Retention
How to Read Retention for a Utility App
Updated 2026-09-144 min read
Short answer
Utility apps are used when a task comes up, not every day, so D1 and D7 — activity on one exact day — look low even when the app is doing its job. Judge a utility by whether people come back at all within a window that matches the need, whether they complete the core task again, whether they keep the app installed, and how each monthly cohort compares with the previous one.
A photo compressor gets 2,000 new users a week. D1 retention is 8%, D7 is 5%, and every benchmark article says that is a failing app. But nobody compresses photos every day. For a utility the question is not whether people open it tomorrow — it is whether they come back to it the next time the need arises, instead of searching for another app.
Why exact-day retention misleads here
D7 counts people who were active on day seven. Someone who compresses photos on day 0, day 11 and day 26 is a loyal user who never appears in D1 or D7. For a habit app, those exact days are a fair test. For a tool used on demand, they measure the calendar rather than loyalty.
Cohort size 2,000 Active on day 1 160 8% Active on day 7 100 5% Active on any day from 1 to 30 540 27% Completed the core task again within 30 days 380 19%
Same people, same month. By exact-day retention the app is failing; by "came back within 30 days" more than a quarter of the cohort returned, and nearly one in five came back to do the job again.
Measures that fit a utility
| Measure | Where to get it | What it tells you |
|---|---|---|
| Came back within 30 days | BigQuery export, or approximated from weekly cohorts in a GA4 cohort exploration (weeks 1 to 4) | Whether people return at all when the need comes back |
| Repeated core task | A cohort exploration with the core event, such as export_complete, as the return criterion | Whether they return to use the app for what it is for |
| Time to second use | Days between the first and second core task, from BigQuery | How often the need recurs — which sets the window for everything else |
| Uninstalls | App Store Connect deletions (opted-in users) or Play Console uninstalls | Whether people keep the tool for next time |
| Cohort trend | The same measure for consecutive monthly cohorts | Whether a release made the tool more or less worth keeping |
Choose the window from the need, not from a benchmark
Start with time to second use. If most repeat users come back within two weeks, a 30-day window catches nearly all returns and a 7-day window cuts many of them off. A scanner used for monthly receipts needs a longer window than a tool for everyday screenshots. Once chosen, keep the window fixed so cohorts stay comparable.
What goes wrong if you use daily retention anyway
- Onboarding gets rebuilt to "fix" a D1 that reflects how the tool is used, not how the first session went
- Push notifications meant to raise D7 reach people with no task that day, and uninstalls rise
- Comparing D7 with social apps or games makes a healthy utility look broken
- Real regressions — a release where the second task fails — hide inside numbers everyone already considers bad
What a real problem looks like in a utility
| Pattern | Usually means |
|---|---|
| 30-day return falling across consecutive cohorts | The tool became less worth coming back to — check releases, ads and paywall changes in that period |
| Returns steady, repeated core task falling | People come back but do not finish — check errors and the completion rate of the core task |
| Uninstalls rising soon after first use | The first task did not work as promised, or ads and prompts outweighed the result |
| New users up, 30-day return down | A new traffic source brings people with a one-off need |
How this looks in Appnaly
Appnaly does not judge retention against industry cut-offs. It compares each app with its own recent cohorts, reads a day-N figure only from cohorts old enough to have reached that day, and offers no judgement below 30 users. For a utility, the most useful pair to watch is D30 against the previous weeks and the completion rate of the core task you mark — both move when the tool stops being worth keeping.