Contents
- A cohort groups the users who arrived over the same period, tracked over time.
- The reading is by columns: comparing week 1 of each cohort says whether acquisition is improving or degrading.
- The drop is always sharp between week 0 and week 1: it's the plateau that follows that counts.
Definition
Cohort retention measures the proportion of users from one arrival group who come back in the following periods.
Formula
Week-N retention = cohort users active in week N, divided by the cohort's initial size.
Worked example
Three weekly cohorts retain 14%, 13% and 22% of their users in week 1 respectively. The third corresponds to a campaign launched on a new audience: this audience comes back markedly more, which justifies reinvesting even though its acquisition cost is higher.
Orders of magnitude
Retention drops sharply between week 0 and week 1 on almost every site. What sets good acquisition apart is the level at which the curve stabilizes afterwards.
The most common mistakes
- Reading rows rather than columns. The useful comparison is between cohorts at the same age, not within the life of a single cohort.
- Judging a too-young cohort. You need at least three or four periods to see a plateau.
- Mixing acquisition sources. An all-channels cohort averages incomparable behaviors.
Where to find the metric
In GA4, a Cohort exploration, to build manually. It lets you segment by source, which is the most useful use.
Frequently asked questions
The Adwise team connects Google Ads, Meta Ads, LinkedIn Ads, GA4 and Search Console so agencies and freelancers can run their campaigns from a single place.