CRM pipeline acceptance worksheet — October 17, 2026
Editorial worksheet, not a product import or completed security assessment.

Data owner / reviewer:
Approved source report and snapshot:
Last successful synchronization / allowed lag:
Reporting database, view, grain and unique key:
Stage ID mapping, expected-close window and time zone:
Amount basis / currency and conversion policy:
Probability source and scale:
Permitted audience, rows and columns:
Application, execution identity, model destination and retention:

For each test record: question, reference IDs/counts/totals, actual query,
actual result, missing-value disclosure, clarification, pass/fail, reviewer.
Keep failed and unresolved questions in the denominator.

Synthetic reference cases (not observed AI behavior):
1. Organization 10, open USD, Q4 2026: IDs 1/2/3/4; 4 deals,
   3 known amounts, known pipeline 38000; 2 weighted deals, subtotal 10000.
2. Alex only: 2 deals, 30000 known, 10000 weighted.
3. Blair only: 2 deals, 1 known amount=8000; 0 weighted, NULL subtotal.
4. At least one contact: IDs 1/2; 2 deals, 30000 known.
5. Missing usable probability or amount: IDs 3/4.
6. Q1 2027: ID 6, 9000 known, 4500 weighted.
7. EUR only: ID 7, 7000 known, 3500 weighted; never add to USD.
8. Second organization: ID 1 is distinct; SQL scope is not authorization.
9. Another contact on deal 1 must not change base pipeline totals.
10. A probability outside 0..1 must be flagged/excluded from weighting.

Deployment tests, to execute separately with approved identities:
- Unauthorized owner/team/org request: no forbidden data.
- Hidden field requested indirectly or via export: no forbidden data.
- Revoked session and source permissions: intended denial behavior.
- Refresh failure: visible freshness failure, not a claim of live data.

Rollout decision / unresolved cases / next review date: