Glossary¶
Canonical terms used across this site, the product, and Venturi’s governed documentation. When a page and this glossary disagree, treat the glossary as the intended vocabulary and tell us.
Attribution¶
- InvocationEvent
- The metadata-only ingestion record for 1 AI invocation. It contains identifiers, model, token, cost, timing, and correlation fields, never prompt or completion bodies. See Ingestion API.
- AttributionRecord
- The versioned, durable result published by platform materialization for an invocation. It carries the resolved
graph edges, output state, stage origin, operational confidence (
coper), and evidence metadata used by dashboards, exports, and the API. See Reading attribution. - Attribution graph
- The connected representation of AI usage and organizational responsibility. Venturi resolves usage across six layers: model invocation, service, code ownership, identity, organizational hierarchy, and budget.
- RAIL (Resource Attribution & Inference Layer)
- Read letter by letter: Resource, Attribution, Inference,
Layer. RAIL is Stage B of Venturi’s three-stage attribution pipeline,
which turns an
InvocationEventinto anAttributionRecord. The Venturi platform runs Stage A deterministic reconciliation (rules R1 to R5). RAIL runs Stage B trained edge-existence inference on the eligible edges that Stage A leaves unresolved and returns an uncapped posterior confidence (cpost) with evidence and alternatives. A separate allocation subsystem runs Stage C fractional allocation, and platform materialization applies the 0.95 confidence cap. RAIL’s training machinery supplies the verified Stage B models; it is part of RAIL, not a separate product. See The attribution pipeline. - Attribution coverage
- The share of capturable AI usage that resolves through the required attribution chain at or above the confidence floor. The objective is 0.95 or higher; the contracted floor depends on your support tier. Venturi reports coverage against a stated denominator; it never claims complete visibility.
- Pathway
- A distinct route by which AI usage reaches your organization (managed provider APIs, cloud-hosted model services, self-hosted models, AI embedded in SaaS, automation platforms). Each pathway has its own capture mechanism.
- Pathway coverage
- The share of observed activity assigned to a named pathway category, with a 100% target. It measures classification, not ownership: a fully classified estate can still hold unresolved spend. See Known limitations.
- Unknown / Shared / Unresolved
- Explicit attribution outcomes. Unknown means the evidence does not identify an owner; Shared means multiple owners legitimately split responsibility; Unresolved means conflicting evidence is awaiting reconciliation. None of these are hidden, zeroed, or silently guessed.
- Multi-truth model
- Usage, cost, and ownership are recorded as distinct truths. A single event can be certain in one dimension and provisional in another.
- Ownership vs allocation
- Two distinct relationships in the attribution graph: ownership is a direct single-owner edge, allocation is a fractional split across several owners. Keeping them separate stops shared responsibility from collapsing into a false single-owner answer.
Confidence¶
- Confidence score (
coper) - The operational confidence attached to every attribution result, from 0 to 1 and capped at 0.95 as a deliberate policy ceiling. It is never called accuracy or certainty. See Confidence and evidence.
attribution_confidence - The descriptive API name for
coper; it is not a second confidence object. In an attribution record’s evidence card the same value appears asinterpretation..confidence - Confidence bands
- Chargeback-ready (
chargeback_ready,coper >= 0.80), Provisional (provisional,0.50 <= coper < 0.80, labeled and never auto-billed), and Estimated (estimated,coper < 0.50, excluded from chargeback). - Evidence card
- The per-result explanation of which signals produced an attribution and at what confidence, available wherever a result is shown.
- Posterior confidence (
cpost) - The uncapped model-internal confidence preserved for training, audit, and
analysis. It is never shown on operational surfaces, which display only the
capped
coper.
Enforcement posture¶
- Decision-time interceptor
- The fail-open component on the AI request path that observes usage at decision time with a fast index lookup, within a 50 ms P99 end-to-end budget (15 ms internal). A cooperative timeout against a cumulative deadline bounds it: on any fault or timeout it forwards the request unmodified and records null attribution. It never runs trained inference and never blocks customer AI traffic.
- Passive
- The observe-only governance mode. Budgets and policies record state and raise alerts; they take no action.
- Advisory (Shape)
- The default governance mode. Venturi raises alerts and recommendations; your systems and people decide.
- Active (Gate)
- The explicit, opt-in enforcement mode, enabled only for a named, operationally trusted workload. Venturi issues a signed enforcement recommendation (deny or reroute) that your enforcement point (gateway, proxy policy, or budget control) decides whether to apply; a budget hard stop uses the same mechanism. The decision-time interceptor never denies, blocks, or reroutes traffic.
- Fail-open / fail-closed
- The interceptor data path fails open: a request that reaches the decision-time interceptor is forwarded unmodified when a Venturi dependency behind it is slow, degraded, or unavailable. With the drop-in proxy, the proxy endpoint’s own reachability stays on your request path (see Proxy endpoint reachability). Everything else fails closed: authentication, authorization, tenant isolation, admin actions, exports, and billing.
- Policy evaluation record
- The audit event written when a governance policy evaluates an invocation or attributed record, making advisory and enforcement behavior reviewable after the fact.
- Manual override record
- The audit event recording a human override of an attribution or policy outcome, preserving the intervention as evidence rather than mutable state.
Deployment¶
- Control plane
- The vendor-managed surface used for licensing, signed updates, configuration metadata, and governed orchestration. Deliberately separate from the data plane, so raw telemetry stays in your dedicated data plane (in your cloud when self-hosted, or in your contracted residency lane in SaaS mode) and never passes through the control plane.
- Data plane
- The dedicated single-tenant runtime containing ingestion, processing, graph state, materialization, and serving. It runs in the customer’s cloud in self-hosted mode or in the contracted residency lane in SaaS mode.
- Customer trust boundary
- The boundary inside which customer telemetry, event processing, graph state, and materialized indexes remain unless explicitly exported.
- Read-only integrations
- Customer system connectors default to read behavior (get, list, describe) and never mutate your systems; a platform gate blocks any connector module that requests write access from shipping.
- Workload Identity Federation (WIF)
- Keyless workload authentication in which a trusted external identity exchanges a signed assertion for short-lived cloud credentials. Venturi uses it by default for Google Cloud and Azure onboarding, so cloud onboarding needs no long-lived cloud credentials by default. The exception is the Azure client-secret fallback, which has a 90-day TTL. See Rotate & offboard.
- External ID
- The customer-specific value Venturi must present when assuming an AWS cross-account role. It mitigates the confused-deputy problem; a role ARN alone is insufficient. Rotation is recommended every 12 months.
- Plan
- The commercial subscription named on your order form: its included quantities, feature entitlements, and contractual ceilings, including the role and feature ceiling that custom roles resolve against and the default API rate limits. It is separate from the support tier (Standard, Premium, or Regulated), which sets support response and service commitments. See Your plan.
Cost and value¶
- Chargeback-eligible
- A cost line whose attribution confidence is at or above the 0.80 floor. Lines below the floor are labeled Provisional or Estimated and are never auto-billed.
- Eco-efficiency
- Capability delivered per unit of environmental cost. It is distinct from the Model Energy Rating and is null when any governed capability or energy input is unknown.
- Identified savings vs realized savings
- An identified saving is an opportunity the platform has found; a realized saving is one you have captured, verified against a frozen baseline. The two are never conflated.
- TRAC
- Total Risk-Adjusted Cost: billed cost plus a confidence risk premium plus carbon liability, which adjusts spend for the uncertainty and externality context of the usage behind it.
- Cost derivation method
- The per-record field describing how a cost value was derived: direct billing evidence, calibrated estimation, or another explicit method, so measured spend is never conflated with inferred spend.
- FOCUS export
- Cost export pinned to the FinOps Open Cost and Usage Specification v1.2: 64 columns, made up of 57 base columns and 7 Venturi extension columns. See Reporting and exports.
Value metrics¶
The AI return measurement guide defines the common evidence contract. Every result identifies its work unit, team, period, acceptance rule, cost scope, comparison and method version.
- Accepted outcome
- A unique business work item that meets the customer’s versioned acceptance rule and maturity window. Existing customer records supply acceptance, rejection, reopening and revocation. An HTTP success or model response alone does not establish business acceptance.
- Fully loaded cost per accepted outcome
- All included workflow costs divided by unique accepted outcomes. Costs include failed attempts, retries, tools, review, rework and allocated support within the declared boundary. A zero or unknown accepted count makes the ratio unavailable. Costs and acceptance evidence remain visible.
- AI cost per successful outcome
- The narrower AI-spend numerator divided by unique accepted outcomes. This preserves the existing cost-per-successful-outcome API meaning. The result identifies its AI-only scope and cannot be presented as fully loaded workflow cost.
- Incremental net benefit
- Attributable incremental financial benefit before included costs, minus complete incremental investment and operating cost. Unknown cost leaves supported components visible; a net total needs complete cost evidence or an independently evidenced complete already-net change.
- Workflow financial ROI
- Incremental net benefit divided by the same complete incremental investment. The denominator must be known, finite and positive. Negative ROI remains visible. A complete already-net numerator still uses the full investment denominator. Revenue, capacity value and modeled risk reduction remain distinct from realized financial benefit.
- AI-spend ROI
- The existing
roi = net_ai_value / ai_costAPI measure. Its AI-spend denominator has a narrower scope than workflow financial ROI. The displayed name, formula, included costs and version must identify that scope. No alias silently changes historical results into fully loaded financial returns. - Gross AI value / net AI value
- Existing cost/value-waterfall terms. Gross value precedes the costs included by its versioned contract; net value subtracts those costs. Components retain their measured-financial, measured-productivity, inferred-productivity or self-reported basis. An aggregate containing valued capacity or modeled risk cannot be labeled realized cash benefit. Attribution confidence is never a monetary multiplier.
- E-FTE (equivalent full-time capacity)
- Net human-work-equivalent hours divided by an explicit period- and role-compatible reference-hours denominator. Comparable team productivity or task effort establishes the baseline; all current human effort and disjoint overhead are deducted. Paid hours, worked hours and active effort are different bases. There is no universal monthly or annual hours constant.
- Gross work equivalence / net capacity
- Gross equivalence is the baseline labor required for the accepted task mix. Net capacity deducts current human effort and overhead. Producing more accepted work at unchanged staffing can establish output-equivalent capacity; it does not establish hours removed from existing work.
- Released effort / redeployment
- Released effort is a supported reduction in labor needed for comparable existing work. Redeployment is evidence that capacity was used elsewhere. These are separate from output equivalence, payroll reductions and hypothetical future hiring avoidance.
- Quality-preserving savings
- A comparable cost improvement subject to fixed acceptance and quality guardrails, including later defects and rework. The older “quality-adjusted savings” label must identify any declared adjustment. An arbitrary evaluator score or ownership confidence cannot scale money.
- Evidence coverage / causal evidence
- Coverage describes how much eligible work has the required observations, joins and comparisons. It is distinct from correctness and from the proportion of benefit caused by AI. Observational before/after results remain descriptive unless their causal identification assumptions are supported. Missing, unresolved, immature and revoked evidence remain explicit.
