Three Ways Cohort Retention Charts Lie to You

Retention curves are one of the most-trusted charts in growth analytics, and also one of the easiest to misread. A few things worth checking before you act on one:

1. Cohort size drift

If early cohorts are tiny and later ones are huge, your "blended" retention curve is really just describing your biggest cohort. Always check n per cohort, not just the percentage.

2. Definition creep

"Active" quietly changing from logged in to completed a core action somewhere between two dashboards will make retention look like it improved when nothing changed except the metric.

3. Survivorship in the denominator

# wrong: denominator shrinks as churned users are excluded upstream
retention = returning_users / active_users_today

# right: denominator is the original cohort size, fixed at cohort start
retention = returning_users / cohort_size_at_day_0

None of this means retention curves are useless — just that they reward a little skepticism before you put them in a slide.