Cohort Integrity Review
Rebuild retention and activation cohorts from raw history so charts stop lying after SDK upgrades.
Retention moves for reasons that have nothing to do with the product: silent user-id resets, timezone mishandling, or onboarding renames that orphan day-one events.
This review rebuilds cohort definitions from raw history, checks identity continuity, and separates measurement noise from real behaviour change.
Deliverables include a cohort outline analysts can maintain, a list of artefacts to stop treating as insight, and limits on what the sample can prove.
Included
- Cohort definition rebuild
- Identity continuity checks
- Noise vs signal memo
- Analyst handoff notes