Retention Rate: Read The Curve, Not The Day

The short version

Retention rate is the share of a cohort that came back after a given number of days. It decides whether an install was a user or a download, and almost every other mobile metric depends on it.

javascript

Day N retention = (users active on day N / users in the cohort) x 100

One convention worth confirming: day 1 retention normally means the user returned on the day after install, not on install day. Sources differ. Check which definition yours uses before comparing anything.

Why it has to be a cohort

Retention is only meaningful for a cohort — users who installed in the same period, followed forward together.

A blended figure across all users mixes cohorts of different ages and different acquisition sources, and it moves for reasons that have nothing to do with user behavior. If retention appears to change and you are looking at a blended number, the change is probably in your acquisition mix.

The shape matters more than any point

Steep day 1 drop. Expectation mismatch. Either the ad promised something the app does not do, or first-run experience loses people before they reach value. Look at the creative and the onboarding, in that order.

Reasonable day 1, collapse by day 7. Users understood the app and found no reason to return. A product and habit problem, not an acquisition problem.

A curve that flattens. The flattening point is the durable core of the cohort. A curve that flattens at a meaningful level is healthier than a higher day 1 that keeps falling.

Retention differing sharply by channel. Useful and expected. This is the strongest argument for paying different acquisition costs by channel.

What retention does not tell you

Whether returning users are paying. High retention with no monetization is real and common.
Why users left. The curve shows the shape, not the cause.
Anything comparable across app categories. Norms differ enough that cross-category comparison is not informative.

Reading the patterns

Retention falling as spend scales. Usually mix, not decline. Broader targeting reaches less qualified users. Judge whether the trade is acceptable rather than treating it as a defect.

Retention falling for one creative only. That creative is over-promising. It shows up as strong IPM with weak retention.

Retention stable, revenue falling. Users stayed and stopped paying. Look at monetization, not acquisition.

Where to see it in Playad

When an MMP is connected, Dashboards carries cohort retention matrices and retention curves, sourced from the MMP rather than from ad platforms. Cohort LTV and ROAS curves sit alongside them and depend on the same base. Without an MMP connected, these views have no data source.

Common mistakes

Reading blended retention instead of cohort retention.
Comparing day 1 figures without confirming both sources define day 1 the same way.
Comparing against another app category's norms.
Treating a retention drop during scaling as a product regression without checking mix.
Optimizing acquisition toward install volume and expecting retention to hold.

Practical rule

Read retention by cohort and by channel, and read the shape rather than a single day. A creative or channel is only worth scaling once its cohort retains, regardless of how good its acquisition cost looks.