WAREHOUSE QUESTION ROUTING AND DEFINITION CHANGE WORKSHEET
Editorial worksheet, not an AskYourDatabase import or automation feature.
Synthetic example: billable usage. No product evaluation has been run.

QUESTION FAMILY RECORD (one per recurring or exploratory task)
Question and intended recipient:
Business owner / technical owner:
Route: maintained SQL / native assistant / connected assistant / clarification:
Reason for route:
Metric definition and version:
Snapshot, timezone and refresh requirement:
Permitted dimensions and filters:
Output: columns, grain, units, ordering, empty/unknown handling:
Execution identity and allowed objects/rows/columns:
Application, model, history and export locations:
Approved reference and held-out check:
Observed result / query / trace / date / product version:
Manual repairs, retries, compute and AI charges:
Publication decision and approver:

SYNTHETIC CHANGE REGISTER: v1 -> v2
v1: successful production units, including retry units.
v2: successful production units, excluding retry units.
Frozen fixture expected total: v1=220; v2=200; delta=20 from event_id=2.
A:120->100; B:100->100; C:0->0. Unit is usage units, not money.
Artifacts to inspect and update explicitly:
[ ] Reporting SQL/view/model and downstream report consumers
[ ] Metric definition, version and effective-date communication
[ ] Native assistant instructions/examples/semantic context where configured
[ ] Connected assistant documentation and SQL examples where configured
[ ] Held-out checks and expected results; preserve v1 reproduction case
[ ] Exports, labels and cached outputs; do not relabel historical values silently
[ ] Access tests after any view or identity change
All boxes are blank until an owner completes and records the work.

QUESTION EXERCISE
Q1 Fixed snapshot total under v1:220.
Q2 Same snapshot under v2:200.
Q3 v2 by account:A100,B100,C0.
Q4 Why 220 vs200? Event2 retry20; v1 and v2 definitions differ.
Q5 Which account generated most value? Clarify 'value'; units do not establish money.
Q6 Next month's total? Cannot determine from this snapshot.
Q7 Another audience's rows? Test actual enforced access; no expected refusal wording.
Q5-Q7 are proposed behavior checks, not observed AI/security test results.

ACTUAL PILOT OUTCOME (leave blank until tested)
Candidate/version/configuration:
Questions correctly answered with accepted scope:
Questions requiring clarification:
Incorrect or unsupported answers:
Unauthorized requests and enforced result:
Maintenance owner and rerun trigger:
Decision and supporting evidence: