	actual_cost; issued_frozen_cost_forecast; forecast_issue_date; out_of_sample_period
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: actual_cost, issued_frozen_cost_forecast, forecast_issue_date, out_of_sample_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Preserve original forecast vintages; later revisions cannot replace the evaluated forecast.

Frozen demand forecast error

Contract: METRIC-ROI-34.frozen_demand_forecast_error.

Actual demand minus the forecast issued before the evaluated period.

Property	Contract
Calculation	actual_demand - issued_frozen_demand_forecast
Unit	declared_demand_unit
Required evidence	actual_demand; issued_frozen_demand_forecast; forecast_issue_date; demand_unit; out_of_sample_period
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: actual_demand, issued_frozen_demand_forecast, forecast_issue_date, demand_unit, out_of_sample_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show the demand unit and signed error. Do not combine incompatible token, request and business-outcome demand.

Emergency buying expense reduction

Contract: METRIC-ROI-34.emergency_buying_expense_reduction.

Attributable reduction in evidenced emergency procurement or overage expense.

Property	Contract
Calculation	baseline_emergency_buying_expense - actual_emergency_buying_expense - disjoint_planning_intervention_cost
Unit	currency
Required evidence	emergency_procurement_invoices; comparison_contract; disjoint_planning_intervention_cost
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: emergency_procurement_invoices, comparison_contract, disjoint_planning_intervention_cost. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show emergency buying expense reduction with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Unused capacity expense reduction

Contract: METRIC-ROI-34.unused_capacity_expense_reduction.

Attributable reduction in actual avoidable unused-capacity expense.

Property	Contract
Calculation	baseline_avoidable_unused_capacity_expense - actual_avoidable_unused_capacity_expense
Unit	currency
Required evidence	capacity_contracts; unused_capacity_invoice_or_commitment_records; avoidable_cost_evidence; comparison_contract
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: capacity_contracts, unused_capacity_invoice_or_commitment_records, avoidable_cost_evidence, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	An unchanged unavoidable commitment is not saved cash. Avoid duplicating commitment-economics benefits.

Planning effort change

Contract: METRIC-ROI-34.planning_effort_change.

Comparable change in total planning effort using the recorded or calibrated labor basis.

Property	Contract
Calculation	current_planning_person_hours - baseline_planning_person_hours
Unit	labor_hour
Required evidence	planning_workflow_records; recorded_or_calibrated_effort; effort_evidence_basis; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: planning_workflow_records, recorded_or_calibrated_effort, effort_evidence_basis, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Operational effort change stays separate from realized expense and from FinOps administration benefits already credited.

35. Self-hosted serving resource efficiency

METRIC-ROI-35. Accepted outcomes per GPU-hour and effective cost per accepted outcome, including idle power, hosting, memory, storage and operating labor. GPU busy time and tokens/second are supporting diagnostics.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Serving telemetry + cost + outcomes; Commitments, provisioned resources, demand, serving/queue records, linked accepted outcomes, deadlines and service-target results

Evidence limits: Allocate effective cost and useful capacity; provider API workloads do not imply access to provider-owned hardware measurements; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Accepted outcomes per gpu hour

Contract: METRIC-ROI-35.accepted_outcomes_per_gpu_hour.

Mature accepted business output per provisioned GPU-hour under a declared serving and allocation boundary.

Property	Contract
Calculation	unique_mature_accepted_outcome_count / provisioned_gpu_hours
Unit	accepted_outcome/GPU_hour
Required evidence	unique_mature_accepted_outcome_count; provisioned_gpu_hours; serving_boundary; resource_allocation_basis; cohort_acceptance_and_maturity_contract
Denominator	reference: provisioned_gpu_hours; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: unique_mature_accepted_outcome_count, provisioned_gpu_hours, serving_boundary, resource_allocation_basis, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Declare provisioned versus busy GPU-hour basis; idle resource use remains visible. Provider API use does not imply hardware telemetry access.

Self hosted cost per accepted outcome

Contract: METRIC-ROI-35.self_hosted_cost_per_accepted_outcome.

Complete self-hosted serving cost including idle power, hosting, memory, storage and operating labor per accepted outcome.

Property	Contract
Calculation	complete_self_hosted_serving_cost / unique_mature_accepted_outcome_count
Unit	currency/accepted_outcome
Required evidence	complete_self_hosted_serving_cost; idle_power_hosting_memory_storage_and_operating_costs; cost_completeness_and_allocation; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: complete_self_hosted_serving_cost, idle_power_hosting_memory_storage_and_operating_costs, cost_completeness_and_allocation, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show self hosted cost per accepted outcome with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Gpu busy time share

Contract: METRIC-ROI-35.gpu_busy_time_share.

GPU busy time divided by available GPU time within the same observed resource window.

Property	Contract
Calculation	observed_gpu_busy_time / observed_available_gpu_time
Unit	ratio
Required evidence	observed_gpu_busy_time; observed_available_gpu_time; gpu_telemetry_boundary
Denominator	reference: observed_available_gpu_time; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: observed_gpu_busy_time, observed_available_gpu_time, gpu_telemetry_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Utilization is diagnostic; busy time does not establish useful accepted work.

Serving token throughput

Contract: METRIC-ROI-35.serving_token_throughput.

Observed compatible token output per elapsed serving time.

Property	Contract
Calculation	observed_output_token_count / observed_serving_seconds
Unit	token/second
Required evidence	observed_output_token_count; observed_serving_seconds; model_and_token_definition; telemetry_boundary
Denominator	reference: observed_serving_seconds; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: observed_output_token_count, observed_serving_seconds, model_and_token_definition, telemetry_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Supporting diagnostic within a declared model and token definition; not equivalent to accepted-business throughput.

36. Useful capacity and queue headroom

METRIC-ROI-36. Accepted outcomes delivered within service targets per provisioned unit, plus queue wait and the binding RPM/TPM/concurrency constraint. Measure overload, abandoned work and declared reserve separately.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Provider limits or server metrics + outcome/deadline evidence; Commitments, provisioned resources, demand, serving/queue records, linked accepted outcomes, deadlines and service-target results

Evidence limits: Allocate effective cost and useful capacity; provider API workloads do not imply access to provider-owned hardware measurements; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Useful capacity per provisioned unit

Contract: METRIC-ROI-36.useful_capacity_per_provisioned_unit.

Accepted outcomes meeting declared service targets per provisioned capacity unit and reporting window.

Property	Contract
Calculation	accepted_outcomes_meeting_service_targets / provisioned_capacity_units
Unit	accepted_outcome/capacity_unit
Required evidence	accepted_outcomes_meeting_service_targets; provisioned_capacity_units; capacity_unit_and_window; service_target_contract
Denominator	reference: provisioned_capacity_units; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: accepted_outcomes_meeting_service_targets, provisioned_capacity_units, capacity_unit_and_window, service_target_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show useful capacity per provisioned unit with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Business queue wait

Contract: METRIC-ROI-36.business_queue_wait.

Elapsed time awaiting service within supported queue boundaries.

Property	Contract
Calculation	declared_statistic(queue_exit_timestamp - queue_entry_timestamp)
Unit	millisecond
Required evidence	queue_entry_exit_records; queue_boundary; queue_wait_statistic
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: queue_entry_exit_records, queue_boundary, queue_wait_statistic. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the named statistic from compatible queue observations; preserve censored work.
Display	Queue delay is elapsed time; it does not imply occupied human labor.

Rpm headroom

Contract: METRIC-ROI-36.rpm_headroom.

Remaining capacity within the actual account/project/model/region limit scope and provider-defined window.

Property	Contract
Calculation	request_limit - requests_in_limit_window
Unit	request
Required evidence	requests_in_limit_window; request_limit; limit_scope; provider_limit_window; measurement_timestamp
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: requests_in_limit_window, request_limit, limit_scope, provider_limit_window, measurement_timestamp. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute within each actual limit scope and time window; do not sum unrelated quota pools.
Display	Show the binding limit scope and sign; preserve over-limit observations. Squad demand is not an independent provider quota unless the contract says so.

Tpm headroom

Contract: METRIC-ROI-36.tpm_headroom.

Remaining capacity within the actual account/project/model/region limit scope and provider-defined window.

Property	Contract
Calculation	token_limit - tokens_in_limit_window
Unit	token
Required evidence	tokens_in_limit_window; token_limit; limit_scope; provider_limit_window; measurement_timestamp
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: tokens_in_limit_window, token_limit, limit_scope, provider_limit_window, measurement_timestamp. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute within each actual limit scope and time window; do not sum unrelated quota pools.
Display	Show the binding limit scope and sign; preserve over-limit observations. Squad demand is not an independent provider quota unless the contract says so.

Concurrency headroom

Contract: METRIC-ROI-36.concurrency_headroom.

Remaining capacity within the actual account/project/model/region limit scope and provider-defined window.

Property	Contract
Calculation	concurrency_limit - concurrent_inflight_requests
Unit	request
Required evidence	concurrent_inflight_requests; concurrency_limit; limit_scope; provider_limit_window; measurement_timestamp
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: concurrent_inflight_requests, concurrency_limit, limit_scope, provider_limit_window, measurement_timestamp. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute within each actual limit scope and time window; do not sum unrelated quota pools.
Display	Show the binding limit scope and sign; preserve over-limit observations. Squad demand is not an independent provider quota unless the contract says so.

Overloaded work count

Contract: METRIC-ROI-36.overloaded_work_count.

Eligible work encountering an evidenced binding resource or quota overload condition.

