# 24 — Adversarial Principal Review of the Four Decision Documents Reviewed: `20-mapping.md`, `21-gaps.md`, `22-contracts.md`, `23-roadmap.md` against the spec digests (`00..06`), the ERP evidence maps (`10..13`), and the original spec at `files/manufacturing_spec/markdown`. Verdict up front: **the entity coverage is complete and the GL/stock rails are respected, but the four documents do not agree on the module's identity (naming/namespace), the roadmap silently reverses two mapping decisions, and several roadmap phases have unscheduled prerequisites.** None of this is fatal; all of it will cause rework if the docs are handed to builders as-is. --- ## 1. Spec entities silently dropped from the mapping **Result: NONE dropped.** All 27 entities in `06-spec-integration.md` §6.2 are present in the mapping's decision table (rows 1–27): Item, BOM_Header, BOM_Line, Routing_Header, Routing_Operation, Work_Center, Tool (7 master data); MPS_Header, MPS_Line, Item_MRP_Settings, MRP_Planned_Order, CRP_Load (5 planning); Production_Order_Header, Order_Component, Order_Operation, Material_Issue_Header/Line, Confirmation_Header/Yield/Detail, Goods_Receipt_Header/Line (10 execution); Standard_Cost, Cost_Center, Variance_Record (3 costing); Pegging, Delivery_Schedule (2 cross-cutting). The SFC layer (spec 05) is correctly treated as a reuse of rows 15+18, matching spec 05 §5.3 ("does not introduce a new master table"). Near-misses worth recording (field/config-level, not entity-level): - **`cost_driver` has no home.** Spec 04 §4.2 makes `cost_driver ∈ {LaborHours, MachineHours, DirectMaterialCost, UnitsProduced}` the overhead-rate denominator and the mapping declares a `CostDriver` enum in the skeleton, but NO table in the mapping carries a `cost_driver` column (work center? overhead pool? settings key?). The confirmation OH posting (`cost_driver × overhead_rate`) cannot be computed without deciding where it lives. - **`MRP_Planned_Order.source_demand_id`** (spec §6.2) is not explicitly listed among `mfg_mrp_planned_orders` columns in mapping row 11 — pegging depends on it. - **Tally arithmetic is wrong:** mapping §2 claims "**20 NEW `mfg_*` tables**" but the document itself enumerates **24** distinct new tables (verified by grep: bom_substitutes, routing_headers, routing_operations, tools, mps_headers, mps_lines, item_mrp_settings, mrp_runs, mrp_planned_orders, mrp_exceptions, crp_loads, work_calendars, calendar_shifts, material_issues, material_issue_lines, confirmations, confirmation_yields, confirmation_details, goods_receipts, goods_receipt_lines, standard_costs, variance_records, peggings, delivery_schedules) — plus `mfg_piecework_log` appearing only in 22 (25th). Sizing inputs derived from "20" are understated. --- ## 2. Contract violations (house dependency rules / rail bypasses) **Positive finding first:** no GL bypass anywhere — every posting in all four documents goes through `Modules\Accounting\Actions\CreateJournalEntry`, and stock mutation is routed through `ApproveIssue`/`ApproveReceipt` with the new `reserve()/release()` correctly contributed INTO Inventory's `StockService`. The Sales-COGS "consuming module owns the GL" pattern is applied correctly. Violations and risks found: - **V1 (real violation): HRM reads a Manufacturing table.** Contracts §1 HRM row: piece-rate payroll feed = "a `mfg_piecework_log` **HRM payroll reads**". A foreign module querying an `mfg_*` table is exactly the inbound dependency that contracts' own Rule 2 ("No module ever imports Manufacturing") forbids. Fix: Manufacturing-side listener pushes rows into an HRM-owned staging table via a new `Modules\HRM\Actions\RecordPieceworkEntry`, mirroring how every other contract in the same document is shaped. - **V2 (violates the adopted decision): the entire contracts document is written for the wrong module.** 22 uses `Modules\Manufacturing\Events`, permission prefix `manufacturing` in `PermissionDependencyRegistry`, `SequenceService::generateNext(company,'manufacturing',…)`, `SettingsService::get('manufacturing.