# Cost accountability rehearsal — source and scope

6 October 2026. Separate synthetic development module; no engine, SOW or production
integration. Builds on `reports/FinOps and AI economics extension.md` and its
`research_notes/FinOps and AI economics extension/` source notes. Independent
read-only source sidecar rechecked these five official pages on 6 October 2026.

| Source | Relevant support | Applicability limit |
| --- | --- | --- |
| [FinOps Cloud Cost Allocation](https://www.finops.org/wg/cloud-cost-allocation/), updated 16 March 2026 | Accounts, hierarchy and metadata support attribution; unallocated costs are visible | Advisory. Does not mandate our lifecycle schema, retention period or guarantee deleted-resource reconstruction. |
| [FOCUS 1.4 §4.4 Split Cost Allocation](https://focus.finops.org/docs/specification/v1-4/attributes/data-generator-calculated-split-cost-allocation-handling/) | Split summable amounts reconstruct their origin; unused capacity can be exposed | Normative within applicable FOCUS generator-calculated split semantics; this rehearsal is not a FOCUS implementation or conformance claim. Residual bucket and drivers here are local choices. |
| [FOCUS 1.4 Billed Cost](https://focus.finops.org/docs/specification/v1-4/columns/cost-and-usage/billed-cost/) | Invoice-based amounts and applicable adjustments; avoid duplicating covered purchases | An estimate is not actual billed cost. No real export, contract or complete invoice is validated here. |
| [FOCUS 1.4 Effective Cost](https://focus.finops.org/docs/specification/v1-4/columns/cost-and-usage/effective-cost/) | Usage-period recognition may differ from billed amounts, including purchase allocation | Effective cost is neither a second bill nor necessarily an estimate. Do not add these views. |
| [FinOps Unit Economics](https://www.finops.org/framework/capabilities/unit-economics/), live undated page | Defined cost scope and outcome metrics, distinct from resource metrics | Advisory. Our accepted-task denominator, human rubric, retries/rejected-work inclusion and review cost treatment are disclosed local measurement choices, not prescribed standards. |

## Local choices and exclusions

All identities, rates, dates, mappings, case facts and USD values are newly authored
teaching assumptions, not the completed MC² record, a client export or provider
prices. The three ledgers are separate and must not be summed. No DoD tagging
schema, component policy, government service availability or provider recommendation
is asserted. The earlier research's inaccessible DoD strategy remains unverified.
No standards version supersession is admitted into the engine.

The shared pool has a fixed 100,000-cent positive source and whole-percent driver
choices. Each scenario conserves assigned plus unallocated amounts exactly; no
rounding remainder occurs in these specific inputs. General currency conversion,
signed adjustments, commitment amortization and arbitrary split rounding are not
implemented. Historical attribution uses four explicitly selected evidence states,
not a general temporal join or collector. Cross-period corrections and actual
invoice reconciliation remain future work requiring client policy and evidence.

AI amounts reproduce the report's invented $16 technology / $150 review example.
Retrieval capacity, model inference, vector service and orchestration are disjoint
by explicit fictional assumption; idle is part of capacity, not an added charge.
Raw provider meter inclusion semantics need independent configuration-specific
validation before any actual use. The slider changes distinct accepted task count,
not spend or observed task quality. Remaining tasks are either rejected or pending
by explicit scenario choice. Review-inclusive unit cost is undefined at zero
accepted tasks and unavailable with unknown review cost; pending outcomes make it
provisional. The exact numerator/denominator is exported; display rounds half up
to cents. This is a scoped teaching cost, not total cost of ownership or savings.

## Canonical bridge

Inspected `src/artifacts.js` and `src/domain-workstreams.js` at baseline e5dc25296:
AR-031 → FIN-1/2; AR-032 → FIN-3/4; AR-035 and AR-042 → FIN-2/5. These are existing
assessment relationships, not new evidence requirements or demonstrated client
capabilities. EAO handles process/architecture handoffs; Service-AIOps adds the
measurement charter. Five-offering scope and 30/51/10/53 remain unchanged.