Metric definitions¶
The original 70 domain measures retain their identities alongside the ROI measurement families in the optimization and adoption guide. This reference is derived from Corpus revision c0fbdf62aa93b8937a26acac1356de625a50c6eb (current/). Output-specific inputs and denominators determine computability; source definitions do not establish a tenant’s deployed scope.
Cost: Feature-level AI COGS
METRIC-COST-01. Sum eligible AI costs by feature and period; revenue share is a separately gated ratio
Boundary: Preserve feature identity and cost basis; do not call AI-only cost fully burdened gross margin.
ROI relationship: 01/06
Required dimensions: feature; period
METRIC-COST-01.
Eligible direct AI cost attributed to the feature and period.
Unit: USD/period. Required evidence: usage; prices; feature-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-01.
Eligible feature AI cost divided by compatible nonzero feature revenue; missing revenue does not hide cost.
Unit: ratio. Required evidence: usage; prices; feature-map; revenue.
Denominator: reference: eligible_feature_revenue; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Cost: Cost per inference turn
METRIC-COST-02. Per-turn cost distribution, median and P95, partitioned by provider/model/cache state
Boundary: Use non-overlapping provider usage buckets, effective-dated rates and scoped missing-cost counts.
ROI relationship: 17/23
Required dimensions: model_tier; provider; cache_state
METRIC-COST-02.
Eligible cost of each turn, partitioned by provider, model and cache state; this observed series supports the cost distribution.
Unit: USD/turn. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-COST-02.
Median of eligible per-turn costs from the same complete population; do not combine precomputed medians.
Unit: USD/turn. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-02.
95th percentile of eligible per-turn costs from the same population; preserve sample size and missing-cost scope.
Unit: USD/turn. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
Cost: Invoice reconciliation coverage
METRIC-COST-03. Attributed eligible cost / eligible invoice cost; separately report scope coverage, signed residual and absolute error
Boundary: The 98% figure is an unverified target. Credits, currency, tax, shared charges and timing require explicit treatment.
ROI relationship: 37
Required dimensions: invoice_cycle; provider
METRIC-COST-03.
Attributed eligible cost divided by compatible eligible invoice cost; never force residuals into attributed owners.
Unit: ratio. Required evidence: usage; prices; invoices.
Denominator: reference: eligible_invoice_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-03.
Matched eligible invoice line count divided by all eligible invoice lines in the registered reconciliation scope.
Unit: ratio. Required evidence: usage; prices; invoices.
Denominator: reference: eligible_invoice_line_count; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-03.
Eligible invoice cost minus matched attributed, allocated and explicitly reconciled cost on one non-overlapping basis.
Unit: USD. Required evidence: usage; prices; invoices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-03.
Sum of absolute matched line-level estimation errors; offsetting positive and negative errors do not cancel.
Unit: USD. Required evidence: usage; prices; invoices.
Denominator: Not applicable. Unavailable state: missing_required_input.
Cost: Shared-cost allocation conservation
METRIC-COST-04. Allocated shared cost plus unallocated residual equals the eligible shared-cost pool within declared rounding tolerance
Boundary: Conservation is not attribution accuracy or proof that a chargeback cannot be disputed.
ROI relationship: 37/39
Required dimensions: allocation_policy_revision
METRIC-COST-04.
Eligible shared costs allocated under the effective policy; preserve rounding and disputed allocations.
Unit: USD. Required evidence: invoices; org-map; allocation-policy.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-04.
Eligible shared-cost pool left unallocated under the same policy and period.
Unit: USD. Required evidence: invoices; org-map; allocation-policy.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-04.
Allocated shared cost plus unallocated residual minus the eligible shared-cost pool; zero is observed only when inputs reconcile within the declared tolerance.
Unit: USD. Required evidence: invoices; org-map; allocation-policy.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-04.
Allocated shared cost divided by the eligible shared-cost pool; a zero pool leaves USD observations available.
Unit: ratio. Required evidence: invoices; org-map; allocation-policy.
Denominator: reference: eligible_shared_cost_pool; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Cost: Prompt-caching economics
METRIC-COST-05. Cache-read/write usage shares; cached/uncached cost; comparable no-cache counterfactual delta. Also expose the source-required outputs: cache_hit_turn_share, cached_turn_cost, uncached_turn_cost, team_cache_cost_delta.
Boundary: Missing cache state is unknown; normalize inclusive token fields and price writes, reads and TTL effects separately.
ROI relationship: 15
Required dimensions: squad; provider; model_tier; cache_state
METRIC-COST-05.
Observed cache-read token units divided by compatible eligible token units; preserve missing cache state as unknown.
Unit: ratio. Required evidence: usage.
Denominator: reference: eligible_cache_usage_units; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-05.
Observed cache-write token units divided by compatible eligible token units; normalize provider fields so inclusive token counters are not double counted.
Unit: ratio. Required evidence: usage.
Denominator: reference: eligible_cache_usage_units; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-05.
Eligible cache-read usage at applicable effective read rates; missing write rates does not suppress a known read cost.
Unit: USD. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-05.
Eligible cache-write usage at applicable write rates and TTL terms; preserve unknown price components.
Unit: USD. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-05.
Eligible uncached usage at applicable effective rates, with inclusive usage normalized to non-overlapping buckets.
Unit: USD. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-05.
Comparable no-cache cost minus observed cached and uncached cost for the same workload and effective prices; require all included components and retain counterfactual assumptions.
Unit: USD. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-05.
Turns with a positive observed cache read divided by cache-eligible turns whose cache state is known. Token read share is separate; unknown cache states cannot be counted as misses.
Unit: ratio. Required evidence: usage.
Denominator: reference: eligible_cache_state_known_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-05.
Total eligible paid cost of turns with an observed cache read divided by those turns, including write, read, uncached and output components. Preserve model, provider, period and cache eligibility.
Unit: USD/turn. Required evidence: usage; prices.
Denominator: reference: eligible_cache_hit_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-05.
Total eligible paid cost of known cache-miss turns divided by those turns. Unknown cache states are excluded and disclosed; a missing write price cannot be treated as zero.
Unit: USD/turn. Required evidence: usage; prices.
Denominator: reference: eligible_known_cache_miss_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-05.
For each effective-dated squad, comparable no-cache cost minus observed total cost for the same eligible turns and rates. Preserve counterfactual assumptions and unknown ownership; this is not realized cash savings.
Unit: USD. Required evidence: usage; prices; org-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
Cost: Unattributed and failed-work spend
METRIC-COST-06. Independent unresolved-owner, failed-attempt, and failed-run spend outputs; disclose overlap rather than summing them as disjoint categories
Boundary: A provider error is an attempt failure, not necessarily a failed workflow. Team views require effective-dated mapping. Evaluate each named output independently. Ratios additionally require a named nonzero eligible cost denominator. Overlap between failed and unresolved spend is reported, never added twice.
ROI relationship: 14/39
Required dimensions: squad
METRIC-COST-06.
Eligible paid cost whose ownership remains unresolved; an incomplete mapping records unresolved ownership instead of inventing an owner.
Unit: USD. Required evidence: usage; prices; identity-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-06.
Eligible paid cost of failed attempts, including attempts later rescued by a successful retry.
Unit: USD. Required evidence: usage; prices; workflow-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-06.
Eligible paid cost of runs whose terminal lifecycle meets the registered failure rule. A failed attempt alone does not mark its run failed.
Unit: USD. Required evidence: usage; prices; workflow-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-06.
unresolved_owner_spend divided by total eligible cost for the same tenant, population, window, currency and cost basis. Preserve the USD observation when the denominator is zero, unknown or incompatible; overlap among categories remains explicit.
Unit: ratio. Required evidence: usage; prices; identity-map.
Denominator: reference: total_eligible_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-06.
failed_attempt_spend divided by total eligible cost for the same tenant, population, window, currency and cost basis. Preserve the USD observation when the denominator is zero, unknown or incompatible; overlap among categories remains explicit.
Unit: ratio. Required evidence: usage; prices; workflow-events.
Denominator: reference: total_eligible_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-06.
failed_run_spend divided by total eligible cost for the same tenant, population, window, currency and cost basis. Preserve the USD observation when the denominator is zero, unknown or incompatible; overlap among categories remains explicit.
Unit: ratio. Required evidence: usage; prices; workflow-events.
Denominator: reference: total_eligible_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Cost: Chargeback/showback coverage and disputes
METRIC-COST-07. Statements issued / eligible org units; customer-registered disputed statements / issued statements per cycle
Boundary: Dispute rate is unavailable without the customer’s intake/closure records; conserve the common invoice pool.
ROI relationship: 40
Required dimensions: org_unit; statement_cycle
METRIC-COST-07.
Issued statement records under the allocation policy divided by eligible organizational units for the same cycle.
Unit: ratio. Required evidence: invoices; org-map; allocation-policy.
Denominator: reference: eligible_organizational_units; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-07.
Customer-registered disputed statements divided by issued statements in the same cycle; missing dispute records is unavailable, not zero.
Unit: ratio. Required evidence: invoices; org-map; allocation-policy; dispute-records.
Denominator: reference: issued_statements; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Cost: Model price efficiency
METRIC-COST-08. Cost per standardized token unit and comparable routing cost delta for a fixed workload and acceptance rule. Also expose the source-required outputs: baseline_standardized_token_cost, standardized_token_cost_delta.
Boundary: A lower token price or different workload mix does not establish realized routing savings or equal quality.
ROI relationship: 13/23
Required dimensions: provider; model_tier; baseline_revision
METRIC-COST-08.
Eligible cost per 1000 standardized tokens for the specified provider, model and workload.
Unit: USD/1000 tokens. Required evidence: usage; prices.
Denominator: reference: eligible_standardized_tokens; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-08.
Matched routing cost difference for the same workload and evidenced acceptance rule; missing routing or outcome evidence preserves observed token cost.
Unit: USD. Required evidence: usage; prices; routing-events; outcomes.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-08.
Frozen comparable baseline cost per 1,000 compatible token units, retaining workload, token composition, model, provider, currency and effective price revision. A baseline can be visible before current-period usage arrives.
Unit: USD/1000 tokens. Required evidence: price-efficiency-baseline.
Denominator: reference: baseline_compatible_token_quantity; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-08.
Current eligible cost per 1,000 compatible token units minus the frozen matched baseline cost per 1,000 tokens. Preserve current and baseline values; incompatible token or cost bases withhold the comparison.
Unit: USD/1000 tokens. Required evidence: usage; prices; price-efficiency-baseline.
Denominator: reference: matched_compatible_token_quantity; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Cost: Budget variance and forecast error
METRIC-COST-09. Actual minus approved budget; relative variance only for nonzero budget; frozen-vintage forecast error on matched actuals
Boundary: Do not backfit forecast vintages. Declare sign, horizon, cohort and zero-baseline behavior.
ROI relationship: 34
Required dimensions: budget_period; forecast_vintage
METRIC-COST-09.
Actual eligible cost minus the approved budget on a matched basis; no forecast is required.
Unit: USD. Required evidence: usage; prices; budgets.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-09.