*')`, and tables `mfg_work_centers` / `mfg_item_settings` / `mfg_pegging`. The mapping (D-01, ratified in the roadmap) decided **extend `Modules/Production`**, prefix **`production.*`**, sequence module key **`'production'`** (already wired), settings keys **`production.*_account_id`** (mapping §2.4 row 25), and EXTEND `production_centers` (no `mfg_work_centers` table exists in the mapping). Implementing 22 as written re-creates the duplicate namespace/permission/sequence waste that mapping §1.4 explicitly calls "pure waste plus a deprecation project". - **V3 (bypass invitation):** Contracts §1 Inventory row lists `StockService::increaseStock/decreaseStock` among what Manufacturing "TAKES". Direct increase/decrease without an `InventoryIssue`/`InventoryReceipt` document breaks the document trail the mapping itself mandates ("never touch balances directly"). The contract should expose only `getIssueCost/getProductCost/getValuationMethod/reserve/release` plus the two Approve Actions. - **V4 (soft):** "CRP reads `cmms_work_orders` directly (read-only query)" — a raw cross-module table read instead of an Action/Service. Forward direction, so tolerable, but inconsistent with the document's own "Nothing else (no … direct table writes/reads except via Actions)" posture; a `CMMS\GetDowntimeWindows` query service would be recipe-clean. - **V5 (unverified reuse claim):** Mapping row 7 gives `mfg_tools.fixed_asset_id` an "FK → Accounting fixed asset for usage-based depreciation". **No evidence document (md/12 included) shows that Moon Accounting has any fixed-asset register at all.** Either the FK target does not exist (the claim is false and Accounting needs a fixed-asset build — a missed gap), or it was never audited. Roadmap Phase 2 books usage depreciation "into overhead" without any Accounting prerequisite for this. --- ## 3. Gap analysis quality (false gaps / missed gaps) **False gaps: none found.** Every Direction-A claim was verified against the evidence maps: A1/A2 (zero GL, zero stock movement — md/10 §5 grep evidence), A5 (`reserved_quantity` orphan — md/11 §3), A6 (ProductType Product|Service only — md/11 §1), A8 (no calendar — md/10/13), A10 (no `cmms_asset_meters` — md/13 §4), A11 (PR creation controller-only — md/11 §5), A12 (ApprovalModule Sales|Purchases only — md/12 §7), A13 (PayrollService derives hourly at runtime — md/13 §2), A14 (no ATP/CTP field on sales_orders — md/13 §1), A18 (no cost_centers.type — md/12 §4). The "excluded candidates" table (UoM engine, cost centers, sequences, DataScope, QMS) is also correct — these genuinely exist. One overstatement: the exclusions table says QMS "NCR/CAPA with `production_order_id` columns waiting" — per md/13 only `qms_inspection_plans`/`qms_inspections` carry `production_order_id`; `qms_non_conformances` links only via `inspection_id`. **Missed real gaps:** 1. **FOVV inputs do not exist.** Spec 04 expects "budget data per cost center and budgeted production volume" (expects-from-ERP #11); FOVV = (actual − **budgeted**) × std fixed OH/unit. Moon `cost_centers` has **no `budget_amount`/`manager_id`** (md/12 §4 column list) and no budgeting feature exists anywhere. The mapping hedges ("+ budget_amount if absent"); 21 has no gap entry; roadmap Phase 3 computes all 7 variances with no budget source scheduled. The 7-variance engine is structurally incomplete without this. 2. **`cost_driver` storage** (see §1) — neither a gap entry nor a mapped column. 3. **Accounting fixed-asset register** (see V5) — required by the mapping's own tool-depreciation design, never audited, never listed as a gap. 4. **Inventory `ReceiptReferenceType`/issue reference enums:** md/11 shows `reference_type => ReceiptReferenceType::Purchase` is an enum; adding `'production_order'` requires extending those Inventory enums — absent from 21 and from mapping §4's "touches outside the module" table (only `MovementType` is listed). 