Retention Fundamentals

• Define Day-N retention and rolling retention precisely
• Explain why retention compounds growth
• Evaluate retention definitions for a given product

• Acquisition fills a leaking bucket; retention is the bucket's size, and fixing leaks compounds every future acquisition dollar.
• Retention predicts LTV: users who stay pay more and cost less to support.
• Retention is a leading indicator of product-market fit — flat D30 curves are the fit signal.
• A product that retains can grow without paid acquisition; one that doesn't can't buy its way out.

One big festival can draw thousands, but it's the café with regulars that's still in business next year. Retention is the regulars; acquisition is the footfall.

• Define the cohort entry: first session (or signup) date per user.
• Define 'active' with exact events, and the day the metric horizons (D1, D7, D30...).
• Day-N = share of the cohort that is active exactly on day N; users who joined after that horizon are excluded from day-N calculations.
• Rolling retention = share active on any day between N and the window end.
• Plot the curve across cohorts to watch trend, and benchmark category-level curves for context.

• A mobile game journals D1/D7/D30 per release; any new-version D1 drop triggers a release review with the engineering team.
• A productivity app reports rolling retention at 30 days to investors, since exact-day retention understates habitual usage.
• A fitness app segments its retention curve by workout entry mode and finds habit studies at the same time of day — the habit signal.

• Pick a definition (Day-N vs rolling) and state it next to every number; mixed definitions are the top cause of retention disputes.
• Anchor windows to near-cosmic events (signup) not calendar dates, so cohorts compare fairly.
• Report Standardese curves on both new cohorts and all-active for full context.
• Pair retention rates with cohort size and purchase-rate to avoid mistaking user thinning for product decline.

• Counting late-arriving users in D7 (users who only signed up 3 days ago) and inflating the number.
• Defining 'active' with passive events so retention is a wall of constant 99%.
• Reading D1 as a forward promise — D1 tells you first-impression quality, D7/D30 tell you habit.
• Comparing retention across products without matching definitions, platforms or acquisition mixes.

• A sudden D1 jump usually means the 'active' definition changed or a SDK fired on first open without permission — diff the events first.
• Retention curves that never decay indicate the active definition includes app-in-background events.
• Check dedupe (user_id vs device_id) — mixed identities fragment cohorts and deflate retention.
• Compare retention on organic vs paid vintages; a paid-campaign cohort with poor retention is an acquisition-quality signal, not a product one.