Budget variance divided by an eligible nonzero approved budget; a zero budget withholds this ratio while preserving absolute variance.
Unit: ratio. Required evidence: usage; prices; budgets.
Denominator: reference: approved_budget; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-09.
Actual eligible cost minus the frozen-vintage forecast for matched tenant, population, horizon, currency and cost basis. A zero forecast still permits this signed currency error; no budget is required.
Unit: USD. Required evidence: usage; prices; forecast-vintages.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-COST-09.
Signed frozen-forecast currency error divided by the absolute frozen forecast cost on the same matched basis. A zero, unknown or incompatible forecast denominator withholds this ratio while preserving currency error.
Unit: ratio. Required evidence: usage; prices; forecast-vintages.
Denominator: reference: absolute_frozen_forecast_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Cost: Feature gross margin
METRIC-COST-10. Feature revenue less explicitly scoped direct and allocated costs, divided by eligible nonzero feature revenue. Also expose the source-required outputs: fully_burdened_feature_margin.
Boundary: Label AI contribution margin unless all applicable COGS are supplied; no inferred revenue or burden.
ROI relationship: 06.
Required dimensions: feature; cost_basis
METRIC-COST-10.
Feature revenue less explicitly scoped direct and allocated costs, divided by eligible nonzero feature revenue. Label AI contribution margin unless all applicable COGS are supplied; no inferred revenue or burden.
Unit: ratio. Required evidence: usage; prices; feature-map; revenue; allocation-policy.
Denominator: reference: eligible_feature_revenue; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-COST-10.
Eligible feature revenue minus every applicable direct, shared and burdened cost component, divided by compatible feature revenue. Complete-cost-basis evidence establishes included, non-applicable and missing components; time-only labor requires authorized rate conversion. Missing burden withholds this variant while preserving contribution margin.
Unit: ratio. Required evidence: usage; prices; feature-map; revenue; allocation-policy; complete-cost-basis.
Denominator: reference: eligible_feature_revenue; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Interception coverage
METRIC-ADOPTION-01. Observed managed eligible invocations / independently defined eligible invocation universe, with scope and uncertainty
Boundary: Detected candidates are not an unbiased enterprise denominator. If the denominator is unknown, report observed counts only.
ROI relationship: 39 (coverage estimate)
Required dimensions: organization; coverage_population
METRIC-ADOPTION-01.
Count of observed managed eligible invocations in the declared population, independent of enterprise denominator evidence.
Unit: count. Required evidence: usage.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-01.
Observed managed eligible invocations divided by an independently validated eligible invocation universe; discovered candidates alone do not establish that universe.
Unit: ratio. Required evidence: usage; discovery-evidence; egress-validation.
Denominator: reference: independently_validated_eligible_invocations; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Agentic versus interactive mix
METRIC-ADOPTION-02. Turns in explicitly evidenced execution modes / mode-classified eligible turns; show unknown/mixed shares separately
Boundary: Service accounts are not automatically agents and human principals are not automatically interactive traffic.
ROI relationship: 09/16/26 (execution-mode diagnostic)
Required dimensions: squad; execution_mode
METRIC-ADOPTION-02.
Evidenced agentic turns divided by mode-classified eligible turns; a service account alone is not agent evidence.
Unit: ratio. Required evidence: usage; identity-map; workflow-events.
Denominator: reference: mode_classified_eligible_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-ADOPTION-02.
Evidenced interactive turns divided by mode-classified eligible turns; a human principal alone is insufficient.
Unit: ratio. Required evidence: usage; identity-map; workflow-events.
Denominator: reference: mode_classified_eligible_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-ADOPTION-02.
Explicitly mixed-mode turns divided by the same mode-classified eligible turns.
Unit: ratio. Required evidence: usage; identity-map; workflow-events.
Denominator: reference: mode_classified_eligible_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-ADOPTION-02.
Unclassified turns divided by all eligible observed turns; name this broader denominator so it is not added to classified-mode shares.
Unit: ratio. Required evidence: usage; identity-map; workflow-events.
Denominator: reference: all_eligible_observed_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Feature and workflow breadth
METRIC-ADOPTION-03. Distinct attributed active feature/workflow IDs; instrumented share only with a complete eligible registry
Boundary: Use stable, effective-dated definitions; do not infer an unobserved feature population from telemetry.
ROI relationship: 30
Required dimensions: feature; workflow; period
METRIC-ADOPTION-03.
Distinct attributed active feature identities in the declared window; missing workflow registration does not hide this count.
Unit: count. Required evidence: usage; feature-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-03.
Distinct attributed active registered workflow identities in the declared window; missing feature mapping does not hide this count.
Unit: count. Required evidence: usage; workflow-registry.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-03.
Instrumented active feature share only when the feature map supplies a complete eligible registry; partial registry withholds the ratio while preserving the observed count.
Unit: ratio. Required evidence: usage; feature-map.
Denominator: reference: complete_eligible_feature_registry; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-ADOPTION-03.
Instrumented active workflow share only over a complete eligible registry; partial registry withholds the ratio while preserving the observed count.
Unit: ratio. Required evidence: usage; workflow-registry.
Denominator: reference: complete_eligible_workflow_registry; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Usage depth by squad
METRIC-ADOPTION-04. Eligible turns per active pseudonymous principal and time window, aggregated at approved cohort sizes. Also expose the source-required outputs: power_user_share.
Boundary: Do not introduce individual productivity surveillance; disclose unmapped identities and privacy suppression.
ROI relationship: 26 (raw depth diagnostic)
Required dimensions: squad; week; usage_threshold_revision
METRIC-ADOPTION-04.
Eligible turns per active pseudonymous principal and time window, aggregated at approved cohort sizes. Do not introduce individual productivity surveillance; disclose unmapped identities and privacy suppression.
Unit: turns/principal/week. Required evidence: usage; identity-map; org-map.
Denominator: reference: active_authorized_principals; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-ADOPTION-04.
Active authorized human principals meeting the versioned customer power-user usage threshold divided by active eligible human principals in the same squad and week. Exclude service accounts from human counts, disclose unmapped principals and apply aggregate privacy suppression.
Unit: ratio. Required evidence: usage; identity-map; org-map; usage-thresholds.
Denominator: reference: active_eligible_human_principals; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Active users and squads
METRIC-ADOPTION-05. Distinct active identities/squads; active share only over a dated, authorized eligible roster
Boundary: Service accounts, people and squads have different denominators; do not merge them into a people count.
ROI relationship: 26
Required dimensions: principal_kind; squad; activity_window
METRIC-ADOPTION-05.
Distinct active authorized pseudonymous principals within the permitted aggregation boundary.
Unit: count. Required evidence: usage; identity-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-05.
Distinct active squads with effective-dated membership and permitted cohort sizes.
Unit: count. Required evidence: usage; identity-map; org-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-05.
Active principals divided by the dated authorized eligible principal roster; an absent roster does not erase the observed count.
Unit: ratio. Required evidence: usage; identity-map; eligible-roster.
Denominator: reference: authorized_eligible_principal_roster; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-ADOPTION-05.
Active squads divided by the dated authorized eligible squad roster; an absent roster does not erase the observed count.
Unit: ratio. Required evidence: usage; identity-map; org-map; eligible-roster.
Denominator: reference: authorized_eligible_squad_roster; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Squad-cohort usage retention
METRIC-ADOPTION-06. Eligible initial cohort squads meeting a fixed activity rule at a mature 90-day horizon / original eligible cohort
Boundary: Immature windows are unavailable, not zero; record reorganization and deprovisioning rules. Renewal prediction is unvalidated.
ROI relationship: 28 (usage retention remains separate)
Required dimensions: squad_cohort; 90_day_horizon
METRIC-ADOPTION-06.
Eligible initial cohort squads meeting a fixed activity rule at a mature 90-day horizon / original eligible cohort. Immature windows are unavailable, not zero; record reorganization and deprovisioning rules. Renewal prediction is unvalidated.
Unit: ratio. Required evidence: usage; org-map; cohort-history.
Denominator: reference: original_eligible_squad_cohort; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Time to first value
METRIC-ADOPTION-07. Elapsed time from agreed deployment milestone to first usable, accepted dashboard/report milestone. Also expose the source-required outputs: deployment_to_first_dashboard, deployment_to_first_chargeback_report, dashboard_to_first_chargeback_report.
Boundary: An empty dashboard is not first value; use customer-agreed milestone definitions and include censored installations.
ROI relationship: 33
Required dimensions: deployment; milestone_kind
METRIC-ADOPTION-07.
Elapsed time from agreed deployment milestone to first usable, accepted dashboard/report milestone. An empty dashboard is not first value; use customer-agreed milestone definitions and include censored installations.
Unit: duration. Required evidence: deployment-milestones; product-events; customer-acceptance.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-07.
Elapsed days from agreed deployment to first usable dashboard using versioned milestone definitions and true event boundaries. Missing endpoints remain censored; negative or incompatible clocks are unavailable. This observed interval is separate from customer acceptance of value.
Unit: duration. Required evidence: deployment-milestones; product-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-07.
Elapsed days from agreed deployment to first usable chargeback report using versioned milestone definitions and true event boundaries. Missing endpoints remain censored; negative or incompatible clocks are unavailable. This observed interval is separate from customer acceptance of value.
Unit: duration. Required evidence: deployment-milestones; product-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-07.
Elapsed days from first usable dashboard to first usable chargeback report using versioned milestone definitions and true event boundaries. Missing endpoints remain censored; negative or incompatible clocks are unavailable. This observed interval is separate from customer acceptance of value.
Unit: duration. Required evidence: deployment-milestones; product-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
Adoption: Dashboard engagement
METRIC-ADOPTION-08. Authorized weekly viewers and query events by permitted aggregate persona/cohort. Also expose the source-required outputs: queries_per_finops_user.
Boundary: Internal product-health signal, not proof of customer ROI. Exclude automation and respect privacy thresholds.
ROI relationship: internal Venturi product health, no direct customer inference ROI
Required dimensions: finops_role_cohort; week
METRIC-ADOPTION-08.
Distinct authorized weekly viewers at permitted aggregate cohort sizes; exclude automation.
Unit: count. Required evidence: product-events; identity-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-08.
Eligible query events by permitted aggregate persona and cohort; not a measure of customer ROI.
Unit: count. Required evidence: product-events; identity-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-ADOPTION-08.
Eligible dashboard queries by authorized FinOps human users divided by active authorized FinOps users in the same declared period. Version the role mapping, exclude automation and apply aggregate cohort suppression.
Unit: queries/principal. Required evidence: product-events; identity-map; eligible-roster.
Denominator: reference: active_authorized_finops_users; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Adoption: Domain activation
METRIC-ADOPTION-09. Count of domains meeting a documented non-demo activity rule per tenant/window
Boundary: An exposed navigation tab is not activation, entitlement, revenue, or verified expansion.
ROI relationship: 30 plus internal domain-activation diagnostic
Required dimensions: customer; domain; period
METRIC-ADOPTION-09.
Count of domains meeting a documented non-demo activity rule per tenant/window. An exposed navigation tab is not activation, entitlement, revenue, or verified expansion.