Property	Contract
Calculation	count_unique(eligible_work_with_supported_overload_condition)
Unit	work_item
Required evidence	work_attempt_lineage; resource_or_quota_overload_records; limit_scope
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: work_attempt_lineage, resource_or_quota_overload_records, limit_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show overloaded work count with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Capacity abandoned work count

Contract: METRIC-ROI-36.capacity_abandoned_work_count.

Eligible work abandoned with a supported queue/capacity consequence link.

Property	Contract
Calculation	count_unique(mature_abandoned_work_with_supported_capacity_link)
Unit	work_item
Required evidence	abandonment_records; queue_capacity_work_links; maturity_contract
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: abandonment_records, queue_capacity_work_links, maturity_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Temporal coincidence is not proof that capacity caused abandonment; retain linkage and causal interpretation separately.

Declared capacity reserve

Contract: METRIC-ROI-36.planned_capacity_reserve.

Explicitly declared reserve capacity under the declared demand and service policy.

Property	Contract
Calculation	declared_planned_capacity - declared_expected_required_capacity
Unit	capacity_unit
Required evidence	capacity_plan; expected_demand_model; service_target_policy; capacity_unit_and_window
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Versioned demand forecast and reserve policy; no arbitrary reserve assumption.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: capacity_plan, expected_demand_model, service_target_policy, capacity_unit_and_window. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Combine only fungible capacity pools under one compatible time and service boundary.
Display	Declared reserve is intentional capacity, distinct from waste or unused unavoidable commitment.

Observed request count

Contract: METRIC-ROI-36.observed_request_count.

Count request occurrences independently of turns, model invocations and accepted workflow instances. Distinguish retry attempts by their source request identity and deduplicate repeated telemetry delivery. Token-only records do not create requests.

Property	Contract
Calculation	count_unique(eligible_request_occurrences)
Unit	request
Required evidence	tenant_qualified_request_occurrence_identity; request_receipt_metadata; declared_provider_account_model_region_window; source_population_coverage
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	A known complete empty population is zero. If coverage is incomplete, preserve a clearly scoped observed count; completeness of the total request population remains unknown. Missing quotas or prices cannot withhold supported request counts.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Observed token count

Contract: METRIC-ROI-36.observed_token_count.

Sum non-overlapping observed input, output, cached and other token kinds only under their declared accounting rule. Preserve per-kind counts; cached subset counters must not be added twice. Provider-reported billed tokens and rate-limit tokens are separate bases.

Property	Contract
Calculation	sum_unique(eligible_token_usage_by_request_and_token_kind)
Unit	token
Required evidence	tenant_qualified_request_usage_identity; observed_token_usage_and_kind; declared_token_accounting_rule; source_population_coverage
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing token usage remains unknown rather than zero or inferred from request count. Supported observed token totals remain available without quota evidence; incomplete population coverage remains explicit.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Request quota utilization

Contract: METRIC-ROI-36.request_quota_utilization.

Match counted request attempts to the effective request quota using identical provider, account, model, region, quota kind and limit window. Respect the provider accounting rule for rejected or failed attempts; do not substitute peak throughput or turn counts. Values above one remain observable.

Property	Contract
Calculation	eligible_quota_counted_requests_in_limit_window / effective_request_quota_for_same_limit_window
Unit	ratio
Required evidence	eligible_quota_counted_requests_in_limit_window; effective_request_quota_for_same_limit_window; provider_account_model_region_and_quota_kind; exact_fixed_or_rolling_limit_window; quota_effective_time_and_freshness; quota_accounting_and_population_coverage
Denominator	reference: effective_request_quota_for_same_limit_window; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown, stale, nonpositive, incompatible or incomplete quota evidence withholds utilization while supported request totals remain available. Never derive current utilization from headroom without the same verified quota and window.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Token quota utilization

Contract: METRIC-ROI-36.token_quota_utilization.

Match observed quota-counted tokens to the effective token quota using the same provider, account, model, region, token accounting basis and fixed or rolling window. Separate token limits that use estimated admission tokens from billed tokens; a conversion requires source evidence. Values above one remain observable.

Property	Contract
Calculation	eligible_quota_counted_tokens_in_limit_window / effective_token_quota_for_same_limit_window
Unit	ratio
Required evidence	eligible_quota_counted_tokens_in_limit_window; effective_token_quota_for_same_limit_window; provider_account_model_region_and_quota_kind; exact_fixed_or_rolling_limit_window; quota_effective_time_and_freshness; token_quota_accounting_and_population_coverage
Denominator	reference: effective_token_quota_for_same_limit_window; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown or stale quota, missing token-accounting evidence, incomplete population, or nonpositive denominator withholds utilization. Supported observed token totals remain available; no peak statistic or provisioned-capacity share substitutes for this ratio.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

37. Reconciled and conserved cost basis

METRIC-ROI-37. Invoice-matched cost coverage, attributable cost coverage and unexplained residuals, by compatible period/currency/cost basis. Direct + shared + unallocated amounts must conserve the source total.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Billing + allocation records; Billing totals, independent coverage denominators, source lineage and labels/validation; resolved acceptance/maturity and dependent cost/comparison completeness for outcome/ROI coverage

Evidence limits: Conserve cost and publish evidence/attribution coverage; unknown denominator means unknown coverage; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Invoice matched cost coverage

Contract: METRIC-ROI-37.invoice_matched_cost_coverage.

Share of the independently evidenced invoice cost basis that is matched to compatible cost records.

Property	Contract
Calculation	invoice_matched_eligible_cost_weight / independently_evidenced_invoice_cost_weight
Unit	ratio
Required evidence	invoice_matched_eligible_cost_weight; independently_evidenced_invoice_cost_weight; invoice_period_currency_cost_basis; credit_discount_tax_weighting_policy
Denominator	reference: independently_evidenced_invoice_cost_weight; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: invoice_matched_eligible_cost_weight, independently_evidenced_invoice_cost_weight, invoice_period_currency_cost_basis, credit_discount_tax_weighting_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Declare signed-versus-absolute cost weighting and credit handling. Matching is not proof of exact per-inference telemetry.

Attributable cost coverage

Contract: METRIC-ROI-37.attributable_cost_coverage.

Share of the declared independent source cost pool with supported ownership attribution.

Property	Contract
Calculation	supported_attributable_cost_weight / independent_source_cost_weight
Unit	ratio
Required evidence	supported_attributable_cost_weight; independent_source_cost_weight; attribution_provenance; cost_weighting_policy
Denominator	reference: independent_source_cost_weight; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: supported_attributable_cost_weight, independent_source_cost_weight, attribution_provenance, cost_weighting_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Keep exact ownership, inferred ownership and policy allocation distinguishable; attribution coverage is not correctness.

Direct cost amount

Contract: METRIC-ROI-37.direct_cost_amount.

Source cost assigned through supported direct cost/ownership records.

Property	Contract
Calculation	sum(unique_direct_cost_components)
Unit	currency
Required evidence	direct_cost_records; source_pool_id; cost_basis
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: direct_cost_records, source_pool_id, cost_basis. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show direct cost amount with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Shared allocated cost amount

Contract: METRIC-ROI-37.shared_allocated_cost_amount.

Declared shared-cost allocation assigned once from the source pool.

Property	Contract
Calculation	sum(unique_shared_cost_allocations)
Unit	currency
Required evidence	shared_cost_pool; allocation_records; allocation_policy_version
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: shared_cost_pool, allocation_records, allocation_policy_version. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Policy allocation does not rewrite request ownership or become exact telemetry.

Unallocated cost amount

Contract: METRIC-ROI-37.unallocated_cost_amount.

Source cost retained explicitly without supported direct assignment or shared allocation.

Property	Contract
Calculation	sum(source_cost_components_remaining_unallocated)
Unit	currency
Required evidence	source_cost_pool; direct_assignment_ledger; shared_allocation_ledger; unallocated_component_records
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: source_cost_pool, direct_assignment_ledger, shared_allocation_ledger, unallocated_component_records. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show unallocated cost amount with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Cost conservation residual

Contract: METRIC-ROI-37.cost_conservation_residual.

Difference between direct plus shared plus unallocated amounts and the authoritative source pool.

Property	Contract
Calculation	direct_cost_amount + shared_allocated_cost_amount + unallocated_cost_amount - source_pool_total
Unit	currency
Required evidence	direct_cost_amount; shared_allocated_cost_amount; unallocated_cost_amount; source_pool_total; compatible_period_currency_cost_basis
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: direct_cost_amount, shared_allocated_cost_amount, unallocated_cost_amount, source_pool_total, compatible_period_currency_cost_basis. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Zero means conservation under this declared basis, not necessarily invoice accuracy or ownership correctness. Preserve signed residual.

Invoice reconciliation signed residual

Contract: METRIC-ROI-37.invoice_reconciliation_signed_residual.

Reconciled recorded cost minus authoritative invoice cost for a compatible scope.

Property	Contract
Calculation	recorded_reconciled_cost - authoritative_invoice_cost
Unit	currency
Required evidence	recorded_reconciled_cost; authoritative_invoice_cost; invoice_scope_and_adjustment_map
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: recorded_reconciled_cost, authoritative_invoice_cost, invoice_scope_and_adjustment_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show invoice reconciliation signed residual with its coverage classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Invoice reconciliation absolute error

Contract: METRIC-ROI-37.invoice_reconciliation_absolute_error.

Absolute discrepancy between reconciled recorded and invoiced costs under the same scope.

Property	Contract
Calculation	abs(recorded_reconciled_cost - authoritative_invoice_cost)
Unit	currency
Required evidence	recorded_reconciled_cost; authoritative_invoice_cost; invoice_scope_and_adjustment_map
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: recorded_reconciled_cost, authoritative_invoice_cost, invoice_scope_and_adjustment_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Sum per-scope absolute errors for an error total; never cancel opposite signed discrepancies. Report signed residual separately.
Display	Show invoice reconciliation absolute error with its coverage classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Covered eligible AI spend

Contract: METRIC-ROI-37.covered_ai_spend.

Sum unique supported covered AI cost under the effective contract, retaining unresolved owner amounts and signed adjustments. Separate currencies, periods and cost bases; tier assignment is a typed effective contract attribute.

Property	Contract
Calculation	sum_unique(eligible_ai_cost_under_effective_covered_spend_contract)
Unit	USD
Required evidence	eligible_ai_cost_records; effective_covered_spend_contract; currency_and_cost_basis; trailing_period_and_tier_rules
Denominator	Not applicable
Evidence class	observed_financial
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown cost or contract applicability makes the corresponding full-scope amount unavailable. Preserve a clearly bounded observed partial amount and coverage without assigning a fabricated tier.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

38. Outcome and causal-evidence coverage

METRIC-ROI-38. Share of eligible work with resolved acceptance, complete costs and a valid comparison; share of spend with decision-grade ROI. Display missing, pending, revoked and insufficient evidence separately.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Outcome/finance/comparison evidence; Billing totals, independent coverage denominators, source lineage and labels/validation; resolved acceptance/maturity and dependent cost/comparison completeness for outcome/ROI coverage

Evidence limits: Conserve cost and publish evidence/attribution coverage; unknown denominator means unknown coverage; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Resolved acceptance coverage

Contract: METRIC-ROI-38.resolved_acceptance_coverage.

Share of eligible work with a supported resolved acceptance disposition and maturity state.

Property	Contract
Calculation	eligible_work_with_resolved_acceptance / independent_eligible_work_count
Unit	ratio
Required evidence	resolved_acceptance_records; independent_eligible_work_count; acceptance_mapping_version; maturity_state
Denominator	reference: independent_eligible_work_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: resolved_acceptance_records, independent_eligible_work_count, acceptance_mapping_version, maturity_state. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show resolved acceptance coverage with its coverage classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Complete cost coverage

Contract: METRIC-ROI-38.complete_cost_coverage.

Share of eligible work with every required cost component resolved for the declared result boundary.

Property	Contract
Calculation	eligible_work_with_complete_result_cost / independent_eligible_work_count
Unit	ratio
Required evidence	output_specific_cost_completeness; independent_eligible_work_count; declared_result_cost_boundary
Denominator	reference: independent_eligible_work_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: output_specific_cost_completeness, independent_eligible_work_count, declared_result_cost_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show complete cost coverage with its coverage classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Valid comparison coverage

Contract: METRIC-ROI-38.valid_comparison_coverage.

Share of eligible work supported by a valid comparison for the declared result and causal interpretation.

Property	Contract
Calculation	eligible_work_with_valid_output_comparison / independent_eligible_work_count
Unit	ratio
Required evidence	output_specific_comparison_validity; independent_eligible_work_count; comparison_and_estimand_contract
Denominator	reference: independent_eligible_work_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: output_specific_comparison_validity, independent_eligible_work_count, comparison_and_estimand_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	A valid descriptive benchmark does not count as causal identification when the dependent output claims an attributable effect.

Joint outcome cost comparison coverage

Contract: METRIC-ROI-38.joint_outcome_cost_comparison_coverage.

Share of eligible work simultaneously meeting acceptance, complete-cost and valid-comparison requirements.

Property	Contract
Calculation	eligible_work_in_intersection_of_resolved_acceptance_complete_cost_valid_comparison / independent_eligible_work_count
Unit	ratio
Required evidence	acceptance_cost_comparison_intersection; independent_eligible_work_count; output_evidence_contract
Denominator	reference: independent_eligible_work_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: acceptance_cost_comparison_intersection, independent_eligible_work_count, output_evidence_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Calculate the actual intersection; do not multiply marginal coverage percentages.

Decision grade roi spend coverage

Contract: METRIC-ROI-38.decision_grade_roi_spend_coverage.

Share of independently enumerated eligible spend with supported financial realization and complete decision-grade ROI evidence.

Property	Contract
Calculation	eligible_spend_with_decision_grade_roi / independent_eligible_spend
Unit	ratio
Required evidence	financial_roi_evidence_records; financial_realization_and_causal_validity; eligible_spend_with_decision_grade_roi; independent_eligible_spend; cost_weighting_policy
Denominator	reference: independent_eligible_spend; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: financial_roi_evidence_records, financial_realization_and_causal_validity, eligible_spend_with_decision_grade_roi, independent_eligible_spend, cost_weighting_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	No opaque combined ROI-confidence score. Keep causal validity, cost completeness, acceptance and realization inspectable.

Dependent result weighted coverage

Contract: METRIC-ROI-38.dependent_result_weighted_coverage.

Output-specific coverage separately under work-count, spend and labor weighting, using independently supported populations.

Property	Contract
Calculation	weight_of_eligible_scope_supported_for_named_output / independent_total_eligible_weight_for_same_output
Unit	ratio
Required evidence	named_dependent_output_id; weighting_basis; supported_eligible_weight; independent_total_eligible_weight; output_evidence_contract
Denominator	reference: independent_total_eligible_weight; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: named_dependent_output_id, weighting_basis, supported_eligible_weight, independent_total_eligible_weight, output_evidence_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Present work-, spend- and labor-weighted views separately. Missing independent denominator means unknown coverage for that view; do not suppress other coverage views.

Missing evidence work count

Contract: METRIC-ROI-38.missing_evidence_work_count.

Eligible work assigned the explicit missing status under a versioned output-specific evidence rule.

Property	Contract
Calculation	count_unique(eligible_work_with_named_output_evidence_status = ‘missing’)
Unit	work_item
Required evidence	eligible_work_inventory; named_dependent_output_id; output_evidence_status_records; status_assignment_rule
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: eligible_work_inventory, named_dependent_output_id, output_evidence_status_records, status_assignment_rule. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Keep this state visible with reason codes. State counts require a defined exclusive status rule before summation; unknown is never zero value.

Pending evidence work count

Contract: METRIC-ROI-38.pending_evidence_work_count.

Eligible work assigned the explicit pending status under a versioned output-specific evidence rule.

Property	Contract
Calculation	count_unique(eligible_work_with_named_output_evidence_status = ‘pending’)
Unit	work_item
Required evidence	eligible_work_inventory; named_dependent_output_id; output_evidence_status_records; status_assignment_rule
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: eligible_work_inventory, named_dependent_output_id, output_evidence_status_records, status_assignment_rule. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Keep this state visible with reason codes. State counts require a defined exclusive status rule before summation; unknown is never zero value.

Revoked acceptance work count

Contract: METRIC-ROI-38.revoked_acceptance_work_count.

Eligible work assigned the explicit revoked status under a versioned output-specific evidence rule.

Property	Contract
Calculation	count_unique(eligible_work_with_named_output_evidence_status = ‘revoked’)
Unit	work_item
Required evidence	eligible_work_inventory; named_dependent_output_id; output_evidence_status_records; status_assignment_rule
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: eligible_work_inventory, named_dependent_output_id, output_evidence_status_records, status_assignment_rule. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Keep this state visible with reason codes. State counts require a defined exclusive status rule before summation; unknown is never zero value.

Insufficient evidence work count

Contract: METRIC-ROI-38.insufficient_evidence_work_count.

Eligible work assigned the explicit insufficient status under a versioned output-specific evidence rule.

Property	Contract
Calculation	count_unique(eligible_work_with_named_output_evidence_status = ‘insufficient’)
Unit	work_item
Required evidence	eligible_work_inventory; named_dependent_output_id; output_evidence_status_records; status_assignment_rule
Denominator	Not applicable
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: eligible_work_inventory, named_dependent_output_id, output_evidence_status_records, status_assignment_rule. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Keep this state visible with reason codes. State counts require a defined exclusive status rule before summation; unknown is never zero value.

39. Ownership, discovery and attribution reliability

METRIC-ROI-39. Exact, inferred and unresolved shares; independently validated correctness where labels exist; provenance and freshness. Discovery coverage requires an independent denominator and stays distinct from ownership accuracy.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Identity/feature joins + validation samples; Billing totals, independent coverage denominators, source lineage and labels/validation; resolved acceptance/maturity and dependent cost/comparison completeness for outcome/ROI coverage

Evidence limits: Conserve cost and publish evidence/attribution coverage; unknown denominator means unknown coverage; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Exact ownership share

Contract: METRIC-ROI-39.exact_ownership_share.

Share of eligible observed attribution units in the exact ownership state under the declared unit and scope.

Property	Contract
Calculation	eligible_units_in_exact_ownership_state / eligible_attribution_unit_count
Unit	ratio
Required evidence	ownership_state_records; eligible_attribution_unit_count; attribution_unit_definition; event_time_identity_scope
Denominator	reference: eligible_attribution_unit_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: ownership_state_records, eligible_attribution_unit_count, attribution_unit_definition, event_time_identity_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Exact, eligible inferred and unresolved states remain distinct from allocation. Resolution coverage and inference confidence do not establish correctness.

Inferred ownership share

Contract: METRIC-ROI-39.inferred_ownership_share.

Share of eligible observed attribution units in the inferred ownership state under the declared unit and scope.

