What is a cohort retention chart?
A cohort is the group of people whose first entry event (by default, any page view or custom event) fell in the same week (or day). Retention is the share of each cohort that is active again in later periods. A week starts on Monday and ends on Sunday, in UTC, as in the date picker. The date of a weekly cohort is the Monday of its week.
In this example, 40% of the people who started in the week of Sep 1 came back in the next week.
Create a retention chart
Retention is a chart type on a dashboard.- Open Dashboards and open or create a dashboard.
- Click Add chart and choose the Retention metric.
- Choose the Entry event. It defines the cohort: people are grouped by the period of their first entry event. Choose Any event (default), Any page view, a Page path such as
/play, or a custom Event such astutorial_completed. - Choose the Returning event. A user is retained in a period when they do this event in it. The choices are the same as for the entry event.
- Choose the Retention type:
- Rolling (default for new charts): a user counts in period N when they return in period N or in any later period. The curve can only go down.
- Recurring: a user counts in period N only when they return in exactly that period. The curve can go down and come back up.
- Choose the Period:
- Weekly (default): weeks 0 to 12.
- Daily: days 0 to 30, which includes D1, D7 and D30.
- Optional: choose a saved segment in Users to count only those people.
- Click Save chart.
How retention is counted
- A user is the explicit
user_id, else the one user that the visitor’s anonymous ID links to, else the anonymous ID. So a user who signs in on a later visit stays in the same cohort. See core concepts. - A user enters a cohort with their first entry event. With Any event, this is their first observed page view or custom event.
identifyevents do not count as activity. - A user is retained in a period when they do the returning event in it (recurring), or in it or a later period (rolling).
- Period 0 is always 100%: every user in the cohort entered in period 0, even when the returning event is different from the entry event.
- Dashboard filters and the segment keep only people with a matching event in the date range.
- Periods that are not complete at the end of the date range show a dash, not a low number.
- Bots and events sent before consent (no visitor ID) are excluded.
Read the chart
- Compare rows to see whether newer cohorts retain better than older ones after a product change.
- Read columns to see where people stop coming back. A large drop between Week 0 and Week 1 often points to onboarding.
- Look for a flat tail. When the curve flattens, you have found the people who stay.