Unit: count. Required evidence: product-events; domain-registry.
Denominator: Not applicable. Unavailable state: missing_required_input.
Adoption: Model tier adoption mix
METRIC-ADOPTION-10. Eligible turns by provider/model tier divided by eligible classified turns; compare matched windows
Boundary: Tier definitions are versioned; cheap-model adoption is not automatically economic efficiency or accepted quality.
ROI relationship: 13/23.
Required dimensions: provider; model_tier; period
METRIC-ADOPTION-10.
Eligible turns by provider/model tier divided by eligible classified turns; compare matched windows. Tier definitions are versioned; cheap-model adoption is not automatically economic efficiency or accepted quality.
Unit: ratio. Required evidence: usage; model-catalog.
Denominator: reference: eligible_classified_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Attribution resolution and independently evaluated precision
METRIC-QUALITY-01. Report exact-join resolution coverage separately from correct reviewed assignments / independently labeled accepted assignments. Also expose the source-required outputs: exact_join_resolution_share, eligible_inference_resolution_share.
Boundary: Exact matching and high model confidence do not prove correctness. Precision needs non-circular labels, sampling scope and uncertainty.
ROI relationship: 39
Required dimensions: resolution_class; inference_eligibility_revision
METRIC-QUALITY-01.
Eligible observed events whose ownership and feature joins resolve, retaining exact, inferred and allocated classes.
Unit: ratio. Required evidence: usage; identity-map; feature-map.
Denominator: reference: eligible_observed_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-01.
Correct independently reviewed accepted assignments divided by independently labeled accepted assignments; record sampling and uncertainty.
Unit: ratio. Required evidence: usage; identity-map; feature-map; independent-labels.
Denominator: reference: independently_labeled_accepted_assignments; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-01.
Eligible observed events resolved by the declared exact-join rule divided by all eligible observed events. Preserve unresolved and inferred classes; exact matching does not prove correctness.
Unit: ratio. Required evidence: usage; identity-map; feature-map.
Denominator: reference: eligible_observed_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-01.
Eligible unresolved-after-exact-join events with an accepted inference assignment divided by the versioned inference-eligible event population. Preserve abstentions and the existing confidence cap; no assignment outside that eligible population may enter the numerator.
Unit: ratio. Required evidence: usage; identity-map; feature-map; inference-eligibility.
Denominator: reference: inference_eligible_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Estimated cost error against invoices
METRIC-QUALITY-02. Matched estimated cost minus invoice actuals; absolute relative error with explicit nonzero denominator and signed residual
Boundary: Separate reconciliation completeness from predictive error; align provider, account, currency, scope and period.
ROI relationship: 37
Required dimensions: provider; invoice_period
METRIC-QUALITY-02.
Matched estimated cost minus compatible invoice actuals, preserving provider, account, currency, population and period.
Unit: USD. Required evidence: usage; prices; invoices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-QUALITY-02.
Sum of absolute matched invoice-line estimation errors; disclose unmatched scope separately.
Unit: USD. Required evidence: usage; prices; invoices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-QUALITY-02.
Absolute matched invoice-line estimation error divided by compatible nonzero invoice actual cost.
Unit: ratio. Required evidence: usage; prices; invoices.
Denominator: reference: matched_invoice_actual_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Six-layer completeness
METRIC-QUALITY-03. Eligible events with all six required layers resolved / eligible events; separate exact, inferred and allocated resolution classes
Boundary: Missing inputs must stay unknown. Completeness is neither correctness nor total enterprise capture.
ROI relationship: 39
Required dimensions: resolution_class
METRIC-QUALITY-03.
Eligible events with all six layers resolved divided by all eligible events; preserve unresolved evidence.
Unit: ratio. Required evidence: usage; identity-map; org-map; feature-map; prices.
Denominator: reference: eligible_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-03.
Eligible fully resolved events in the declared exact resolution class divided by eligible events. Disclose overlap and classification precedence; completeness does not prove correctness.
Unit: ratio. Required evidence: usage; identity-map; org-map; feature-map; prices.
Denominator: reference: eligible_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-03.
Eligible fully resolved events in the declared inferred resolution class divided by eligible events. Disclose overlap and classification precedence; completeness does not prove correctness.
Unit: ratio. Required evidence: usage; identity-map; org-map; feature-map; prices.
Denominator: reference: eligible_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-03.
Eligible fully resolved events in the declared allocated resolution class divided by eligible events. Disclose overlap and classification precedence; completeness does not prove correctness.
Unit: ratio. Required evidence: usage; identity-map; org-map; feature-map; prices.
Denominator: reference: eligible_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Data freshness and query latency
METRIC-QUALITY-04. Event/ingest-to-served lag and query latency as separate distributions with clock and workload context
Boundary: P95 under one second is a target until measured under a specified load; do not average percentile aggregates.
ROI relationship: freshness/serving trust diagnostic supporting 36/38, not customer productivity by itself
Required dimensions: workload; statistic
METRIC-QUALITY-04.
Served time minus source event time for matched records, retaining clock uncertainty and late arrivals.
Unit: duration. Required evidence: pipeline-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-QUALITY-04.
Served time minus ingestion time for matched records; do not substitute for event freshness.
Unit: duration. Required evidence: pipeline-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-QUALITY-04.
Query completion minus query start under the declared workload. Preserve distribution and percentile labels without averaging percentile aggregates.
Unit: duration. Required evidence: serving-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
Quality: Failed-run cost by team
METRIC-QUALITY-05. Eligible cost of failed attempts and failed workflows, reported separately by effective owner. Also expose the source-required outputs: failed_run_share.
Boundary: Retry-rescued attempts remain paid costs without falsely classifying the whole run as failed.
ROI relationship: 14
Required dimensions: squad; run_maturity_rule
METRIC-QUALITY-05.
Eligible paid cost of failed attempts by effective owner, including retry-rescued attempts.
Unit: USD. Required evidence: usage; prices; identity-map; org-map; workflow-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-QUALITY-05.
Eligible paid cost of terminally failed runs by effective owner; an attempt failure alone does not make a failed run.
Unit: USD. Required evidence: usage; prices; identity-map; org-map; workflow-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-QUALITY-05.
Failed-attempt spend divided by total eligible team cost on the same period and basis; report overlap with failed runs.
Unit: ratio. Required evidence: usage; prices; identity-map; org-map; workflow-events.
Denominator: reference: eligible_team_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-05.
Failed-run spend divided by total eligible team cost on the same period and basis; never add overlapping shares.
Unit: ratio. Required evidence: usage; prices; identity-map; org-map; workflow-events.
Denominator: reference: eligible_team_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-05.
Distinct eligible terminally failed runs divided by eligible terminal runs under the same maturity rule, separately by authorized squad. A failed attempt rescued by retry is not a failed run; unknown and open runs remain disclosed.
Unit: ratio. Required evidence: workflow-registry; workflow-events.
Denominator: reference: eligible_terminal_runs; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Accepted-result rate
METRIC-QUALITY-06. Distinct eligible runs with an unrevoked customer-registered accepted result / eligible runs under a fixed maturity rule
Boundary: Core evidence capture is a prerequisite; expanded analytics may be Phase II. Venturi does not judge prompt/output content.
ROI relationship: 01/09/38
Required dimensions: workflow; acceptance_rule_revision
Canonical measure: METRIC-WORKFLOW-06
METRIC-QUALITY-06.
Distinct eligible completed/mature runs with unrevoked customer acceptance / eligible runs under the declared cohort rule. Basic acceptance evidence belongs to Core; customer semantics/version and absence/rejection distinction are mandatory.
Unit: ratio. Required evidence: workflow-registry; workflow-events; outcomes.
Denominator: reference: eligible_mature_runs; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Evidence coverage
METRIC-QUALITY-07. Registered eligible workflows/runs with valid evidence in the window / registered eligible workflows/runs
Boundary: Publish workflow coverage and run coverage as different outputs; stale or invalid evidence does not count.
ROI relationship: 38
Required dimensions: workflow; run
Canonical measure: METRIC-WORKFLOW-07
METRIC-QUALITY-07.
Distinct eligible registered workflows with valid, current evidence divided by eligible registered workflows. Evidence need not indicate success.
Unit: ratio. Required evidence: workflow-registry; workflow-events; outcomes.
Denominator: reference: eligible_registered_workflows; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-07.
Distinct eligible runs with valid, current evidence divided by eligible runs under the matched cohort rule.
Unit: ratio. Required evidence: workflow-registry; workflow-events; outcomes.
Denominator: reference: eligible_runs; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Step latency versus budget
METRIC-QUALITY-08. Step duration distribution and breach share against versioned customer budgets on matched step/run boundaries
Boundary: Provider call latency alone excludes app queueing and human waits; missing budgets yield no compliance score.
ROI relationship: 24
Required dimensions: step; budget_revision; statistic
METRIC-QUALITY-08.
Duration from the registered true step start to end; unfinished steps remain censored. A missing customer budget does not erase observed duration.
Unit: duration. Required evidence: workflow-registry; workflow-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-QUALITY-08.
Eligible steps exceeding their effective customer duration budget divided by steps with compatible versioned budgets and complete boundaries.
Unit: ratio. Required evidence: workflow-registry; workflow-events; step-budgets.
Denominator: reference: eligible_budgeted_steps; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Fallback rescue effectiveness
METRIC-QUALITY-09. Eligible failed routed attempts whose successor chain meets the defined success or accepted-result criterion / eligible routed failures
Boundary: Limit to observed routed traffic. Provider transport success and accepted business outcomes are separate endpoints.
ROI relationship: 25
Required dimensions: fallback_chain; success_criterion
METRIC-QUALITY-09.
Eligible routed failures whose linked successor chain meets the declared lifecycle success criterion, divided by eligible routed failures.
Unit: ratio. Required evidence: usage; routing-events; workflow-events.
Denominator: reference: eligible_routed_failures; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-09.
Eligible routed failures whose linked successor chain has mature unrevoked customer acceptance, divided by eligible routed failures.
Unit: ratio. Required evidence: usage; routing-events; workflow-events; outcomes.
Denominator: reference: eligible_routed_failures; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Quality: Audit-trail integrity
METRIC-QUALITY-10. Verified audit objects passing a declared validation suite / eligible checked objects; discrepancy counts and untested population
Boundary: A hash existing in a record is not evidence of append-only durability, authorization or complete lineage.
ROI relationship: 37/43.
Required dimensions: validation_suite; discrepancy_state
METRIC-QUALITY-10.
Eligible checked audit objects passing the declared validation suite divided by eligible checked objects.
Unit: ratio. Required evidence: audit-records; integrity-tests.
Denominator: reference: eligible_checked_audit_objects; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-QUALITY-10.
Distinct discrepancies in the declared checked audit population, keyed to suite and object revision.
Unit: count. Required evidence: audit-records; integrity-tests.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-QUALITY-10.
Enumerated eligible audit objects explicitly not tested in the authoritative test inventory. Missing test inventory is unavailable, not all-clear.
Unit: count. Required evidence: audit-records; integrity-tests.
