cohort-analysislisted

# Cohort Analysis Group users by when they first joined, then track what fraction return in each period after. A cohort retention table turns a flat churn number into a curve you can read: where the cliff is, whether the curve flattens (a sticky core), and whether newer cohorts retain better than older ones. **Grounded in:** *Lean Analytics* — Croll & Yoskovitz: cohort retention curves read against a benchmark, not zero. **Go deeper (The Product Channel):** [Product Metrics](https://sidsaladi.substack.com/p/week-5-week-in-product-series-product) ## When to use this - Retention or churn looks bad but you can't tell if it's getting better or worse over time. - You shipped an onboarding, activation, or pricing change and want to see if it moved retention for users who joined after. - You need to know *where* in the lifecycle users drop — day 1, week 1, month 2 — not just that they do. - Leadership asks "are the users we're acquiring now any good?" and you need cohort-over-cohort evidence. - You want to find your retention floor (the curve flattening) to size a real, durable user base vs. a leaky bucket. ## Before you start (gather these) - **The cohort definition**: what groups users (usually signup/first-purchase date) and the grain (daily, weekly, monthly). - **The return event**: what "retained" means — opened app, completed a core action, paid again. Vanity logins are weak; a value-moment event is strong. - **The period grid**: how you'll bucket time after joining (Day 0