DECISION

**COMPLETE RESERVED MODEL.** Publish WM-REC-002 *Report / Statement* as a standalone aggregate owning two identities — the reusable **report-definition version** and the issued **report statement (issue)** — with every execution, artifact, publication, edition and disclosure-shape concern held as an external reference.

BOUNDARY PROOF

Tested candidate by candidate against dossier text, not intuition.

*Owned.* (1) **Reusable report-definition versions.** WM-ACT-053 excludes reusable definitions by name ("Pipeline, job and workflow definitions … remain external"), and its only candidate parent is WM-DAT-005 Data Pipeline, an execution graph, not an analytical specification. No dossier model owns a reusable report spec. (2) **Parameter and reporting-grain contract.** WM-ACT-053 owns *resolved* parameters of one execution; the declared contract that constrains all executions has no owner. Grain is two-layered: WM-XCT-003 owns the *permitted* grain (a ceiling on what may leave), WM-REC-002 owns the *asserted* grain (what the statement claims). (3) **Report issue identity.** Not a WM-REC-001 version — that model's out-of-scope excludes "subject-matter semantics of the record's content (… financial postings)", which is exactly what a statement asserts. Not a WM-MED-003 edition (bibliographic identity, no parameter contract). Not a WM-ACT-044 case (an operation, not a claim). (4) **Reporting period, cutoff, as-of.** WM-ACT-053 has "effective data interval" and watermark as execution scoping; the asserted period is a semantic commitment that survives re-execution. (5) **Correction/restatement lineage.** Four neighbouring lineages exist and none says *the number changed because an input fact changed*: ACT-053 rerun/backfill (execution), ACT-044 correction/takedown (delivery propagation), MED-003 erratum/corrigendum (bibliographic), REC-001 supersedes (record version). Restatement class and reason are owned here.

*External references, with owned pinning.* Source snapshots and runs → WM-ACT-053 (the issue owns the echoed identifier tuple, never the run lifecycle). Disclosure binding → WM-XCT-003, stored as the exact output of its `assemble-disclosure-provenance-tuple`. Artifacts → WM-REC-001 record/version/instantiation. Publication → WM-ACT-044 case; edition → WM-MED-003. Metric and query references → reference-first, but the dossier names no metric-definition owner, so an inline measure binding pinned to a WM-DAT-004 schema version is the interim (see H2).

*Rejected alternatives.* REUSE ONLY / composition-only leaves the pinned tuple homeless: nothing owns the join of definition version + parameters + snapshots + cutoff + policy version, so the negative case (arbitrary HR–finance join) has no object to refuse it. PROFILE of WM-REC-001 collapses statement identity into a document version and fails when an issue has zero artifacts or many. NEW MODEL is unwarranted: the reservation already carries the matching purpose, NAV.INF.REC.RPT, legacy alias N2 and `factor_overlap` 0.09. The flag "часто subtype N1" describes a frequent instantiation pattern, not identity; adopt REFERENCES, not subtype.

MINIMAL CONTRACT

**Identities (all disjoint):** `reportDefinitionVersionId` and `reportIssueId` (owned); `runId` (ACT-053), `recordId/versionId/instantiationId` (REC-001), `publicationCaseId` (ACT-044), `editionId` (MED-003), `policyVersionId + templateFingerprint` (XCT-003) — all foreign.

**Lifecycle.** Definition version: draft → issued → superseded → withdrawn. Issue: planned → computed → issued → superseded | withdrawn. No issue returns to computed.

**Core fields (18):** 1 reportDefinitionId · 2 reportDefinitionVersionId · 3 definitionVersionState · 4 parameterContract (name, type, domain, required, default) · 5 assertedGrain (kept dimensions + level) · 6 measureBinding (metric ref or expression + pinned schema version) · 7 reportIssueId · 8 issueState · 9 boundParameterValues · 10 reportingPeriod · 11 dataCutoff (RFC 3339) · 12 asOfBasis (instant + transaction/valid-time code) · 13 inputSnapshotPin (dataset version, partition, digest) 1..n · 14 executionEvidenceRef (runId) 1..n · 15 disclosureProvenanceTuple · 16 suppressionManifest (slice key, basis ref, withheld count — never withheld values) · 17 restatementLink (supersedes, class, reason) · 18 artifactRef / publicationCaseRef 0..n.