Denominator: Not applicable. Unavailable state: missing_required_input.
Security & Compliance: Shadow-AI discovery with economic context
METRIC-SECURITY-01. Observed candidate unmanaged channels and attributed cost where matching usage/rates exist; independent detection evaluation. Also expose the source-required outputs: observed_uncontrolled_endpoint_count, unmanaged_traffic_share.
Boundary: Network metadata alone does not establish tokens or exact spend. Detection model evidence and denominator validity are separate gates. Independent detection evaluation retains its own evidence gate and is not an output inferred from either count or cost.
ROI relationship: 39/41/42
Required dimensions: detection_scope; evaluation_revision; coverage_population
METRIC-SECURITY-01.
Distinct observed candidate unmanaged channels within the declared detection scope; this count does not assert detector accuracy or enterprise-wide coverage.
Unit: count. Required evidence: discovery-evidence.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-01.
Matched eligible paid cost for detected channels with compatible usage and rates; missing prices withhold cost while preserving the observed count.
Unit: USD. Required evidence: discovery-evidence; usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-01.
Distinct observed endpoints independently validated as outside the declared managed AI path, within the authorized discovery scope. Keep unvalidated candidates separate and do not imply enterprise-wide discovery completeness.
Unit: count. Required evidence: discovery-evidence; egress-validation.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-01.
Independently validated unmanaged AI invocations divided by the independently validated eligible AI invocation universe in the same tenant, population and period. Detection-derived classifications require dated two-class evaluation evidence under the governed model contract. Unknown traffic and denominator gaps cannot become zero or an assumed managed-coverage complement.
Unit: ratio. Required evidence: usage; discovery-evidence; egress-validation; detection-evaluation.
Denominator: reference: independently_validated_eligible_invocations; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Security & Compliance: Shadow-AI detection delay
METRIC-SECURITY-02. Detection time minus independently observed first eligible unmanaged event, with left/right-censoring
Boundary: Do not timestamp unseen activity by the first later proxy observation; report earliest-observed delay, not unknowable true onset.
ROI relationship: 41 (detection-time diagnostic)
Required dimensions: endpoint; statistic; censoring_state
METRIC-SECURITY-02.
Detection time minus independently observed first eligible unmanaged event, with left/right-censoring. Do not timestamp unseen activity by the first later proxy observation; report earliest-observed delay, not unknowable true onset.
Unit: duration. Required evidence: discovery-evidence; egress-validation.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
Security & Compliance: Processing-boundary assurance
METRIC-SECURITY-03. Checked processing/storage paths conforming to the declared deployment boundary; runtime violations and unchecked scope. Also expose the source-required outputs: in_boundary_processing_share.
Boundary: Configuration is necessary but insufficient. External provider processing must not be described as inside a customer VPC.
ROI relationship: 43
Required dimensions: declared_boundary; processing_location
METRIC-SECURITY-03.
Checked eligible processing and storage paths conforming to the declared deployment boundary divided by eligible checked paths.
Unit: ratio. Required evidence: deployment-config; runtime-egress-evidence.
Denominator: reference: eligible_checked_paths; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SECURITY-03.
Observed runtime boundary violations in the declared checked scope. External provider processing remains external.
Unit: count. Required evidence: deployment-config; runtime-egress-evidence.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-03.
Count eligible paths marked unassessed in an inventory whose completeness has been established; missing runtime evidence does not establish compliance.
Unit: count. Required evidence: deployment-config; runtime-egress-evidence.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-03.
Eligible processing events positively evidenced inside the declared customer boundary divided by all eligible processing events in the matched population. Reconcile event location and coverage; path conformance alone cannot produce this share. External provider processing remains external.
Unit: ratio. Required evidence: usage; deployment-config; runtime-egress-evidence; processing-event-evidence.
Denominator: reference: eligible_processing_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Security & Compliance: Content-retention assurance
METRIC-SECURITY-04. Checked data stores/logs/traces/dumps meeting the metadata-only retention contract; findings and unchecked scope. Also expose the source-required outputs: content_free_event_share.
Boundary: Content-free collection and zero retention are different claims. Synthetic-canary tests do not prove every production event was clean.
ROI relationship: 43
Required dimensions: retention_window; evidence_coverage
METRIC-SECURITY-04.
Eligible checked stores, logs, traces and dumps meeting the declared metadata-only contract divided by eligible checked surfaces.
Unit: ratio. Required evidence: deployment-config; retention-evidence; integrity-tests.
Denominator: reference: eligible_checked_retention_surfaces; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SECURITY-04.
Observed retention-contract findings keyed to surface and evidence revision; canary results retain their bounded scope.
Unit: count. Required evidence: deployment-config; retention-evidence; integrity-tests.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-04.
Enumerated eligible surfaces explicitly untested in the coverage inventory; an unknown inventory cannot produce zero unchecked scope.
Unit: count. Required evidence: deployment-config; retention-evidence; integrity-tests.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-04.
Eligible events positively covered by metadata-only attestations that no raw prompt, output content or source-specified raw IP is retained across every applicable surface and retention window, divided by eligible events. Store checks or synthetic canaries alone cannot produce this event percentage. Do not collect content or raw IP to compute it.
Unit: ratio. Required evidence: usage; deployment-config; retention-evidence; integrity-tests; event-retention-attestations.
Denominator: reference: eligible_retention_scope_events; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Security & Compliance: Principal hygiene
METRIC-SECURITY-05. Observed actions from unresolved, expired or unregistered pseudonymous principals against an effective-dated authorized roster
Boundary: Unknown identity is a data-quality finding, not proof of malicious or unauthorized behavior.
ROI relationship: 39/41
Required dimensions: principal_status
METRIC-SECURITY-05.
Observed actions in the unresolved principal class under effective-dated mapping and roster rules; this classification does not prove malicious behavior.
Unit: count. Required evidence: usage; identity-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-05.
Observed actions in the unresolved principal class divided by all eligible observed actions in the same authorized scope; disclose category overlap.
Unit: ratio. Required evidence: usage; identity-map.
Denominator: reference: eligible_observed_actions; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SECURITY-05.
Observed actions in the expired principal class under effective-dated mapping and roster rules; this classification does not prove malicious behavior.
Unit: count. Required evidence: usage; identity-map; eligible-roster.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-05.
Observed actions in the expired principal class divided by all eligible observed actions in the same authorized scope; disclose category overlap.
Unit: ratio. Required evidence: usage; identity-map; eligible-roster.
Denominator: reference: eligible_observed_actions; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SECURITY-05.
Observed actions in the unregistered principal class under effective-dated mapping and roster rules; this classification does not prove malicious behavior.
Unit: count. Required evidence: usage; identity-map; eligible-roster.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-05.
Observed actions in the unregistered principal class divided by all eligible observed actions in the same authorized scope; disclose category overlap.
Unit: ratio. Required evidence: usage; identity-map; eligible-roster.
Denominator: reference: eligible_observed_actions; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Security & Compliance: Advisory policy alerts and availability
METRIC-SECURITY-06. Policy evaluations, advisory breaches, evaluations unavailable/stale and gap duration on the eligible observed path. Also expose the source-required outputs: decision_time_alert_count.
Boundary: Preserve always-forward advisory Gate and no synchronous RAIL. Missing evaluation must never appear as zero violations/all-clear.
ROI relationship: 41/43
Required dimensions: month; policy_revision
METRIC-SECURITY-06.
Observed eligible advisory policy evaluations in the declared window.
Unit: count. Required evidence: policy-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-06.
Observed advisory breaches under the effective policy; this does not establish policy availability during an observation gap.
Unit: count. Required evidence: policy-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-06.
Available and current evaluations over the declared eligible observed path, with policy and serving coverage reconciled.
Unit: ratio. Required evidence: policy-events; serving-observability.
Denominator: reference: eligible_policy_evaluations; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SECURITY-06.
Duration of unavailable or stale evaluation intervals against the declared policy and serving boundaries.
Unit: duration. Required evidence: policy-events; serving-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-06.
Distinct generated decision-time governance alerts in the declared calendar month, keyed by alert ID and policy revision. Retain generated, sent and acknowledged states separately; a breach event alone does not prove that an alert was generated or delivered.
Unit: count. Required evidence: alert-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
Security & Compliance: Decision-lineage completeness
METRIC-SECURITY-07. Eligible scoped cost linked to work, valid evidence and an actual versioned decision / eligible scoped cost
Boundary: Work-to-cost lineage can exist earlier; full work-evidence-decision lineage requires those customer events and scope.
ROI relationship: 39/43
Required dimensions: cost_basis; lineage_revision
Canonical measure: METRIC-WORKFLOW-05
METRIC-SECURITY-07.
Eligible scoped cost traceable to registered work and actual lifecycle evidence divided by eligible scoped cost; acceptance and decisions are not required for this partial lineage.
Unit: ratio. Required evidence: usage; prices; workflow-registry; workflow-events.
Denominator: reference: eligible_scoped_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SECURITY-07.
Eligible cost linked through work, valid evidence and an actual versioned decision divided by eligible scoped cost.
Unit: ratio. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes; decision-events.
Denominator: reference: eligible_scoped_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Security & Compliance: Audit-package readiness
METRIC-SECURITY-08. Time to generate a scoped package; validated packages and delivery acknowledgments. Also expose the source-required outputs: generated_package_count.
Boundary: Generation is not auditor acceptance, certification or regulatory compliance; qualify export and customer-retention scope.
ROI relationship: 43
Required dimensions: package_revision; generation_mode
METRIC-SECURITY-08.
Elapsed time from the recorded package request to generation completion; successful generation is not integrity acceptance.
Unit: duration. Required evidence: audit-records.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-08.
Scoped generated packages whose declared integrity checks passed; untested packages remain separate.
Unit: count. Required evidence: audit-records; integrity-tests.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-08.
Recorded acknowledgments from the intended destination for a scoped package; generation alone is not delivery.
Unit: count. Required evidence: audit-records; grc-destination.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-08.
Distinct automatically generated audit packages completed in the declared period, keyed by package and revision. Generation, integrity validation and destination acknowledgment remain separate states.
Unit: count. Required evidence: audit-records.
Denominator: Not applicable. Unavailable state: missing_required_input.
Security & Compliance: Interception coverage for security
METRIC-SECURITY-09. Same canonical managed-coverage primitive as Adoption, rendered through a security lens
Boundary: Reuse Adoption’s denominator, revision and uncertainty; do not compute a second contradictory percentage.
ROI relationship: 39
Required dimensions: organization; coverage_population
Canonical measure: METRIC-ADOPTION-01
METRIC-SECURITY-09.
Count of observed managed eligible invocations in the declared population, independent of enterprise denominator evidence.
Unit: count. Required evidence: usage.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-09.
Observed managed eligible invocations divided by an independently validated eligible invocation universe; discovered candidates alone do not establish that universe.
Unit: ratio. Required evidence: usage; discovery-evidence; egress-validation.
Denominator: reference: independently_validated_eligible_invocations; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Security & Compliance: AI-channel incident exposure and MTTR
METRIC-SECURITY-10. Registered relevant incidents and time-to-resolution from the customer’s incident process, joined to attribution evidence
Boundary: Venturi supplies evidence; customer process outcomes are not generated by telemetry alone.