5. **`production_centers` capacity/rate columns as a scheduled work item** — mapping row 6 specifies ~14 new WC columns (daily_capacity, multiplier, efficiency, utilization, setup/labor/machine rates, labor_calc, calendar_id, bottleneck_flag…), but no gap entry and no roadmap phase ever schedules this EXTEND (see §4.1). Direction-B critique is strong (subcontracting B3, reversals B7, return-to-stock B14 are real spec blind spots correctly caught), and the "ATP/CTP is not a spec gap" exclusion is correct. --- ## 4. Roadmap: hidden prerequisites & sizing 1. **The work-center extension is scheduled nowhere.** Phase 2 posts labor/machine/OH from WC rates and computes effective capacity = daily × multiplier × eff% × util%, but no phase's table scope includes extending `production_centers` with those columns (Phase 1 adds only `cost_center_id`). Phase 4 CRP needs `calendar_id` + capacity fields too. Hidden prerequisite spanning Phases 2 and 4. 2. **Per-item `costing_method` ordering knot.** Phase 3 (costing) needs it; the mapping houses it on `mfg_item_mrp_settings`, which the roadmap builds in **Phase 4**. The roadmap "solves" this by migrating `products.cost_method` to an enum in Phase 3 — directly contradicting mapping row 1 ("do NOT bloat `products`") and row 23. Pick one home and one phase. 3. **Phase 3 FOVV has no budget source** (see §3.1) — either descope FOVV in v1 (declare 6 variances) or schedule a budget input. 4. **Phase 2 tool depreciation** depends on the unverified Accounting fixed-asset capability and the "usage-driven depreciation variant [that] is net-new in Accounting" (mapping row 7) — not in Phase 2's prerequisite list. 5. **Phase 1 reopens a settled decision:** "OR use Inventory documents directly with reference linkage (**recommended**; fewer tables)" — the mapping decisively created `mfg_material_issues`/`mfg_goods_receipts` precisely because Inventory documents lack `issue_type`/`issue_level`/`operation_no`/grade fields. If the roadmap's recommendation is followed, those spec fields have nowhere to live; if not, the sentence is noise. Also Phase 1 `substitute_items` JSON contradicts mapping's `mfg_bom_substitutes` table. 6. **The "addendum-first" mandate has no owner.** 21's headline: every spec-side gap (B1–B25) "must be resolved as a written addendum BEFORE the build order starts, or it resurfaces as rework". The roadmap allocates zero time/phase to authoring ~25 addenda; D-01..D-14 cover only some of them (B5, B6, B11, B19, B20 etc. have no decision row). 7. **Sizing legends conflict:** 21 defines S ≤ 3 dev-days / M ≤ 3 dev-weeks / L > 3 dev-weeks; 23 defines S ≈ ≤1 wk / M 2–4 wks / L 4–8 wks. A board member mapping gap-sizes onto phase-sizes will mis-budget. Phase 1 itself (BOM versioning + multi-level + full state-machine rename rippling into Angular atomically + two posting rails + reservations + FE dialogs) is at the aggressive end of "L" for a single senior dev; the bottom-line "factory genuinely operational at END OF PHASE 1" assumes none of the D-decisions stall. 8. **Quality-status orphan window:** GR lines carry `quality_status` resolved from QMS (mapping row 22), but QMS hooks arrive only in Phase 6 — receipts shipped Phases 1–5 need a defined default behaviour (auto-Released?) that no document states. --- ## 5. Contradictions between the four documents | # | Topic | 20-mapping | 21-gaps | 22-contracts | 23-roadmap | |---|---|---|---|---|---| | C1 | Module identity / namespace | Extend `Modules/Production`, permission prefix `production.*`, sequence key `'production'` | (follows 20) | **`Modules\Manufacturing`**, prefix `manufacturing`, sequence `'manufacturing'` | Extend Production (D-01) | | C2 | Settings keys | `production.*_account_id` (§2.4 row 25) | `manufacturing.*_account_id` (A7) | `manufacturing.*` (§3 #13, §4) | `manufacturing.