Property	Contract
Calculation	eligible_units_in_inferred_ownership_state / eligible_attribution_unit_count
Unit	ratio
Required evidence	ownership_state_records; eligible_attribution_unit_count; attribution_unit_definition; event_time_identity_scope
Denominator	reference: eligible_attribution_unit_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: ownership_state_records, eligible_attribution_unit_count, attribution_unit_definition, event_time_identity_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Exact, eligible inferred and unresolved states remain distinct from allocation. Resolution coverage and inference confidence do not establish correctness.

Unresolved ownership share

Contract: METRIC-ROI-39.unresolved_ownership_share.

Share of eligible observed attribution units in the unresolved ownership state under the declared unit and scope.

Property	Contract
Calculation	eligible_units_in_unresolved_ownership_state / eligible_attribution_unit_count
Unit	ratio
Required evidence	ownership_state_records; eligible_attribution_unit_count; attribution_unit_definition; event_time_identity_scope
Denominator	reference: eligible_attribution_unit_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: ownership_state_records, eligible_attribution_unit_count, attribution_unit_definition, event_time_identity_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Exact, eligible inferred and unresolved states remain distinct from allocation. Resolution coverage and inference confidence do not establish correctness.

Independently validated ownership correctness

Contract: METRIC-ROI-39.independently_validated_ownership_correctness.

Correct ownership predictions among independently labeled evaluable attribution units under a declared sampling design.

Property	Contract
Calculation	independently_correct_ownership_labels / independently_labeled_evaluable_units
Unit	ratio
Required evidence	independent_ownership_labels; frozen_predictions; sampling_design; evaluable_label_count
Denominator	reference: independently_labeled_evaluable_units; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: independent_ownership_labels, frozen_predictions, sampling_design, evaluable_label_count. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Report label provenance, scope, uncertainty and sampling bias. Do not substitute operational confidence or resolved-share for accuracy.

Independent discovery coverage

Contract: METRIC-ROI-39.independent_discovery_coverage.

Discovered eligible systems or invocations divided by an independently enumerated eligible population in the same unit.

Property	Contract
Calculation	discovered_eligible_units / independent_total_eligible_units
Unit	ratio
Required evidence	discovered_unit_ids; independent_inventory_or_usage_denominator; eligible_unit_definition; scope_and_period
Denominator	reference: independent_total_eligible_units; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: discovered_unit_ids, independent_inventory_or_usage_denominator, eligible_unit_definition, scope_and_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Discovery coverage requires an independent denominator and remains distinct from ownership correctness.

Attribution provenance coverage

Contract: METRIC-ROI-39.attribution_provenance_coverage.

Share of eligible attribution results retaining the required replayable source and method lineage.

Property	Contract
Calculation	eligible_results_with_complete_provenance / eligible_attribution_result_count
Unit	ratio
Required evidence	source_event_references; join_or_inference_method_versions; eligible_attribution_result_count; provenance_completeness_rule
Denominator	reference: eligible_attribution_result_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: source_event_references, join_or_inference_method_versions, eligible_attribution_result_count, provenance_completeness_rule. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show attribution provenance coverage with its coverage classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Attribution evidence age

Contract: METRIC-ROI-39.attribution_evidence_age.

Age of source ownership evidence at the declared observation time.

Property	Contract
Calculation	observation_timestamp - latest_valid_source_evidence_timestamp
Unit	second
Required evidence	observation_timestamp; latest_valid_source_evidence_timestamp; evidence_validity_interval
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: observation_timestamp, latest_valid_source_evidence_timestamp, evidence_validity_interval. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the named age distribution from compatible records; do not treat future timestamps as fresh evidence.
Display	Show source age and validity. Freshness does not establish correct ownership or causal attribution.

Source-to-observation lag

Contract: METRIC-ROI-39.source_to_observation_lag.

Elapsed seconds from a source event to first successful ingestion of the same received source revision.

Property	Contract
Calculation	observed_at - source_event_at
Unit	second
Required evidence	source_event_at; observed_at; matched_source_record_revision; versioned_timestamp_mapping; clock_context
Input definitions	source event at: Event timestamp retained from the matched source record revision, normalized with timezone and precision.; observed at: First successful ingestion timestamp of that same revision in the declared pipeline; never last-seen, replay or query time.; matched source record revision: Same tenant, source system, source record ID and source revision for both endpoints.; clock context: Declared clock basis, precision, uncertainty and source-specific compatibility policy.
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline for this observed interval. Comparisons require compatible source revisions, timestamp mappings, clock basis and reporting populations.
Causal interpretation	Descriptive source telemetry only; freshness does not establish attribution correctness, discovery coverage, business impact or causal benefit.
Missing evidence	Missing or incompatible source_event_at, observed_at, record revision, timestamp mapping or clock context makes this output unavailable with a specific reason. Missing served_at does not withhold this interval. Preserve negative intervals with clock-inconsistent status; missing observations never become zero.
Aggregation	Calculate observation, mean, p50 and p95 from compatible received source revisions with sample counts and coverage; deduplicate replay and recompute percentiles from observations. No ratio denominator applies.
Display	Show source, revision, reporting window, timestamp mapping, clock uncertainty, sample count, freshness policy and signed duration. Received-record scope does not establish coverage of unreceived activity.

Unresolved attribution unit count

Contract: METRIC-ROI-39.unresolved_attribution_unit_count.

Count tenant-qualified units whose exact and eligible inferred ownership remains unresolved. Declare event or run grain; allocation-only labels do not establish ownership.

Property	Contract
Calculation	count_unique(eligible_attribution_units_in_unresolved_state)
Unit	attribution_unit
Required evidence	eligible_attribution_unit_ids; versioned_resolution_rule; source_population_completeness
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	A complete known empty cohort is zero; an unknown observed population is unavailable. The total-universe denominator is not required for this count.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Shared-principal workflow resolution share

Contract: METRIC-ROI-39.shared_principal_workflow_resolution_share.

Count each shared-key or machine-account event once when exact or eligible inferred evidence resolves its accountable workflow. Exclude allocation-only assignments and show exact, inferred and unresolved strata.

Property	Contract
Calculation	unique_eligible_shared_principal_events_with_exact_or_eligible_inferred_workflow / eligible_shared_principal_event_count
Unit	ratio
Required evidence	shared_principal_inventory; eligible_shared_principal_event_ids; workflow_resolution_evidence; versioned_resolution_rule
Denominator	reference: eligible_shared_principal_event_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	coverage
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown shared-principal classification or incomplete eligible denominator withholds the full-scope share; retain supported counts and disclose a separately scoped covered cohort.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Observed third-party entity count

Contract: METRIC-ROI-39.observed_third_party_entity_count.

Count separately by entity_type provider, endpoint, subprocessor or jurisdiction with source provenance. Required evidence is conditional by entity_type: provider needs attributed provider identity; endpoint needs an observed endpoint identifier; subprocessor needs an authoritative vendor inventory; jurisdiction needs authoritative processing-location metadata. No entity type inherits the evidence requirement of another type. Endpoint hostnames do not establish a subprocessor or processing jurisdiction; those require existing vendor inventories or approved published records.

Property	Contract
Calculation	count_unique(observed_entities_by_declared_entity_type)
Unit	entity
Required evidence	typed_entity_identity_and_evidence; versioned_entity_identity_mapping; declared_entity_type_and_window
Denominator	Not applicable
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing subprocessor or jurisdiction evidence withholds those entity-type counts while supported provider and endpoint counts remain available. Observed inventory never establishes estate-wide completeness.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

40. FinOps administration and accountability value

METRIC-ROI-40. Change in time/cost to reconcile bills, allocate shared spend, close chargeback disputes and perform planning. Monetary benefit requires actual avoided effort expense or separately valued capacity.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Finance operations and dispute/time records; Existing finance/control/audit workflow records, effort and accepted completion; advisory evaluation/decision traces and supported decision outcomes for availability, false positives and lost work; separate customer-enforcement evidence where assessed

Evidence limits: Measure process cost/time using the same recorded/inferred effort distinction; advisory availability and customer enforcement remain separate; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Finops process effort change

Contract: METRIC-ROI-40.finops_process_effort_change.

Comparable change in labor consumed by bill reconciliation, shared allocation, chargeback disputes or planning.

Property	Contract
Calculation	current_process_person_hours - baseline_comparable_process_person_hours
Unit	labor_hour
Required evidence	finance_process_type; process_work_records; recorded_or_calibrated_effort; effort_evidence_basis; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: finance_process_type, process_work_records, recorded_or_calibrated_effort, effort_evidence_basis, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Report each named finance process separately and retain recorded/inferred effort. Negative means less effort.

Finops process cost per completed item

Contract: METRIC-ROI-40.finops_process_cost_per_completed_item.

Complete finance-administration process cost per accepted completed bill, allocation, dispute or planning item.

Property	Contract
Calculation	complete_finance_process_cost / accepted_completed_finance_process_item_count
Unit	currency/process_item
Required evidence	finance_process_type; complete_finance_process_cost; accepted_completed_finance_process_item_count; completion_contract
Denominator	reference: accepted_completed_finance_process_item_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: finance_process_type, complete_finance_process_cost, accepted_completed_finance_process_item_count, completion_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show finops process cost per completed item with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Realized finops expense reduction

Contract: METRIC-ROI-40.realized_finops_expense_reduction.

Attributable reduced cash expense in finance administration after disjoint intervention cost.

Property	Contract
Calculation	comparable_baseline_finops_cash_expense - actual_finops_cash_expense - disjoint_intervention_cash_cost
Unit	currency
Required evidence	dated_finops_expense_records; comparison_contract; disjoint_intervention_cash_cost; benefit_cost_inclusion_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: dated_finops_expense_records, comparison_contract, disjoint_intervention_cash_cost, benefit_cost_inclusion_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Hours saved without reduced spending remain operational capacity, not cash savings.