ROI relationship: 41.
Required dimensions: incident_kind; lifecycle_state
METRIC-SECURITY-10.
Distinct registered relevant incidents joined to attribution evidence under the customer’s incident definitions.
Unit: count. Required evidence: incident-records; identity-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SECURITY-10.
Mean resolution time for eligible resolved incidents from true incident lifecycle boundaries. Report open and censored populations separately; no resolved incidents yields unavailable.
Unit: duration. Required evidence: incident-records; identity-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
Capacity: AI rate-limit headroom
METRIC-CAPACITY-01. Remaining capacity against fresh provider/account/model/region/window-specific TPM/RPM quotas; separately classified 429 events
Boundary: Observed usage and 429s do not reveal the true quota by themselves; shared quota allocation to squads must be explicit.
ROI relationship: 36
Required dimensions: squad; week; quota_kind; provider_account; model; region; quota_window
METRIC-CAPACITY-01.
Observed 429 events classified by available provider and error evidence, with unknown classes retained; no configured quota is required.
Unit: count. Required evidence: usage.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-CAPACITY-01.
Remaining capacity against fresh matching provider, account, model, region, quota kind and time-window limits; stale or absent limits withhold headroom.
Unit: ratio. Required evidence: usage; quota-config.
Denominator: reference: fresh_compatible_quota; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Capacity: Token throughput by squad and feature
METRIC-CAPACITY-02. Non-overlapping observed token usage per time unit and measured peaks over a fixed aggregation window
Boundary: Peak-versus-provisioned needs the provisioned quota. Preserve input/output/cache semantics and missing usage.
ROI relationship: 35/36
Required dimensions: squad; feature; day; provider_account; quota_window
METRIC-CAPACITY-02.
Non-overlapping observed tokens per time unit by authorized squad and feature.
Unit: tokens/time. Required evidence: usage; identity-map; org-map; feature-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-CAPACITY-02.
Maximum observed throughput across the fixed aggregation windows, preserving input, output and cache semantics.
Unit: tokens/time. Required evidence: usage; identity-map; org-map; feature-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-CAPACITY-02.
Observed peak throughput divided by fresh compatible provisioned quota for the same provider, account, region and window.
Unit: ratio. Required evidence: usage; identity-map; org-map; feature-map; quota-config.
Denominator: reference: compatible_provisioned_quota; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Capacity: Provider concentration
METRIC-CAPACITY-03. Largest provider share and HHI=sum of squared provider shares, separately for eligible cost and turns
Boundary: HHI is not the largest share. Normalize shares, state scale, and consider common upstream failure domains.
ROI relationship: 25
Required dimensions: provider; weighting_basis
METRIC-CAPACITY-03.
Largest provider share of eligible observed turns.
Unit: ratio. Required evidence: usage.
Denominator: reference: eligible_observed_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-CAPACITY-03.
Sum of squared provider shares of eligible observed turns on the same population.
Unit: ratio. Required evidence: usage.
Denominator: reference: eligible_observed_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-CAPACITY-03.
Largest provider share of eligible cost on the declared basis.
Unit: ratio. Required evidence: usage; prices.
Denominator: reference: eligible_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-CAPACITY-03.
Sum of squared provider shares of eligible cost; missing rates do not withhold turn concentration.
Unit: ratio. Required evidence: usage; prices.
Denominator: reference: eligible_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Capacity: Customer data-plane utilization
METRIC-CAPACITY-04. Authorized node CPU/memory distributions and scaling events for an enumerated deployment scope
Boundary: Customer-VPC deployment alone does not establish telemetry permissions or collection coverage.
ROI relationship: 35 for the measured Venturi-overhead boundary only, not provider GPU utilization
Required dimensions: node; deployment
METRIC-CAPACITY-04.
Authorized node CPU utilization series and declared statistics for the enumerated deployment scope; missing telemetry is unknown.
Unit: ratio. Required evidence: node-observability.
Denominator: reference: authorized_node_cpu_capacity; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-CAPACITY-04.
Authorized node memory utilization series and declared statistics for the same deployment scope.
Unit: ratio. Required evidence: node-observability.
Denominator: reference: authorized_node_memory_capacity; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-CAPACITY-04.
Observed scaling events for the enumerated nodes; deployment location alone does not establish telemetry coverage.
Unit: count. Required evidence: node-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
Capacity: Fast-path responsiveness
METRIC-CAPACITY-05. Serving-query latency and cache-hit ratio under a specified workload, plus errors and stale responses
Boundary: Internal serving health is not end-to-end workflow performance; cache misses and stale hits need distinct treatment.
ROI relationship: internal serving diagnostic
Required dimensions: workload; statistic
METRIC-CAPACITY-05.
Serving-query duration observations and declared statistics under a specified workload, distinct from end-to-end workflow latency.
Unit: duration. Required evidence: serving-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-CAPACITY-05.
Eligible cache hits divided by eligible cache lookups; distinguish stale hits and misses.
Unit: ratio. Required evidence: serving-observability.
Denominator: reference: eligible_cache_lookups; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-CAPACITY-05.
Observed serving errors under the declared workload and window.
Unit: count. Required evidence: serving-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-CAPACITY-05.
Responses failing the declared freshness rule, even if returned from cache; unknown freshness is separately unavailable.
Unit: count. Required evidence: serving-observability.
Denominator: Not applicable. Unavailable state: missing_required_input.
Capacity: Queue and wait time
METRIC-CAPACITY-06. Wait duration on explicitly instrumented router/application queues, scoped by run/attempt and traffic mode
Boundary: Passthrough has no Venturi queue. Never fabricate zero queue time for an uninstrumented application queue.
ROI relationship: 24/36
Required dimensions: queue_kind; statistic
METRIC-CAPACITY-06.
Wait duration from true instrumented router queue boundaries, scoped by run, attempt and traffic mode.
Unit: duration. Required evidence: routing-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-CAPACITY-06.
Wait duration from true instrumented application queue boundaries; this output does not require a separate router queue.
Unit: duration. Required evidence: application-queue-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
Capacity: Commitment coverage and savings
METRIC-CAPACITY-07. Eligible usage/spend covered by valid commitments; realized cost delta using matched contract terms. Also expose the source-required outputs: realized_commitment_savings_rate.
Boundary: Allocated commitment economics require contract dates, amortization and usage scope; not all invoice sources expose these fields.
ROI relationship: 22
Required dimensions: contract; billing_period
METRIC-CAPACITY-07.
Compatible eligible billed usage quantity covered by valid commitments divided by eligible billed usage quantity. Missing invoice quantities withhold this output.
Unit: ratio. Required evidence: invoices; commitment-terms.
Denominator: reference: eligible_billed_usage_quantity; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-CAPACITY-07.
Eligible spend covered by valid commitments divided by eligible spend under the same amortization and term rules.
Unit: ratio. Required evidence: invoices; commitment-terms.
Denominator: reference: eligible_commitment_scope_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-CAPACITY-07.
Matched uncommitted contract-basis cost minus actual amortized committed cost, including unused commitment expense and explicit comparison assumptions.
Unit: USD. Required evidence: invoices; commitment-terms.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-CAPACITY-07.
Matched uncommitted contract-basis cost minus actual amortized committed cost, divided by nonzero matched uncommitted cost. Include unused commitment expense and identical scope, period and amortization. A currency delta is not this rate.
Unit: ratio. Required evidence: invoices; commitment-terms.
Denominator: reference: matched_uncommitted_contract_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Capacity: Elasticity under bursts
METRIC-CAPACITY-08. Matched burst intervals meeting a predeclared quality/latency/error SLO / eligible observed burst intervals
Boundary: Observed throughput correlation does not prove Venturi absorbed a spike; distinguish provider, customer app and serving tiers.
ROI relationship: 36
Required dimensions: burst_definition; slo_revision
METRIC-CAPACITY-08.
Matched burst intervals meeting a predeclared quality/latency/error SLO / eligible observed burst intervals. Observed throughput correlation does not prove Venturi absorbed a spike; distinguish provider, customer app and serving tiers.
Unit: ratio. Required evidence: usage; pipeline-observability.
Denominator: reference: eligible_observed_burst_intervals; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Capacity: Capacity forecast accuracy
METRIC-CAPACITY-09. Frozen-vintage forecast versus actual eligible demand at the declared horizon and aggregation grain. Also expose the source-required outputs: frozen_token_forecast, actual_token_demand.
Boundary: Mature matched observations are required; no retrospectively changed forecasts or fabricated M6/M12 history.
ROI relationship: 34/36
Required dimensions: forecast_vintage; forecast_horizon_months; target_period
METRIC-CAPACITY-09.
Actual eligible tokens minus the frozen-vintage token forecast for a mature matched horizon and grain.
Unit: tokens. Required evidence: usage; forecast-vintages.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-CAPACITY-09.
Signed token forecast error divided by the absolute frozen forecast tokens; zero forecast preserves token error and withholds this ratio.
Unit: ratio. Required evidence: usage; forecast-vintages.
Denominator: reference: absolute_frozen_forecast_tokens; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-CAPACITY-09.
Projected eligible token demand from a frozen vintage at the separately identified six-month or 12-month horizon. Record issue time, target period, population and grain; actuals are not required for the forecast series.
Unit: tokens. Required evidence: forecast-vintages.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-CAPACITY-09.
Observed eligible token demand for the target period and population, independently available from a frozen forecast. Incomplete periods remain provisional and cannot be compared as mature forecast actuals.
Unit: tokens. Required evidence: usage.
Denominator: Not applicable. Unavailable state: missing_required_input.
Capacity: Idle committed-capacity cost
METRIC-CAPACITY-10. Eligible amortized reserved/committed cost not consumed under explicit contract and utilization rules
Boundary: Do not present provider-native parity functionality as a proven differentiator.
ROI relationship: 22.
Required dimensions: commitment; utilization_period
METRIC-CAPACITY-10.
Eligible amortized reserved/committed cost not consumed under explicit contract and utilization rules. Do not present provider-native parity functionality as a proven differentiator.
Unit: USD. Required evidence: invoices; commitment-terms.
Denominator: Not applicable. Unavailable state: missing_required_input.
Sustainability: Estimated energy per inference
METRIC-SUSTAINABILITY-01. Model/hardware/region/time-qualified energy model applied to compatible observed usage, with uncertainty
Boundary: Roadmap extension only; token factors are estimates, not metered customer energy. Unknown factors produce null.
ROI relationship: 44
Required dimensions: model_tier; factor_revision
METRIC-SUSTAINABILITY-01.
Model/hardware/region/time-qualified energy model applied to compatible observed usage, with uncertainty. Roadmap extension only; token factors are estimates, not metered customer energy. Unknown factors produce null.
Unit: kWh/1000 turns. Required evidence: usage; energy-factors.
Denominator: reference: eligible_observed_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Sustainability: Estimated carbon per feature
METRIC-SUSTAINABILITY-02. Compatible estimated energy multiplied by time/location-matched carbon factors and attributed to features/workflows
Boundary: State system boundary, provenance and uncertainty; attribution does not establish reporting-standard eligibility.