*` (Phases 0/1/2) | | C3 | Item-settings table | `mfg_item_mrp_settings` | `mfg_item_settings` (A6) | `mfg_item_settings` | `mfg_item_mrp_settings` | | C4 | Pegging / calendar tables | `mfg_peggings`, `mfg_work_calendars`+`mfg_calendar_shifts` | `mfg_calendars` (A8) | `mfg_pegging` | `mfg_pegging`, `mfg_calendars`+`mfg_calendar_shifts` | | C5 | Work-center store | EXTEND `production_centers` | — | **`mfg_work_centers`** (nonexistent table, §1 CMMS + §5) | EXTEND `production_centers` | | C6 | Sales promise column | `sales_orders.promised_date` | `promised_date` (A14) | `promised_delivery_date` | `production_promised_date` (D-12) | | C7 | Event names | `ConfirmationPosted`, `OperationCompleted`, `ProductionOrderClosed`, `MaterialIssued` | — | `OperationConfirmed`, `ProductionOrderSettled`, `MaterialIssued` | `MaterialsIssued`, `FinishedGoodsReceived`, `ProductionOrderClosed`, `OperationCompleted` | | C8 | BOM substitutes | `mfg_bom_substitutes` child table | — | — | `substitute_items` JSON v1 | | C9 | Issue/GR documents | NEW `mfg_*` doc tables (definitive) | — | (follows 20) | "OR use Inventory documents directly (recommended)" | | C10 | Consignment model | `warehouses.is_consignment` flag (rows 17) | either/or (A15) | "consignment bucket" (unspecified) | `ownership` dimension on balances/movements (Phase 6) | | C11 | Per-item costing-method home | `mfg_item_mrp_settings`; "do NOT bloat products" | extension table (A6) | `mfg_item_settings` | `products.costing_method` enum migration (Phase 3) | | C12 | Labor-rate resolution | WC rate v1, HRM deferred (§4 XS) | WC rate recommended (A13) | NET-NEW `HRM\GetLaborRate` Action in the confirmation path | WC rate v1 (D-06) but HRM columns added as Phase 2 prerequisite | | C13 | Settlement action | `CloseProductionOrder` computes variances | `SettleProductionOrder` (A1) | `SettleProductionOrder` | `CloseProductionOrder` (Phase 1, no variances) | | C14 | NEW-table tally | "20 NEW" claimed, 24 enumerated | — | +`mfg_piecework_log` (25) | — | C1/C2 are the dangerous ones: registry contributor name, sequence module key, and settings-key namespace are **runtime identifiers** — if Phase 0 seeds `production.*` keys while the confirmation Action (built from 22) resolves `manufacturing.wip_account_id`, postings fail at the first material issue. --- ## 6. What is sound (credit where due) - Entity coverage is genuinely complete (27/27) with defensible verdicts; the extension-table pattern for Item and the Accounting-owns-cost-centers ruling resolve the spec's two ownership ambiguities correctly. - Evidence discipline is high: every Direction-A gap traces to a file:line in md/10–13, and no claimed gap turned out to exist in the ERP. - The three roadmap inversions (integrations woven in, execution before planning, MPS-lite) are well-argued and consistent with the spec's own contradiction (real-time GL rule vs integrations-last build order). - The GL/stock rails are never bypassed; the JE family table (22 §4) matches md/12 §9 exactly. ## 7. Required fixes before board sign-off 1. Rewrite 22 against the D-01 decision: `Modules\Production`, `production.*` permissions, `'production'` sequence key, ONE settings namespace (recommend `production.*` to match the existing permission/sequence space), `production_centers` not `mfg_work_centers`. Or, if the board prefers the `manufacturing` namespace, change D-01's naming addendum — but pick one. 2. Re-route the piecework payroll feed through an HRM-owned Action/table (V1) and drop `increaseStock/decreaseStock` from the Manufacturing-facing Inventory contract (V3). 3. Audit Accounting for a fixed-asset register; add a gap entry (and Phase 2 prerequisite) or drop `fixed_asset_id` from `mfg_tools` v1. 4. Add gap entries + phase scheduling for: FOVV budget inputs, `cost_driver` storage, `production_centers` capacity/rate columns, Inventory reference-type enums. 5. Resolve C6–C11 into single named decisions (add D-15+ rows); fix the "20 NEW" tally; align the two sizing legends; assign an owner + timebox for the B1–B25 addenda demanded by 21.