M03 OFFLINE EXPERIMENT BRIEF INPUT
Blank entries remain unknown and may block launch. This is a planning checklist, not a Meta import file.

[ ] Exactly two attorney-approved creative IDs and approval references
[ ] One business question; the single creative variable changed
[ ] Primary metric definition, data source and attribution window
[ ] Overall firm allowance, currency and maximum experiment portion, all-in
[ ] Fee/tax estimate and source; unknown means launch blocked
[ ] Intended start, end and fixed review dates, with time zone
[ ] Allocation assumptions, no automatic scaling and what remains constant
[ ] Named reviewer and person authorized to stop the test
[ ] Current account screenshot/capture of eligible experiment options and displayed requirements
[ ] Aggregate historical volume or explicit unknown
[ ] Privacy-safe measurement path and failure detection; no raw leads/case outcomes tied to people
[ ] Stop conditions: supplied all-in cap, broken measurement, policy/privacy issue, end date

FILL-IN FIELDS
Creative A ID and approval:
Creative B ID and approval:
One question:
One variable changed:
Primary metric definition/source/attribution window:
Overall allowance and currency:
Experiment allowance, all-in:
Known fees/taxes and evidence:
Dates/time zone:
Reviewer and stop authority:
Eligibility evidence and capture date:
Historical aggregate volume:
Measurement/privacy evidence:
Allocation assumption and constant conditions:

OPTIONAL EXISTING RESULTS
Attach an authorized aggregate platform export only if available. Record its exact original headers and reporting dates. Preserve variant, spend, primary outcome counts and any actual confidence/result fields. Missing confidence is unknown, not zero and not an invented estimate. State whether delivery was an actual eligible randomized experiment or an ordinary unequal-delivery report.

OUTPUT REVIEW CHECKLIST
One question, two approved creative IDs, one primary metric, fixed review dates, all-in cap, known blockers, stop authority, result rule including inconclusive, and no automatic scaling.