ROI relationship: 45
Required dimensions: feature; workflow; factor_revision
METRIC-SUSTAINABILITY-02.
Compatible estimated carbon attributed to the feature; preserve factor versions, system boundary and uncertainty.
Unit: kgCO2e. Required evidence: usage; energy-factors; carbon-factors; feature-map.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SUSTAINABILITY-02.
Compatible estimated carbon attributed to the registered workflow; missing feature mapping does not erase this separately scoped estimate.
Unit: kgCO2e. Required evidence: usage; energy-factors; carbon-factors; workflow-registry.
Denominator: Not applicable. Unavailable state: missing_required_input.
Sustainability: AI carbon intensity
METRIC-SUSTAINABILITY-03. Compatible estimated carbon divided by eligible nonzero scoped AI cost
Boundary: A changed price can change this ratio without changing physical efficiency; report physical and financial series separately.
ROI relationship: 45 (retain the declared dollar denominator only as a secondary diagnostic)
Required dimensions: cost_basis; factor_revision
METRIC-SUSTAINABILITY-03.
Estimated physical carbon series for the eligible population and period, independent of prices.
Unit: kgCO2e. Required evidence: usage; energy-factors; carbon-factors.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SUSTAINABILITY-03.
Financial cost series for the same eligible population and period, independent of environmental factors.
Unit: USD. Required evidence: usage; prices.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-SUSTAINABILITY-03.
Compatible estimated carbon divided by scoped nonzero AI cost; a price change alone is not physical efficiency.
Unit: kgCO2e/USD. Required evidence: usage; prices; energy-factors; carbon-factors.
Denominator: reference: eligible_scoped_ai_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Sustainability: Estimated caching carbon delta
METRIC-SUSTAINABILITY-04. Explicit comparable no-cache versus observed-cache energy scenario, then compatible carbon conversion
Boundary: Caching token discounts are not metered energy savings; do not claim actual avoided emissions without a supported counterfactual.
ROI relationship: 45
Required dimensions: cache_state; counterfactual_revision
METRIC-SUSTAINABILITY-04.
Explicit comparable no-cache versus observed-cache energy scenario, then compatible carbon conversion. Caching token discounts are not metered energy savings; do not claim actual avoided emissions without a supported counterfactual.
Unit: kgCO2e. Required evidence: usage; energy-factors; carbon-factors; cache-counterfactual.
Denominator: Not applicable. Unavailable state: missing_required_input.
Sustainability: Estimated routing carbon delta
METRIC-SUSTAINABILITY-05. Quality-compatible baseline versus routed energy scenario over matched work, then carbon conversion
Boundary: No same-quality or realized reduction claim from down-tiering alone; preserve both scenario provenance and uncertainty.
ROI relationship: 45
Required dimensions: routing_baseline; factor_revision
METRIC-SUSTAINABILITY-05.
Quality-compatible baseline versus routed energy scenario over matched work, then carbon conversion. No same-quality or realized reduction claim from down-tiering alone; preserve both scenario provenance and uncertainty.
Unit: kgCO2e. Required evidence: usage; routing-events; energy-factors; carbon-factors; outcomes.
Denominator: Not applicable. Unavailable state: missing_required_input.
Sustainability: Region and renewable context
METRIC-SUSTAINABILITY-06. Observed regional usage joined to applicable disclosed energy attributes; unmatched regions reported unknown. Also expose the source-required outputs: low_carbon_region_inference_share.
Boundary: A low-carbon region is not proof that a specific inference used renewable energy; cloud provider region alone is insufficient.
ROI relationship: 45
Required dimensions: region; carbon_threshold_revision; period
METRIC-SUSTAINABILITY-06.
Observed usage share by resolved region in the declared population; region is a dimension of this output.
Unit: ratio. Required evidence: usage; region-map.
Denominator: reference: eligible_observed_usage; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SUSTAINABILITY-06.
Eligible usage without a supported region match divided by eligible observed usage; never treat it as zero renewable usage.
Unit: ratio. Required evidence: usage; region-map.
Denominator: reference: eligible_observed_usage; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SUSTAINABILITY-06.
Scoped observed regional usage linked to compatible published energy attributes divided by eligible observed usage. This is contextual coverage, not proof of renewable energy used by an inference.
Unit: ratio. Required evidence: usage; region-map; energy-disclosures.
Denominator: reference: eligible_observed_usage; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-SUSTAINABILITY-06.
Eligible inference turns in regions meeting a versioned, time-matched carbon-intensity threshold divided by eligible inference turns. Unknown region or factor coverage cannot become zero; incomplete classification withholds the full-population share. This roadmap estimate does not prove renewable energy was consumed.
Unit: ratio. Required evidence: usage; region-map; carbon-factors; environmental-classification-policy.
Denominator: reference: eligible_observed_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Sustainability: Estimated emissions-intensity trend
METRIC-SUSTAINABILITY-07. Comparable estimated emissions per turn across matched windows with factor/model versions controlled
Boundary: Separate factor revisions from operational changes; no fabricated quarter-over-quarter history.
ROI relationship: 44/45
Required dimensions: quarter; factor_revision
METRIC-SUSTAINABILITY-07.
Comparable estimated emissions per turn across matched windows with factor/model versions controlled. Separate factor revisions from operational changes; no fabricated quarter-over-quarter history.
Unit: kgCO2e/turn. Required evidence: usage; energy-factors; carbon-factors.
Denominator: reference: eligible_observed_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Sustainability: Emissions disclosure data readiness
METRIC-SUSTAINABILITY-08. Eligible modeled emissions with required provenance and supported boundary / eligible modeled emissions
Boundary: Data readiness is not GHG/Scope conformity or assured reporting; identify exclusions and required customer review.
ROI relationship: 45 (coverage diagnostic)
Required dimensions: reporting_boundary; coverage_population
METRIC-SUSTAINABILITY-08.
Eligible modeled emissions with required provenance and supported boundary / eligible modeled emissions. Data readiness is not GHG/Scope conformity or assured reporting; identify exclusions and required customer review.
Unit: ratio. Required evidence: usage; energy-factors; carbon-factors.
Denominator: reference: eligible_modeled_emissions; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Sustainability: Provider environmental factor scorecard
METRIC-SUSTAINABILITY-09. Published comparable factors by model/hardware/region/time, with provenance, exclusions and uncertainty
Boundary: Venturi hosts external evidence; missing or non-comparable factors do not justify a provider ranking.
ROI relationship: 45 (external-factor dependency)
Required dimensions: provider; region; model; factor_revision
METRIC-SUSTAINABILITY-09.
Published comparable factors by model/hardware/region/time, with provenance, exclusions and uncertainty. Venturi hosts external evidence; missing or non-comparable factors do not justify a provider ranking.
Unit: gCO2e/token. Required evidence: energy-disclosures; carbon-factors.
Denominator: reference: compatible_factor_token_basis; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Sustainability: Avoided-human-work carbon scenario
METRIC-SUSTAINABILITY-10. Customer-evidenced displaced activity times an explicitly justified system-boundary emissions model
Boundary: Double-dependent roadmap hypothesis; released employee time does not imply eliminated commuting/building emissions.
ROI relationship: 45 only with explicit lifecycle boundaries and evidence, otherwise unavailable.
Required dimensions: human_baseline; outcome_class; carbon_scenario
METRIC-SUSTAINABILITY-10.
Customer-evidenced displaced activity times an explicitly justified system-boundary emissions model. Double-dependent roadmap hypothesis; released employee time does not imply eliminated commuting/building emissions.
Unit: kgCO2e. Required evidence: outcomes; human-baselines; carbon-factors.
Denominator: Not applicable. Unavailable state: missing_required_input.
Workflow Economics: Cost per workflow run
METRIC-WORKFLOW-01. Eligible run costs by declared basis, with attempt/failed/retry costs retained and explicit completion/acceptance state. Also expose the source-required outputs: fully_loaded_cost_per_completed_run.
Boundary: Inference-only cost is not fully loaded. Show included/missing components and unassigned paid work separately.
ROI relationship: 01 (run-cost drilldown retained)
Required dimensions: workflow_class; cost_basis; completion_cohort
METRIC-WORKFLOW-01.
Eligible paid cost per run on the declared basis, including failed and retried attempts. Identify missing cost components; inference-only cost is not fully loaded.
Unit: USD/run. Required evidence: usage; prices; workflow-registry; workflow-events.
Denominator: reference: eligible_workflow_runs; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-01.
Paid work explicitly unresolved by complete run attribution evidence. Missing lifecycle evidence cannot label all paid work unassigned.
Unit: USD. Required evidence: usage; prices; workflow-registry; workflow-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-WORKFLOW-01.
All applicable paid inference, tool, infrastructure, human review and allocated shared costs for the declared delivery cohort divided by distinct completed runs, by workflow class. Include failed attempts and retries. Complete-cost-basis evidence must establish included or non-applicable components; missing components withhold this variant while preserving observed cost.
Unit: USD/run. Required evidence: usage; prices; workflow-registry; workflow-events; complete-cost-basis.
Denominator: reference: eligible_completed_runs; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Cost per accepted outcome and yield
METRIC-WORKFLOW-02. All eligible scoped workflow spend including unsuccessful work / distinct unrevoked accepted outcomes; separately accepted outcomes / 1000 eligible turns
Boundary: Deduplicate acceptance, handle revocations and maturity, and return unavailable for zero/unknown denominators. No IDD dependency for observed managed work.
ROI relationship: 01/38
Required dimensions: workflow_class; acceptance_rule_revision
METRIC-WORKFLOW-02.
All eligible scoped workflow spend, including unsuccessful work and retries, divided by distinct mature unrevoked accepted outcomes; preserve the declared complete cost basis.
Unit: USD/outcome. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes.
Denominator: reference: distinct_mature_unrevoked_accepted_outcomes; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-02.
Distinct mature unrevoked accepted outcomes per 1000 eligible turns in the same cohort; missing prices does not hide yield.
Unit: outcomes/1000 turns. Required evidence: usage; workflow-registry; workflow-events; outcomes.
Denominator: reference: eligible_observed_turns; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Workflow unit-cost trend
METRIC-WORKFLOW-03. Quarter-over-quarter USD per accepted outcome for matched cohorts and cost bases; a separately named cost-per-run trend is available when acceptance evidence is unavailable
Boundary: Do not call a cost-per-run series cost-per-outcome. Preserve the quarter-over-quarter outcome comparison and its acceptance evidence; match workflow cohort, cost basis and evidence maturity across quarters.
ROI relationship: 01 (matched accepted-outcome trend)
Required dimensions: workflow_class; quarter; cost_basis; acceptance_rule_revision
METRIC-WORKFLOW-03.
Matched period cost-per-run trend; this is separately named and never substitutes for the accepted-outcome trend.
Unit: USD/run. Required evidence: usage; prices; workflow-registry; workflow-events.
Denominator: reference: eligible_runs_per_matched_period; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-03.