Finops capacity valuation scenario

Contract: METRIC-ROI-40.finops_capacity_valuation_scenario.

Separate economic valuation of supported finance-administration capacity under an explicit labor-rate assumption.

Property	Contract
Calculation	supported_finops_released_hours * declared_role_labor_value_rate
Unit	currency
Required evidence	supported_finops_released_hours; declared_role_labor_value_rate; valuation_basis; benefit_overlap_map
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Supported comparable effort change and explicitly declared role/rate valuation.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: supported_finops_released_hours, declared_role_labor_value_rate, valuation_basis, benefit_overlap_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Label capacity valuation, keep it outside realized cash ROI and avoid duplicating planning or reused-capability value.

FinOps process effort saved

Contract: METRIC-ROI-40.finops_process_effort_saved.

Compare total non-overlapping person-hours for compatible finance close or reconstruction work. This is the sign-correct inverse of finops_process_effort_change and keeps that existing output unchanged.

Property	Contract
Calculation	baseline_comparable_process_person_hours - current_process_person_hours
Unit	person_hour
Required evidence	baseline_comparable_process_person_hours; current_process_person_hours; matched_process_role_period_and_effort_basis
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing comparable baseline or current effort withholds the saved-hours comparison. Preserve negative saved hours; observed current effort and separately supported cash evidence remain available.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

41. Observed security and incident economics

METRIC-ROI-41. Change in actual incident response cost, remediation effort, AI-related loss and recovery time, with severity/exposure adjustment. Alert counts and unknown endpoints are diagnostic evidence.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Incident records + attribution + cost; Incident, exposure, remediation, cost and loss-history records, plus qualified external loss references where useful

Evidence limits: Report observed incident economics and separately modeled expected loss; rare absent incidents alone do not establish prevention; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Registered ai related incident count

Contract: METRIC-ROI-41.registered_ai_related_incident_count.

Existing registered incidents with supported AI-related exposure under the declared incident definition.

Property	Contract
Calculation	count_unique(registered_incidents_with_supported_ai_exposure)
Unit	incident
Required evidence	incident_registry; supported_ai_exposure_links; incident_definition; reporting_period
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incident_registry, supported_ai_exposure_links, incident_definition, reporting_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	This count remains available without incident prices or an independent exposure denominator. Alerts and unknown endpoints are separate diagnostics.

Observed incident response cost

Contract: METRIC-ROI-41.observed_incident_response_cost.

Actual evidenced expense of responding to AI-related incidents within the declared scope.

Property	Contract
Calculation	sum(unique_incident_response_expense_entries)
Unit	currency
Required evidence	incident_registry; response_expense_records; incident_cost_linkage; cost_boundary
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incident_registry, response_expense_records, incident_cost_linkage, cost_boundary. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show observed incident response cost with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Incident remediation effort

Contract: METRIC-ROI-41.incident_remediation_effort.

Recorded person-hours consumed by incident remediation.

Property	Contract
Calculation	sum(nonoverlapping_recorded_incident_remediation_hours)
Unit	labor_hour
Required evidence	remediation_work_records; recorded_person_time; incident_work_linkage
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: remediation_work_records, recorded_person_time, incident_work_linkage. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Keep recorded versus inferred effort distinct. Missing cost rates do not remove supported hours.

Realized ai related incident loss

Contract: METRIC-ROI-41.realized_ai_related_incident_loss.

Actual independently evidenced monetary loss attributed to registered AI-related incidents.

Property	Contract
Calculation	sum(unique_realized_incident_loss_entries)
Unit	currency
Required evidence	registered_incident_links; realized_loss_records; loss_deduplication_and_scope
Denominator	Not applicable
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: registered_incident_links, realized_loss_records, loss_deduplication_and_scope. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Separate response cost, realized business loss and recovered funds; prevent duplicate loss entries.

Incident recovery time

Contract: METRIC-ROI-41.incident_recovery_time.

Elapsed incident recovery time under a declared start, recovered state and statistic.

Property	Contract
Calculation	declared_statistic(recovered_timestamp - incident_start_timestamp)
Unit	hour
Required evidence	incident_start_records; recovered_state_evidence; recovery_statistic; unresolved_incident_censoring
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incident_start_records, recovered_state_evidence, recovery_statistic, unresolved_incident_censoring. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute the statistic over compatible incident records and preserve unresolved incidents.
Display	Resolution and restoration can differ; show the declared endpoint. Recovery time does not require a dollar valuation.

Severity exposure adjusted incident cost change

Contract: METRIC-ROI-41.severity_exposure_adjusted_incident_cost_change.

Comparable incident cost/loss change adjusted under a frozen severity and exposure model.

Property	Contract
Calculation	current_standardized_incident_cost_or_loss - baseline_standardized_incident_cost_or_loss
Unit	currency/declared_exposure_unit
Required evidence	incident_cost_or_loss_records; severity_definition; independent_exposure_denominator; frozen_adjustment_method; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: incident_cost_or_loss_records, severity_definition, independent_exposure_denominator, frozen_adjustment_method, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Standardize both sides under one declared severity/exposure model before comparing; retain sparse-event uncertainty.
Display	A rare period with no incidents does not establish prevention. Missing exposure denominator withholds this comparison, not incident counts or observed costs.

42. Modeled risk reduction

METRIC-ROI-42. Change in event probability times loss severity across explicit scenarios, less control costs. Publish assumptions, ranges and residual risk; never label expected loss reduction as realized cash savings.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Exposure data + calibrated loss model; Incident, exposure, remediation, cost and loss-history records, plus qualified external loss references where useful

Evidence limits: Report observed incident economics and separately modeled expected loss; rare absent incidents alone do not establish prevention; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Baseline expected risk loss

Contract: METRIC-ROI-42.baseline_expected_risk_loss.

Modeled expected loss for the named reference scenario and horizon.

Property	Contract
Calculation	sum(event_probability_s * loss_severity_s)
Unit	currency
Required evidence	baseline_event_probabilities; baseline_loss_severities; exposure_scope; scenario_horizon; calibration_evidence
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Explicit reference scenario with calibrated probability/severity assumptions and dependency treatment.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_event_probabilities, baseline_loss_severities, exposure_scope, scenario_horizon, calibration_evidence. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Sum only mutually exclusive scenarios or expected losses whose overlap/dependencies are explicitly modeled.
Display	Publish probability, severity, horizon and ranges; not realized cash.

Residual expected risk loss

Contract: METRIC-ROI-42.residual_expected_risk_loss.

Modeled expected loss remaining with the stated controls.

Property	Contract
Calculation	sum(controlled_event_probability_s * controlled_loss_severity_s)
Unit	currency
Required evidence	controlled_event_probabilities; controlled_loss_severities; control_effect_assumptions; exposure_scope; scenario_horizon
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Same scenario boundary and horizon as the reference risk model.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: controlled_event_probabilities, controlled_loss_severities, control_effect_assumptions, exposure_scope, scenario_horizon. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Respect scenario dependencies and overlap; propagate parameter and model uncertainty.
Display	Show residual risk even when expected risk reduction is positive.

Risk control cost

Contract: METRIC-ROI-42.risk_control_cost.

Complete observed or explicitly assumed control cost under a declared cost boundary.

Property	Contract
Calculation	sum(unique_control_cost_components)
Unit	currency
Required evidence	control_cost_components; cost_evidence_class; cost_boundary; cost_completeness
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: control_cost_components, cost_evidence_class, cost_boundary, cost_completeness. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Preserve observed versus assumed input provenance. This modeled risk-ledger component is not itself a realized benefit.

Net expected risk reduction

Contract: METRIC-ROI-42.net_expected_risk_reduction.

Modeled expected loss reduction after complete control cost.

Property	Contract
Calculation	baseline_expected_risk_loss - residual_expected_risk_loss - risk_control_cost
Unit	currency
Required evidence	baseline_expected_risk_loss; residual_expected_risk_loss; risk_control_cost; scenario_dependency_and_overlap_model
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Compatible reference and control scenarios over one horizon.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: baseline_expected_risk_loss, residual_expected_risk_loss, risk_control_cost, scenario_dependency_and_overlap_model. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible expected losses and costs before differencing; propagate shared scenario uncertainty.
Display	Publish central estimate, assumption ranges and negative values. Never label expected-loss reduction as realized cash savings.

43. Governance, compliance and audit efficiency

METRIC-ROI-43. Time/cost per completed audit, evidence package, control review or policy exception; measure Venturi advisory evaluation availability, false positives and lost accepted work alongside savings; evaluate separately owned customer enforcement only with customer evidence.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Control evidence + process/time/cost records; Existing finance/control/audit workflow records, effort and accepted completion; advisory evaluation/decision traces and supported decision outcomes for availability, false positives and lost work; separate customer-enforcement evidence where assessed

Evidence limits: Measure process cost/time using the same recorded/inferred effort distinction; advisory availability and customer enforcement remain separate; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Governance effort per completed item

Contract: METRIC-ROI-43.governance_effort_per_completed_item.

Recorded or calibrated effort per completed audit, evidence package, control review or policy exception.

Property	Contract
Calculation	complete_governance_person_hours / accepted_completed_governance_item_count
Unit	labor_hour/process_item
Required evidence	governance_process_type; complete_recorded_or_calibrated_effort; effort_evidence_basis; accepted_completed_governance_item_count; completion_contract
Denominator	reference: accepted_completed_governance_item_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	baseline_estimate
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: governance_process_type, complete_recorded_or_calibrated_effort, effort_evidence_basis, accepted_completed_governance_item_count, completion_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Report each process type and effort evidence basis separately. A generated package is not automatically a completed accepted audit.

Governance cost per completed item

Contract: METRIC-ROI-43.governance_cost_per_completed_item.

Complete economic cost per accepted governance process item.

Property	Contract
Calculation	complete_governance_process_cost / accepted_completed_governance_item_count
Unit	currency/process_item
Required evidence	governance_process_type; complete_governance_process_cost; accepted_completed_governance_item_count; completion_contract
Denominator	reference: accepted_completed_governance_item_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_financial
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: governance_process_type, complete_governance_process_cost, accepted_completed_governance_item_count, completion_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show governance cost per completed item with its observed_financial classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Advisory evaluation availability

Contract: METRIC-ROI-43.advisory_evaluation_availability.

Eligible Venturi advisory evaluations receiving a supported timely result under the advisory contract.

Property	Contract
Calculation	eligible_evaluations_with_timely_supported_result / eligible_advisory_evaluations
Unit	ratio
Required evidence	advisory_evaluation_traces; eligible_advisory_evaluations; advisory_result_contract; evaluation_deadline
Denominator	reference: eligible_advisory_evaluations; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: advisory_evaluation_traces, eligible_advisory_evaluations, advisory_result_contract, evaluation_deadline. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Venturi advice availability is distinct from separately owned customer enforcement. Advisory failure does not authorize blocking customer work.

Advisory false positive rate

Contract: METRIC-ROI-43.advisory_false_positive_rate.

Independently validated false adverse advisory decisions divided by evaluated benign eligible decisions.

Property	Contract
Calculation	validated_false_adverse_advisories / independently_validated_benign_eligible_decisions
Unit	ratio
Required evidence	advisory_decision_traces; independent_decision_labels; benign_eligible_decision_count; sampling_and_label_contract
Denominator	reference: independently_validated_benign_eligible_decisions; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: advisory_decision_traces, independent_decision_labels, benign_eligible_decision_count, sampling_and_label_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Keep false-positive-rate denominator distinct from the fraction of alerts judged false. Unknown labels do not become negatives.

Supported lost accepted work

Contract: METRIC-ROI-43.supported_lost_accepted_work.

Accepted business work lost under a supported advisory or separately owned enforcement consequence contract.

Property	Contract
Calculation	count_unique(work_items_with_supported_accepted_work_loss)
Unit	accepted_outcome
Required evidence	decision_to_work_consequence_links; independent_acceptance_evidence; loss_attribution_contract; decision_owner
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Supported alternative accepted result or validated consequence evidence; potential lost work otherwise remains a scenario.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: decision_to_work_consequence_links, independent_acceptance_evidence, loss_attribution_contract, decision_owner. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Name whether the consequence comes from Venturi advice, a user decision or customer enforcement. Advisory evaluation alone does not enforce.

Customer enforcement availability

Contract: METRIC-ROI-43.customer_enforcement_availability.

Availability of separately owned customer enforcement only where the customer provides its own execution evidence.

Property	Contract
Calculation	eligible_customer_enforcement_actions_meeting_contract / eligible_customer_enforcement_actions
Unit	ratio
Required evidence	customer_enforcement_execution_records; customer_control_contract; eligible_customer_enforcement_actions; customer_owner_identity
Denominator	reference: eligible_customer_enforcement_actions; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	coverage
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: customer_enforcement_execution_records, customer_control_contract, eligible_customer_enforcement_actions, customer_owner_identity. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Do not infer customer enforcement from Venturi advisory traces. Missing customer evidence leaves only this dependent output unavailable.

Governance process effort improvement

Contract: METRIC-ROI-43.governance_process_effort_improvement.

Comparable reduction in complete governance effort for the same accepted process scope.

Property	Contract
Calculation	baseline_governance_person_hours - current_governance_person_hours
Unit	labor_hour
Required evidence	governance_process_type; recorded_or_calibrated_baseline_and_current_effort; effort_evidence_basis; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	An AI-attributable effect requires supported randomized or identified observational comparison over the complete eligible cohort, including failures and non-adopters. Otherwise publish a descriptive comparison or scenario under a distinct label.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: governance_process_type, recorded_or_calibrated_baseline_and_current_effort, effort_evidence_basis, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Show process scope and added review/lost work alongside savings. Any monetary benefit requires separate cash evidence or labeled capacity valuation.

Decision-time policy breach rate

Contract: METRIC-ROI-43.decision_time_policy_breach_rate.

Use decision-policy-evaluation identity for both numerator and denominator. Count an evaluation with several breach findings once; preserve policy revisions and advisory versus customer-opted-in enforcement states.

Property	Contract
Calculation	unique_eligible_decision_policy_evaluations_with_breach / eligible_decision_policy_evaluation_count
Unit	ratio
Required evidence	versioned_policy_evaluation_records; unique_eligible_decision_policy_evaluations_with_breach; eligible_decision_policy_evaluation_count; evaluation_coverage_and_effective_policy
Denominator	reference: eligible_decision_policy_evaluation_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	observed_operational
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Missing evaluation coverage or denominator makes the full-scope rate unavailable. Supported breach and evaluation counts remain separate; absence of a result is not a passing policy evaluation.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

Control evidence coverage

Contract: METRIC-ROI-43.control_evidence_coverage.

Evaluate current evidence sufficiency for each applicable workflow/control requirement, grouped by declared risk tier. Missing, stale, unsupported and not-applicable states remain distinct; evidence coverage does not establish compliance certification or control effectiveness.

Property	Contract
Calculation	eligible_workflow_control_requirements_with_current_sufficient_evidence / eligible_workflow_control_requirement_count
Unit	ratio
Required evidence	versioned_workflow_control_requirement_inventory; risk_tier_mapping; evidence_sufficiency_and_freshness_rules; eligible_workflow_control_requirement_count; workflow_control_evidence_records_with_validity
Denominator	reference: eligible_workflow_control_requirement_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	coverage
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	Unknown applicability inventory or incomplete denominator withholds coverage. Retain individual evidence states; untested controls cannot become passing controls.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

44. Energy efficiency per accepted outcome

METRIC-ROI-44. Measured or modeled kWh / accepted outcomes, including the declared serving boundary and allocation method; show total energy as well as intensity.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Power telemetry or qualified provider factors + outcomes; Relevant power/site/provider/environmental evidence, allocation boundaries, accepted output and qualified factor provenance

Evidence limits: Report measured or modeled energy/carbon/water intensity and totals; no universal token-to-physical-resource conversion; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Measured energy total

Contract: METRIC-ROI-44.measured_energy_total.

Measured energy consumption allocated to the declared serving boundary and period.

Property	Contract
Calculation	sum(measured_energy_kwh_allocated_once_within_boundary)
Unit	kWh
Required evidence	energy_meter_or_qualified_measured_provider_records; physical_serving_boundary; allocation_policy; measurement_period
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: energy_meter_or_qualified_measured_provider_records, physical_serving_boundary, allocation_policy, measurement_period. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show physical boundary, direct measurement provenance and allocation. Provider API usage alone does not reveal provider hardware energy.

Modeled energy total

Contract: METRIC-ROI-44.modeled_energy_total.

Estimated energy consumption from qualified factors under the declared physical boundary.

Property	Contract
Calculation	sum(compatible_activity_quantity * qualified_energy_factor) + disjoint_modeled_boundary_energy
Unit	kWh
Required evidence	compatible_activity_quantities; qualified_energy_factors; factor_location_time_model_provenance; physical_boundary; allocation_and_uncertainty_model
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Qualified activity-to-energy factors for the actual workload boundary; no universal token-to-energy conversion.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: compatible_activity_quantities, qualified_energy_factors, factor_location_time_model_provenance, physical_boundary, allocation_and_uncertainty_model. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Sum compatible allocated physical energy and propagate shared factor uncertainty.
Display	Label modeled energy, factor vintage and uncertainty. Exact token counts do not make estimated energy measured.

Measured energy per accepted outcome

Contract: METRIC-ROI-44.measured_energy_per_accepted_outcome.

Measured declared-boundary energy per unique mature accepted business outcome.

Property	Contract
Calculation	measured_energy_total / unique_mature_accepted_outcome_count
Unit	kWh/accepted_outcome
Required evidence	measured_energy_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_energy_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: measured_energy_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_energy_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show absolute measured energy alongside intensity and preserve serving boundary.

Modeled energy per accepted outcome

Contract: METRIC-ROI-44.modeled_energy_per_accepted_outcome.

Modeled declared-boundary energy per unique mature accepted business outcome.

Property	Contract
Calculation	modeled_energy_total / unique_mature_accepted_outcome_count
Unit	kWh/accepted_outcome
Required evidence	modeled_energy_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_energy_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	modeled_scenario
Baseline	Qualified factor model retained from modeled_energy_total.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: modeled_energy_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_energy_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Label modeled intensity and publish total energy, allocation, factor provenance and uncertainty.

Comparable energy benchmark index

Contract: METRIC-ROI-44.energy_benchmark_index.

Index measured or modeled energy intensity against a declared compatible reference using the same work unit, accepted-quality requirement, model/hardware boundary and period. Keep measured and modeled bases separate and retain the underlying intensity.