**Invariants.** I1 An issued issue is immutable; change requires a successor. I2 Issue cannot reach *issued* without fields 2, 9, 13, 11, 12, 14, 15. I3 Bound values validate against the parameter contract of that exact definition version. I4 Asserted grain is no wider than the meet of applicable shapes; subsumption is decided by WM-XCT-003, not re-implemented. I5 reportingPeriod.end ≤ dataCutoff ≤ asOf ≤ issuedAt. I6 Every referenced run's effective data interval covers the reporting period, or a coverage exception is declared. I7 Correction creates a successor; the predecessor is retained in issued/superseded state, never overwritten or deleted. I8 The restatement chain is acyclic with one head per (definitionVersion, parameters, period). I9 The five identity classes are disjoint; no field may carry one as another. I10 Any withheld slice appears in the suppression manifest with the withholding policy version; silent omission is prohibited. I11 Every source in a definition version resolves to a pinned schema/contract version and a binding reference; an unresolved path blocks issuance. I12 Publication is not issuance: an issue may exist unpublished, and a publication case without a valid issue reference may not claim report status. I13 The explanation tuple stays resolvable for the retention period assigned under WM-REC-001, including after template withdrawal (tombstone).

**Functions.** F1 *Version report definition* — fix an immutable definition version with its parameter contract, grain and measure bindings. F2 *Issue report statement* — bind parameters, pin snapshots, cutoff, as-of, run evidence and disclosure tuple; delegates execution, never performs it. F3 *Restate* — create a successor issue with class and reason, retaining the predecessor. F4 *Explain issue* — return the full reconstruction tuple including suppression manifest, artifacts and publications.

DASHBOARD DISPOSITION

A **dashboard definition is a profile of the report-definition version**: same parameter contract, asserted grain, measure bindings and disclosure binding, with added layout and refresh attributes that WM-REC-002 does not own. A **dashboard view is not a report issue** — it is an ephemeral evaluation with no pinned cutoff, and WM-REC-002 does not model it; a citable view must be converted by snapshot-on-cite, minting a real issue under F2. It is *not* a "governed projection": WM-XCT-003 projection is output-shape policy, not a definition. The negative case is blocked structurally by I11 plus I4 — a panel whose sources are not declared in a governed definition version cannot be evaluated, so arbitrary joins never reach disclosure evaluation to bypass it.

ACCEPTANCE WALKTHROUGH

Issue: F2 mints issue A pinning definition v1.2, parameters, three snapshot digests, cutoff, as-of, run R1 and policy tuple P7/fingerprint F9; a sensitive region slice is withheld and recorded in A's suppression manifest with P7 (I10). A is published via an ACT-044 case referencing A (I12). Correct: an input fact is restated upstream; a new snapshot digest appears. A is untouched (I1, I7). Successor: F3 mints issue B, same definition version and parameters, same period, `restatementClass = input-restatement`, supersedes A, new run R2. A remains issued-and-superseded, still explainable via F4 from its own pins (I13). Suppression: B re-evaluates under P7 as pinned, so the sensitive slice is suppressed under the *exact* version that governed A, even if a newer policy exists. Dashboard drift: changing a dashboard profile alters no issue, because A and B carry their own pins — explainability is independent of the live view.

HOLDS AND PUBLICATION RECOMMENDATION

H1 *Inherited-neighbour hold*: all five neighbours are reviewable-draft with `publishableCanonical: false`, and ACT-044, ACT-053 and MED-003 are single-provider waivers; WM-REC-002 cannot claim a stronger status than its weakest referenced authority. H2 *Metric ownership gap*: no dossier model owns metric definitions; field 6's inline fallback must be revisited when an owner exists. H3 *Time-model gap*: WM-REC-001 excludes calendar and period arithmetic, so fiscal-calendar semantics behind fields 10–12 are unowned. H4 *Subsumption is declarative*: WM-XCT-003 carries an open multi-grain validation hold, so I4 is a declared check, not a demonstrated one. H5 *Single-reviewer hold*: this adjudication is one independent reviewer; the definition/issue split needs adversarial second-provider review before canonical status. H6 *Alignment-only*: state alignment and mapping, with lossiness recorded; no conformance claims. H7 *Live-view deferral*: the runtime dashboard-view concern needs a registry placeholder decision.

Recommendation: publish WM-REC-002 at `0.1.0-research.1`, `entry_kind: aggregate`, `status: reviewable-draft`, `publishableCanonical: false`; add candidate REFERENCES rows to WM-ACT-053, WM-ACT-044, WM-REC-001, WM-XCT-003 and WM-MED-003 — no CONTAINS, and explicitly reject the N1 subtype reading.