Quarter-over-quarter cost per accepted outcome using matched cohort, cost basis, maturity and revocation semantics.
Unit: USD/outcome. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes.
Denominator: reference: accepted_outcomes_per_matched_period; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Workflow portfolio spend mix
METRIC-WORKFLOW-04. Eligible spend share by registered workflow class; separately valued-workflow share if validated value inputs exist. Also expose the source-required outputs: high_roi_workflow_spend_share.
Boundary: Basic mix is producible conditionally. High-ROI share is unavailable without value/baseline data and an approved ROI definition.
ROI relationship: 02/12/30
Required dimensions: workflow_class; roi_threshold_revision
METRIC-WORKFLOW-04.
Eligible spend share by registered workflow class in a matched window and cost basis.
Unit: ratio. Required evidence: usage; prices; workflow-registry.
Denominator: reference: eligible_workflow_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-04.
Separately defined share for workflows with validated customer value evidence; any high-ROI classification additionally requires an approved complete ROI calculation.
Unit: ratio. Required evidence: usage; prices; workflow-registry; outcomes; outcome-values.
Denominator: reference: eligible_workflow_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-04.
Eligible cost assigned to workflows whose complete accepted-work ROI exceeds the versioned customer threshold divided by total eligible workflow cost. Use the governed ROI formula and matched value/baseline evidence. Unknown classifications cannot become low ROI or zero; retain basic mix and disclose unclassified scope.
Unit: ratio. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes; outcome-values; human-baselines; complete-cost-basis; roi-classification-policy.
Denominator: reference: eligible_workflow_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Work-evidence-decision lineage
METRIC-WORKFLOW-05. Eligible scoped cost traceable through work, registered evidence and a recorded versioned decision / eligible scoped cost
Boundary: Do not infer business decisions from dashboard views or assign full evidence lineage to cost-only records.
ROI relationship: 39
Required dimensions: cost_basis; lineage_revision
METRIC-WORKFLOW-05.
Eligible scoped cost traceable to registered work and actual lifecycle evidence divided by eligible scoped cost; acceptance and decisions are not required for this partial lineage.
Unit: ratio. Required evidence: usage; prices; workflow-registry; workflow-events.
Denominator: reference: eligible_scoped_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-05.
Eligible cost linked through work, valid evidence and an actual versioned decision divided by eligible scoped cost.
Unit: ratio. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes; decision-events.
Denominator: reference: eligible_scoped_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Outcome acceptance rate
METRIC-WORKFLOW-06. Distinct eligible completed/mature runs with unrevoked customer acceptance / eligible runs under the declared cohort rule
Boundary: Basic acceptance evidence belongs to Core; customer semantics/version and absence/rejection distinction are mandatory.
ROI relationship: 01/09/38
Required dimensions: workflow; acceptance_rule_revision
METRIC-WORKFLOW-06.
Distinct eligible completed/mature runs with unrevoked customer acceptance / eligible runs under the declared cohort rule. Basic acceptance evidence belongs to Core; customer semantics/version and absence/rejection distinction are mandatory.
Unit: ratio. Required evidence: workflow-registry; workflow-events; outcomes.
Denominator: reference: eligible_mature_runs; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Workflow evidence coverage
METRIC-WORKFLOW-07. Eligible registered workflows/runs with valid evidence in a matched window / eligible registered workflows/runs
Boundary: Use distinct named workflow and run outputs. Evidence receipt is not a favorable outcome or production validation.
ROI relationship: 38
Required dimensions: workflow; run
METRIC-WORKFLOW-07.
Distinct eligible registered workflows with valid, current evidence divided by eligible registered workflows. Evidence need not indicate success.
Unit: ratio. Required evidence: workflow-registry; workflow-events; outcomes.
Denominator: reference: eligible_registered_workflows; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-07.
Distinct eligible runs with valid, current evidence divided by eligible runs under the matched cohort rule.
Unit: ratio. Required evidence: workflow-registry; workflow-events; outcomes.
Denominator: reference: eligible_runs; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Workflow cycle time and time economics
METRIC-WORKFLOW-08. Run end minus run start from true workflow boundaries; saved time/value only against comparable customer baseline and approved rates
Boundary: First-to-last inference span excludes other workflow work. Released time is capacity, not automatically cash savings.
ROI relationship: 04/08
Required dimensions: workflow_class; human_baseline; valuation_basis
METRIC-WORKFLOW-08.
Run end minus run start using true workflow boundaries and explicit unfinished or censored states.
Unit: duration. Required evidence: workflow-registry; workflow-events.
Denominator: Not applicable. Unavailable state: missing_required_input.
Statistical contract: statistics: observation; mean; p50; p95; result identity fields: statistic; observation_id_if_observation; aggregation_window; population_revision; empty population: unavailable; percentile composition: recompute_from_eligible_observations
METRIC-WORKFLOW-08.
Comparable customer baseline duration minus observed workflow duration for the same accepted work and quality scope; duration is not a cash saving.
Unit: duration. Required evidence: workflow-registry; workflow-events; outcomes; human-baselines.
Denominator: Not applicable. Unavailable state: missing_required_input.
METRIC-WORKFLOW-08.
Comparable released time valued at approved customer burdened rates; keep modeled value separate from realizable cash savings.
Unit: USD. Required evidence: workflow-registry; workflow-events; outcomes; human-baselines; burdened-rates.
Denominator: Not applicable. Unavailable state: missing_required_input.
Workflow Economics: Workflow margin
METRIC-WORKFLOW-09. Attributed revenue/price less explicitly scoped workflow direct and allocated cost, divided by eligible nonzero revenue
Boundary: Cost attribution alone cannot produce margin; distinguish contribution margin from fully burdened margin.
ROI relationship: 06
Required dimensions: workflow_class; price_revision; cost_basis
METRIC-WORKFLOW-09.
Attributed revenue/price less explicitly scoped workflow direct and allocated cost, divided by eligible nonzero revenue. Cost attribution alone cannot produce margin; distinguish contribution margin from fully burdened margin.
Unit: ratio. Required evidence: usage; prices; workflow-registry; workflow-events; revenue; allocation-policy.
Denominator: reference: eligible_workflow_revenue; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Workflow Economics: Workflow ROI and human-versus-AI cost
METRIC-WORKFLOW-10. (Incremental attributable benefit before the included costs minus complete incremental cost) / complete incremental cost; compare like-for-like accepted task costs. Also expose the source-required outputs: ai_cost_per_accepted_task, human_cost_per_accepted_task.
Boundary: Outcome value and human baseline are customer inputs; no automatic savings, causality, or ROI claim from adoption.
ROI relationship: 01/02/04.
Required dimensions: workflow_class; comparison_side; value_baseline_revision
METRIC-WORKFLOW-10.
Human baseline cost per accepted task minus comparable AI cost per accepted task, a single signed USD/task difference. Retain the separately named human and AI levels. Require complete monetary baselines, all paid attempt lifecycle links, matched acceptance semantics and authorized rates for time-only baselines; this difference is not proof of realized savings.
Unit: USD/task. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes; human-baselines; complete-cost-basis.
Denominator: reference: like_for_like_accepted_tasks; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-10.
Incremental attributable benefit before included costs minus complete incremental cost, divided by that nonzero cost; reconcile net-benefit inputs to avoid subtracting costs twice. Link every eligible paid attempt, including failed retries, through the run lifecycle before comparing accepted tasks.
Unit: ratio. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes; human-baselines; outcome-values; complete-cost-basis.
Denominator: reference: complete_incremental_cost; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-10.
Complete eligible AI delivery cost, including failed attempts, retries and applicable review or shared costs, divided by distinct eligible accepted tasks. This AI series does not require a human cost baseline.
Unit: USD/task. Required evidence: usage; prices; workflow-registry; workflow-events; outcomes; complete-cost-basis.
Denominator: reference: eligible_accepted_ai_tasks; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
METRIC-WORKFLOW-10.
Complete monetary human baseline cost divided by distinct comparable accepted human tasks, using the customer task and acceptance definitions. This human series does not require AI price feeds; time-only baselines require authorized rate conversion.
Unit: USD/task. Required evidence: workflow-registry; outcomes; human-baselines; complete-cost-basis.
Denominator: reference: eligible_accepted_human_tasks; required state: nonzero_compatible; unavailable reasons: zero_denominator; unknown_denominator; incompatible_denominator. Unavailable state: missing_required_input.
Data handling¶
- Metadata-only capture
- Venturi captures invocation metadata, never prompt or completion content. There is no content field in the event schema and no setting that turns content capture on.
- Minimum cohort size (k = 5)
- Every cohort read path (adoption views, dashboards, exports, API responses, board packs, and webhooks) enforces a server-side minimum cohort size of 5. A smaller cohort rolls into its parent unit and cannot be reconstructed by differencing overlapping cohorts. An Admin can raise the threshold but never lower it below 5. Individual-level adoption views do not exist.
- WORM (write once, read many)
- The audit log’s storage contract. Entries are append-only in the product, and the audit stream is stored under S3 Object Lock with a 5-year retention period. Governance mode is the default: only a principal holding the explicit bypass-governance-retention permission can shorten or remove a lock early, and that use is recorded. Compliance mode, in which no user (including the account root user) can shorten or remove a lock early, is configured at onboarding for deployments that require it. See Audit logs.
- Crypto-shred
- Erasure by destroying the per-subject or per-tenant encryption keys, completed within 30 days and backed by a deletion certificate.
- Break-glass access
- The only support access path: customer-approved, time-boxed, MFA-gated, audit-logged, and visible to you. There is no standing support access. See Support-access architecture.
Product surfaces¶
- Projection
- A decision surface derived from the attribution graph (for example the chargeback view or the coverage map), not a standalone module.
- Adoption intelligence
- Attributed usage signals about organizational uptake of AI, reported cohort-only. It is not HR analytics and is never used for individual performance scoring.
- Decision surface state
- The evidence-use label on a product surface: observe-only (informational), decision-support (human-guided decisions with caveats), or operationally trusted (evidence quality supports tighter operational reliance).
- Detected-only
- A pathway or source observed through discovery evidence without enough request-level evidence for attribution. It remains visible in the discovery ledger and is excluded from chargeback.
- Fan-out warning
- A dashboard warning that one upstream action appears to be triggering a disproportionately large volume of downstream AI calls, so multiplied cost is investigated rather than misread as normal demand.
- Model Energy Rating
- The 1-to-5 star rating shown on model energy surfaces. It is computed at render time as a relative 20% quintile band of stored watt-hours per query within one task (and, for text generation, one size class) and one catalog version, so a star is comparable only within that population and can change when a new catalog version recalibrates the bands. See Energy methodology.
- System of record
- The authoritative, replayable record for a class of business data. Venturi is the system of record for AI. It joins activity, accountable ownership, connected workflows and systems, normalized cost, and accepted business outcomes. Customers choose domain offerings for their decisions; Workflow Economics Core is not a universal prerequisite. Each numerical result retains its source, scope, period and evidence state.
Sign in to send feedback. Opens in a new tab.
Copy page details
Docs feedback: Glossary
Page: https://docs.venturi.systems/reference/glossary/