test(12): complete UAT - 7 passed, 0 issues

.planning/phases/12-orders-invoices-company-matching/12-UAT.md

@@ -0,0 +1,53 @@
+---
+status: complete
+phase: 12-orders-invoices-company-matching
+source: [12-01-SUMMARY.md, 12-02-SUMMARY.md, 12-03-SUMMARY.md, 12-04-SUMMARY.md, 12-05-SUMMARY.md]
+started: 2026-07-11T11:29:27Z
+updated: 2026-07-11T11:35:57Z
+---
+
+## Current Test
+
+[testing complete]
+
+## Tests
+
+### 1. Cold Start Smoke Test
+expected: Restart the app container from scratch. Migrations 093/094/095 apply (or are already applied) without error, the app boots, and a basic health query against pax8_order_items/pax8_companies returns live data.
+result: pass
+
+### 2. Full PAX8 sync populates historical order items
+expected: Triggering a full PAX8 sync (POST /api/pax8/sync) populates pax8_order_items with per-company id, billing period, and dual-cost data across the full historical invoice set (not a partial subset).
+result: pass
+
+### 3. Confident company auto-matches are created and correct
+expected: pax8_companies rows exist with autotask_company_id set and match_confidence >= 0.90 (pg_trgm), and a sample of these auto-matches are genuinely the same real company on both sides.
+result: pass
+
+### 4. Ambiguous/no-match companies are flagged, never silently guessed
+expected: PAX8 companies with no clear Autotask match, or multiple similarly-scored candidates, get a pax8_company_match_review row instead of being auto-linked.
+result: pass
+
+### 5. Re-running sync is idempotent for resolved matches
+expected: Running the sync a second time does not change the auto-match set already established (same pax8_company_id/autotask_company_id/match_confidence triples).
+result: pass
+
+### 6. Partner cost figures retain full decimal precision (CR-01 regression check)
+expected: partner_cost/partner_cost_total values with more than 2 decimal places (e.g. real PAX8 costs like 0.0182 or 22.176) are stored without silent rounding.
+result: pass
+
+### 7. Sync History does not show fabricated deletion counts (CR-02 regression check)
+expected: A sync run that flags companies for manual review does not report those flagged companies as "records_deleted" in sync_history / the admin Sync History table.
+result: pass
+
+## Summary
+
+total: 7
+passed: 7
+issues: 0
+pending: 0
+skipped: 0
+
+## Gaps
+
+[none]