Rolling Retention

Rolling retention is the share of a cohort active on day N or any day after it — equivalently, those whose last activity reaches at least day N.

This variant answers the same retention question more softly: instead of 'did they return exactly today?' it asks 'have we not lost them yet?'. Because it accepts activity on day N or later, it is always greater than or equal to classic retention for that day and gives a more realistic picture for low-frequency products. The denominator is again the acquisition-cohort size; the numerator is anyone whose last visit fell on day N or after. Its trap is that, since it peers into the future, it is incomplete for days near today — a user returning tomorrow is not yet counted, so the recent tail of the curve dips artificially and must not be read as churn. So read rolling retention for days whose full window has passed, not for yesterday.