Property	Contract
Calculation	qualified_energy_per_work_unit / positive_compatible_reference_energy_per_work_unit
Unit	ratio
Required evidence	qualified_energy_per_work_unit; positive_compatible_reference_energy_per_work_unit; matched_workload_quality_and_system_boundary; factor_versions_and_uncertainty
Denominator	reference: positive_compatible_reference_energy_per_work_unit; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	modeled_scenario
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	An absent, nonpositive or non-comparable reference withholds the index while qualified underlying energy observations remain available. External factors cannot create a customer performance claim.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

45. Carbon efficiency and avoided emissions

METRIC-ROI-45. Energy and other included lifecycle emissions allocated to accepted outcomes; compare the same workload boundary and publish location/time/factor uncertainty. Monetize only actual energy or carbon expense under an applicable cost policy.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Energy/factor provenance + boundaries + outcomes; Relevant power/site/provider/environmental evidence, allocation boundaries, accepted output and qualified factor provenance

Evidence limits: Report measured or modeled energy/carbon/water intensity and totals; no universal token-to-physical-resource conversion; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Measured carbon total

Contract: METRIC-ROI-45.measured_carbon_total.

Directly measured emissions allocated to the explicitly included physical lifecycle boundary.

Property	Contract
Calculation	sum(directly_measured_emissions_allocated_once_within_boundary)
Unit	kgCO2e
Required evidence	direct_emissions_measurement_records; included_lifecycle_boundary; location_and_time; allocation_policy; measurement_unit_conversion
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: direct_emissions_measurement_records, included_lifecycle_boundary, location_and_time, allocation_policy, measurement_unit_conversion. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Use this output only for direct measured-emission evidence. Energy multiplied by emissions factors belongs in modeled_carbon_total.

Modeled carbon total

Contract: METRIC-ROI-45.modeled_carbon_total.

Estimated energy-related and other included lifecycle emissions under qualified location/time factors.

Property	Contract
Calculation	sum(allocated_energy_kwh * qualified_emissions_factor_kgco2e_per_kwh) + disjoint_other_included_lifecycle_emissions
Unit	kgCO2e
Required evidence	allocated_energy_and_evidence_basis; qualified_location_time_emissions_factors; other_included_lifecycle_emissions; lifecycle_boundary; factor_provenance_and_uncertainty
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Qualified factors and declared lifecycle boundary; retain measured-versus-modeled energy provenance.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: allocated_energy_and_evidence_basis, qualified_location_time_emissions_factors, other_included_lifecycle_emissions, lifecycle_boundary, factor_provenance_and_uncertainty. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Sum nonoverlapping included emissions and propagate shared factor/allocation uncertainty.
Display	Show factor vintage, location/time, included lifecycle categories and uncertainty. Do not treat estimates as direct measurements.

Measured carbon per accepted outcome

Contract: METRIC-ROI-45.measured_carbon_per_accepted_outcome.

Directly measured declared-boundary emissions per mature accepted outcome.

Property	Contract
Calculation	measured_carbon_total / unique_mature_accepted_outcome_count
Unit	kgCO2e/accepted_outcome
Required evidence	measured_carbon_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_emissions_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: measured_carbon_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_emissions_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show measured carbon per accepted outcome with its observed_operational classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Modeled carbon per accepted outcome

Contract: METRIC-ROI-45.modeled_carbon_per_accepted_outcome.

Modeled lifecycle emissions per mature accepted outcome under the declared boundary.

Property	Contract
Calculation	modeled_carbon_total / unique_mature_accepted_outcome_count
Unit	kgCO2e/accepted_outcome
Required evidence	modeled_carbon_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_emissions_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	modeled_scenario
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: modeled_carbon_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_emissions_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Label modeled intensity and display absolute total, factor provenance and boundary.

Avoided emissions estimate

Contract: METRIC-ROI-45.avoided_emissions_estimate.

Signed difference in comparable lifecycle emissions for the same accepted workload and qualified alternative.

Property	Contract
Calculation	comparable_baseline_emissions - current_comparable_emissions
Unit	kgCO2e
Required evidence	comparable_baseline_emissions; current_comparable_emissions; same_workload_lifecycle_boundary; acceptance_quality_and_constraint_parity; comparison_contract
Denominator	Not applicable
Evidence class	baseline_estimate
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes. Match physical lifecycle, location/time and workload boundaries.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: comparable_baseline_emissions, current_comparable_emissions, same_workload_lifecycle_boundary, acceptance_quality_and_constraint_parity, comparison_contract. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Recompute each comparator and current result from compatible non-overlapping components before subtraction; aggregate signed effects only over compatible, non-overlapping benefit or work scopes.
Display	Keep measured/model input provenance and uncertainty. Routing comparisons require quality, security, residency and latency qualification; price-only alternatives cannot establish avoided emissions.

Realized energy expense reduction

Contract: METRIC-ROI-45.realized_energy_expense_reduction.

Supported attributable reduction in actual energy expense for the declared comparable boundary.

Property	Contract
Calculation	baseline_comparable_energy_expense - actual_energy_expense - disjoint_energy_intervention_cost
Unit	currency
Required evidence	energy_invoices_or_applicable_billed_energy_cost; comparison_contract; disjoint_energy_intervention_cost; benefit_overlap_map
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: energy_invoices_or_applicable_billed_energy_cost, comparison_contract, disjoint_energy_intervention_cost, benefit_overlap_map. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Physical emissions reduction is not itself cash savings. Do not add energy expense reduction twice to serving cost or procurement benefits.

Realized carbon expense reduction

Contract: METRIC-ROI-45.realized_carbon_expense_reduction.

Supported reduction in actual applicable carbon-related charges under the customer’s cost policy.

Property	Contract
Calculation	baseline_applicable_carbon_expense - actual_applicable_carbon_expense - disjoint_carbon_intervention_cost
Unit	currency
Required evidence	applicable_carbon_cost_policy; dated_actual_carbon_charges; comparison_contract; disjoint_carbon_intervention_cost
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: applicable_carbon_cost_policy, dated_actual_carbon_charges, comparison_contract, disjoint_carbon_intervention_cost. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	An internal shadow carbon price is a scenario, not an actual charge or realized saving.

Internal carbon price scenario value

Contract: METRIC-ROI-45.internal_carbon_price_scenario_value.

Explicit modeled valuation of emissions change under an internal carbon-price assumption.

Property	Contract
Calculation	avoided_emissions_estimate * assumed_internal_price_per_kgco2e
Unit	currency
Required evidence	avoided_emissions_estimate; assumed_internal_price_per_kgco2e; scenario_cost_policy; scenario_horizon
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Supported physical comparison plus explicit internal price assumption.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: avoided_emissions_estimate, assumed_internal_price_per_kgco2e, scenario_cost_policy, scenario_horizon. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Label shadow-price scenario; exclude from realized cash ROI and do not add to actual carbon-charge savings for the same effect.

Comparable carbon benchmark index

Contract: METRIC-ROI-45.carbon_benchmark_index.

Index compatible carbon intensity against a declared reference with matched work unit, quality, system boundary, regional/time factors and measurement basis. Retain factor revisions and uncertainty separately from operational change.

Property	Contract
Calculation	qualified_emissions_per_work_unit / positive_compatible_reference_emissions_per_work_unit
Unit	ratio
Required evidence	qualified_emissions_per_work_unit; positive_compatible_reference_emissions_per_work_unit; matched_workload_quality_and_system_boundary; factor_versions_and_uncertainty
Denominator	reference: positive_compatible_reference_emissions_per_work_unit; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared entity or work unit, population, period, maturity, policy and value basis.
Evidence class	modeled_scenario
Baseline	A declared compatible reference is required wherever the formula compares periods, paths, targets or benchmarks. An observed level otherwise requires no baseline.
Causal interpretation	Preserve descriptive, estimated and modeled evidence classes. Association, target attainment and source coverage do not establish an AI-caused financial or operational effect.
Missing evidence	An absent, nonpositive or incompatible reference withholds the index. Missing regional or emissions factors remains unknown, and the index does not prove avoided emissions or renewable-energy use.
Aggregation	Resolve each named output independently. Combine compatible disjoint components before division, never average ratios or percentile aggregates. Preserve signed differences and source coverage.
Display	Display the named unit, population, tenant-authorized grouping, period, source and method revisions, evidence state, coverage, numerator and denominator where applicable, and missing-evidence reason.

46. Water and other resource efficiency

METRIC-ROI-46. Measured or explicitly modeled water consumed/withdrawn per accepted outcome and absolute totals, with location and allocation. Keep withdrawals distinct from consumption; monetize only applicable resource costs.

Readable evidence: Dated tenant, identity, event-time ownership, workflow and reporting-cohort contract; Provider/site resource evidence + boundary + outcomes; Relevant power/site/provider/environmental evidence, allocation boundaries, accepted output and qualified factor provenance

Evidence limits: Report measured or modeled energy/carbon/water intensity and totals; no universal token-to-physical-resource conversion; Read existing customer records; no new customer instrumentation or customer-system writes. Source capability does not prove deployed runtime support.; Preserve output-specific availability, negative results and separate financial realization, causal validity, labor uncertainty and ownership confidence.

Measured water consumed total

Contract: METRIC-ROI-46.measured_water_consumed_total.

Directly measured water consumed, distinct from water withdrawn, within the declared site and allocation boundary.

Property	Contract
Calculation	sum(measured_consumed_water_allocated_once)
Unit	liter
Required evidence	measured_water_consumption_records; site_location_and_period; physical_boundary; allocation_policy
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: measured_water_consumption_records, site_location_and_period, physical_boundary, allocation_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Consumption excludes returned water under the stated definition; never substitute withdrawal.

Modeled water consumed total

Contract: METRIC-ROI-46.modeled_water_consumed_total.

Modeled water consumption under qualified site/provider evidence and compatible allocation factors.

