Skip to main content

Retention

Retention analysis answers: of the users who did something, how many came back and did it (or anything) again later?

What a retention analysis is​

A retention analysis is defined by:

  • Cohort event — the event that puts a user into a cohort (e.g. signed_up, or $app_installed for install cohorts)
  • Return event — the event that counts as "coming back" (a specific event, or any event)
  • Cohort grain — group cohorts by day, week, or month
  • Periods — which offsets to measure (e.g. day 1, 7, 14, 30)

The result is a cohort table: each row is a cohort (users who first did the cohort event in that day/week/month), each column is the percentage still active N periods later.

Cohort        Size   D0     D1    D7    D14
Jan 1 week     234   100%   45%   32%   28%
Jan 8 week     289   100%   52%   38%   31%

Reading it: compare columns down a column across cohorts — if D7 retention improves for newer cohorts, whatever you changed is working.

How to instrument for retention​

  1. Have a clear cohort event. signed_up or $app_installed (tracked automatically by the SDKs) are the usual choices. If you track nothing else, lifecycle events alone give you install-cohort retention out of the box.
  2. Use stable user IDs. Retention joins the cohort event and return events on user ID across days or weeks. Call identify() with the same ID on every platform and login — a user who appears under a new ID looks like a churned user plus a new user.
  3. Pick a return event that means value. "Any event" measures app opens; a specific event like document_edited measures whether users return to the thing that matters.

Availability​

Retention analysis is available on paid plans.

Next steps​

  • Identifying Users — stable IDs across sessions and platforms
  • Funnels — how users get to activation in the first place