cohort-analysislisted

# Cohort Analysis ## When to use this skill Use when the user wants to understand how behavior changes over the **lifecycle of users**, segmented by when they entered. Triggers: - "Show retention by signup cohort…" - "Build a cohort table…" - "N-day retention for…" - "Compare cohorts by acquisition channel" - "Is retention improving?" If the user wants step-by-step conversion within a single session/flow, use `funnel-analysis` instead. ## Required inputs | Input | Why it matters | |---|---| | Cohort anchor event | Defines when a user enters the cohort (signup, first purchase, install) | | Cohort grain | Day, week, or month | | Retention event | What action defines "retained" (any session, key feature use, purchase) | | Time window per period | Daily / weekly / monthly columns | | Lookback window | How many periods to display | ## Workflow 1. **Confirm the anchor event.** A user belongs to exactly one cohort. If they could be re-anchored (e.g., resubscribe), pre-specify the rule (first only, latest, both as separate cohorts). 2. **Pick the cohort grain.** - Daily: short time-scale features, fast-moving products - Weekly: most product analytics defaults — smooths weekday/weekend variance - Monthly: long sales/subscription cycles 3. **Define "retained" precisely.** Most common definitions: - **Any-activity**: at least one logged event in period - **Key-action**: specific event (e.g., completed_order) - **N-day**: active on day N exactly (rolling vs ab