Property	Contract
Calculation	sum(compatible_activity_quantity * qualified_consumptive_water_factor)
Unit	liter
Required evidence	compatible_activity_quantities; qualified_consumptive_water_factors; site_location_time_and_factor_provenance; physical_boundary; allocation_and_uncertainty_model
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Qualified site/provider consumptive factors; no universal token-to-water conversion.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: compatible_activity_quantities, qualified_consumptive_water_factors, site_location_time_and_factor_provenance, physical_boundary, allocation_and_uncertainty_model. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Label modeled consumption, location and uncertainty; preserve the distinction from withdrawal.

Measured water withdrawn total

Contract: METRIC-ROI-46.measured_water_withdrawn_total.

Directly measured gross water withdrawal within the declared site and boundary.

Property	Contract
Calculation	sum(measured_withdrawn_water_allocated_once)
Unit	liter
Required evidence	measured_water_withdrawal_records; site_location_and_period; physical_boundary; allocation_policy
Denominator	Not applicable
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: measured_water_withdrawal_records, site_location_and_period, physical_boundary, allocation_policy. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Withdrawal and consumption are distinct physical flows and must not be added as a single water total.

Modeled water withdrawn total

Contract: METRIC-ROI-46.modeled_water_withdrawn_total.

Modeled water withdrawal under qualified site/provider evidence.

Property	Contract
Calculation	sum(compatible_activity_quantity * qualified_withdrawal_water_factor)
Unit	liter
Required evidence	compatible_activity_quantities; qualified_withdrawal_water_factors; site_location_time_and_factor_provenance; physical_boundary; allocation_and_uncertainty_model
Denominator	Not applicable
Evidence class	modeled_scenario
Baseline	Qualified withdrawal factors; no universal token-to-water conversion.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: compatible_activity_quantities, qualified_withdrawal_water_factors, site_location_time_and_factor_provenance, physical_boundary, allocation_and_uncertainty_model. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Show modeled water withdrawn total with its modeled_scenario classification, period, unit, source provenance, coverage and scope. Preserve negative values where the formula allows them; keep measured, estimated and modeled evidence separate.

Measured water consumed per accepted outcome

Contract: METRIC-ROI-46.measured_water_consumed_per_accepted_outcome.

Declared-boundary water intensity with consumption/withdrawal and measured/modeled semantics preserved.

Property	Contract
Calculation	measured_water_consumed_total / unique_mature_accepted_outcome_count
Unit	liter/accepted_outcome
Required evidence	measured_water_consumed_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_resource_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: measured_water_consumed_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_resource_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show absolute total alongside intensity, location, period, boundary, allocation and evidence class. Never merge consumed and withdrawn water.

Modeled water consumed per accepted outcome

Contract: METRIC-ROI-46.modeled_water_consumed_per_accepted_outcome.

Declared-boundary water intensity with consumption/withdrawal and measured/modeled semantics preserved.

Property	Contract
Calculation	modeled_water_consumed_total / unique_mature_accepted_outcome_count
Unit	liter/accepted_outcome
Required evidence	modeled_water_consumed_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_resource_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	modeled_scenario
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: modeled_water_consumed_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_resource_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show absolute total alongside intensity, location, period, boundary, allocation and evidence class. Never merge consumed and withdrawn water.

Measured water withdrawn per accepted outcome

Contract: METRIC-ROI-46.measured_water_withdrawn_per_accepted_outcome.

Declared-boundary water intensity with consumption/withdrawal and measured/modeled semantics preserved.

Property	Contract
Calculation	measured_water_withdrawn_total / unique_mature_accepted_outcome_count
Unit	liter/accepted_outcome
Required evidence	measured_water_withdrawn_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_resource_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	observed_operational
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: measured_water_withdrawn_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_resource_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show absolute total alongside intensity, location, period, boundary, allocation and evidence class. Never merge consumed and withdrawn water.

Modeled water withdrawn per accepted outcome

Contract: METRIC-ROI-46.modeled_water_withdrawn_per_accepted_outcome.

Declared-boundary water intensity with consumption/withdrawal and measured/modeled semantics preserved.

Property	Contract
Calculation	modeled_water_withdrawn_total / unique_mature_accepted_outcome_count
Unit	liter/accepted_outcome
Required evidence	modeled_water_withdrawn_total; unique_mature_accepted_outcome_count; cohort_acceptance_and_maturity_contract; matched_resource_outcome_cohort
Denominator	reference: unique_mature_accepted_outcome_count; required state: finite_positive_compatible; unavailable reasons: missing_denominator; nonfinite_denominator; nonpositive_denominator; incompatible_cohort_or_basis; compatible rule: Same tenant, declared cohort, period, unit, acceptance version and cost or labor basis.
Evidence class	modeled_scenario
Baseline	No baseline is required for this observed level. Any claimed improvement requires a separately identified comparable baseline.
Causal interpretation	No causal claim from this observation; do not infer AI attribution from ownership, temporal proximity or association.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: modeled_water_withdrawn_total, unique_mature_accepted_outcome_count, cohort_acceptance_and_maturity_contract, matched_resource_outcome_cohort. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate compatible, non-overlapping numerator components and denominators before division; never average ratios or count the same business object twice.
Display	Show absolute total alongside intensity, location, period, boundary, allocation and evidence class. Never merge consumed and withdrawn water.

Realized water resource expense reduction

Contract: METRIC-ROI-46.realized_water_resource_expense_reduction.

Attributable reduction in actual applicable water or declared resource expense, net of disjoint intervention cost.

Property	Contract
Calculation	baseline_applicable_resource_expense - actual_applicable_resource_expense - disjoint_resource_intervention_cost
Unit	currency
Required evidence	resource_type_and_unit; applicable_resource_tariff_or_contract; dated_resource_expense_records; comparison_contract; disjoint_resource_intervention_cost
Denominator	Not applicable
Evidence class	observed_financial
Baseline	Use a frozen comparator with compatible task mix, role, accepted-work scope, quality, maturity, period and economic boundary; retain uncertainty and known concurrent changes.
Causal interpretation	Reconcile dated financial evidence and a supported attributable comparison, preserve the complete investment scope, exclude duplicate benefits and costs, and retain negative results.
Missing evidence	Publish unavailable for this output when any required dependency is missing, nonfinite or incompatible: resource_type_and_unit, applicable_resource_tariff_or_contract, dated_resource_expense_records, comparison_contract, disjoint_resource_intervention_cost. Retain available sibling outputs; never replace missing evidence with zero.
Aggregation	Aggregate only compatible, non-overlapping scoped components; preserve signed differences and the declared cohort boundary.
Display	Monetize only applicable resource charges. Do not infer cash savings from a physical intensity reduction alone.

Adoption

Adoption intelligence (Adoption / Adoption intelligence) reads the same attribution graph for adoption: which parts of your organization have genuinely embedded AI, and which lag. It is the second projection of the graph (no second product, no new data collection) and it answers the question a CTO, CHRO, or program sponsor brings to a board: is our AI investment actually taking hold, and where are the enablement gaps?

Cohort-only, always

Adoption is reported at the cohort level (by team, function, role, or tenure), never as individual surveillance.

Privacy is built in, not bolted on

• Minimum cohort size is k = 5. Any cohort smaller than five is suppressed and rolled into its parent unit, never errored, never shown. A suppressed cohort can never be reconstructed by differencing overlapping cohorts. An Admin may raise the threshold; it can never be lowered below 5.
• Individual-level views do not exist. The server enforces the minimum cohort size of 5 on every cohort read path (adoption views, dashboards, filters, drill-downs, exports, API responses, board packs, and webhooks); no role or setting can bypass that invariant.
• Venturi never acts against a person. It never notifies, blocks, or enforces against a resolved individual. Employees whose usage is attributed have full data-subject rights and are not platform users.

This cohort-only posture is what keeps adoption intelligence out of profiling, and it is non-negotiable. For the privacy commitments behind it, see data privacy & retention and data-subject rights; for the read-only, no-content boundary it sits inside, see trust & security.

What you can see and do

• Adoption depth by cohort: how deeply each team, function, or tenure band has adopted AI, with the same signal-type basis everywhere: whether a figure is direct, inferred, or estimated.
• Real underuse vs. visibility limits: the surface distinguishes a team that genuinely is not using AI from a team Venturi cannot see yet, so you act on reality, not on a coverage gap.
• Enablement gaps: the cohorts where adoption lags, so enablement budget can follow traction instead of guesswork.
• Governed cohort reporting: schedule a saved adoption view as a report, rendered under each recipient’s access with cohorts below five suppressed and EU rows respected.

Read the basis before you act

An adoption figure marked inferred or estimated is a directional signal, not a hard count. The surface always names the basis so a leader presenting to the board distinguishes a measured fact from a modeled estimate, and never mistakes a visibility limit for genuine underuse.

Alerts and budgets

Both surfaces feed your notifications: you can alert on a budget threshold, a spend anomaly, a new high-value optimization opportunity, or an adoption shift. Alerts reach the in-app notification center and, per preference, email, Slack, Microsoft Teams, or a signed webhook. Admins set tenant notification routing under Settings / Admin / Notifications. Budgets and the policies behind these alerts are configured in governance & exports.

Next steps

• Use Governance & exports to set budgets, govern policies, and export the savings and adoption numbers.
• Use Optimization & governance to understand the full savings model and verified-savings basis.
• Use Adoption intelligence to see the capability end to end.
• Use Coverage & unknowns to make sure adoption reflects reality, not a data gap.

Give feedback on this page

Sign in to send feedback. Opens in a new tab.

Copy page details

Docs feedback: Optimization & adoption
Page: https://docs.venturi.systems/dashboard/optimization-and-adoption/