Written by Krisztián Stark
Key takeaways
- Retention rate is the share of a group of new users who are active again a set number of days after they started.
- Classic retention counts users active on exactly day N; rolling retention counts anyone active on day N or later, and is never lower.
- Average retention across cohorts by adding up users, not by averaging percentages, and always split it by acquisition source.
What is retention rate in apps?
Retention rate is the percentage of users who come back to an app a given number of days after they first used it. If 1,000 people install an app on one day and 250 of them open it again the next day, the day 1 (D1) retention rate is 25%. D7 and D30 retention measure the same thing one week and one month later.
Retention is the clearest sign of whether people find lasting value in a product. Acquisition fills the top of the funnel, but every later number, from daily active users to revenue per user, depends on how many people stay.
The retention rate formula
Retention is always measured for a cohort: a group of users who started on the same day, or in the same week.
Day N retention = users from the cohort active on day N / users in the cohort x 100The install or sign-up day is day 0. Day 1 is the next day, day 7 is a week later, and so on. Analytics tools differ on whether a “day” is a calendar day in a time zone or a 24-hour period from the first open, so do not compare figures from two tools without checking.
Classic, rolling, and range retention
| Type | Counts a user as retained on day N if they were active | Best for |
|---|---|---|
| Classic (N-day) | On exactly day N | Daily-use apps and games, and comparing with published benchmarks |
| Rolling (unbounded) | On day N or any day after it | Products used now and then, where a user may skip day 7 but return on day 9 |
| Range (bracket) | At least once inside a range, such as days 7 to 13 | Weekly-use products, where one exact day is too noisy |
Rolling retention is never lower than classic retention for the same cohort, because it accepts more days. When a partner quotes a D30 number, ask which type it is.
Reading a cohort retention table
Most tools show retention as a table: one row per cohort, one column per day. Read down a column to see whether newer users stick better than older ones, and along a row to see how fast one cohort fades.
| Install week | Users | D1 | D7 | D30 |
|---|---|---|---|---|
| Week 1 | 2,000 | 26% | 10% | 4% |
| Week 2 | 500 | 34% | 16% | 7% |
| Week 3 | 1,500 | 30% | 13% | 5% |
Illustrative math: averaging three cohorts correctly
The simple average of the three D7 figures above is (10% + 16% + 13%) ÷ 3 = 13%. That is wrong, because the cohorts are different sizes. Add up the users instead: 200 + 80 + 195 = 475 users retained on day 7, out of 4,000 installs, which is 11.9%. The small week 2 cohort made the simple average look better than reality. The cohorts are invented; the weighting rule applies everywhere.
What D1, D7, and D30 retention tell you
- D1 mostly reflects the first session: onboarding, loading time, and whether the app does what the ad or store page promised.
- D7 shows whether users found a reason to make the app part of their week.
- D30 is close to your long-term audience, and it is the one that drives lifetime value.
For what these numbers usually look like in different categories, with published figures from Adjust and GameAnalytics, read App retention: day 1, 7 and 30 benchmarks. This page sticks to the definitions.
Retention rate vs churn rate
Churn is the other side of retention: the share of users who stop using the product. For a single cohort and day, churn = 100% minus retention, so D30 retention of 6% means 94% of that cohort had churned by day 30, at least for that day. For subscription businesses, churn is usually measured per month among paying customers instead, which is a different calculation, so check which one a report uses.
A related measure is a stickiness ratio such as DAU/MAU, which shows how often retained users return. Retention tells you whether people stay; stickiness tells you how often they show up.
Common retention mistakes
- Mixing classic and rolling figures when comparing channels or benchmarks.
- Averaging percentages across cohorts of different sizes, as in the example above.
- Blending all sources together. Paid, organic, and rewarded users behave differently, so compare cost per retained user by source, as CAC vs CPA vs LTV explains.
- Counting reinstalls or new accounts made by the same person as new users, which inflates cohorts and lowers retention.
- Judging D30 too early. A cohort needs 30 full days before its D30 figure is final.
Paying for users who come back, not just installs
On Sharklio, an advertiser can build a multi-step offer campaign with up to 15 paid steps, each confirmed by a postback from your own server. A later step can be a retention goal, such as reaching a level that takes several days of play, so part of the price is only paid for users who actually stay. You set the price per step per country, and every campaign is reviewed by staff before it goes live. See task vs offer campaigns or Sharklio for advertisers.
Related terms: DAU/MAU ratio and ARPDAU.
Frequently asked questions
How do you calculate D1 retention?
Take the users who installed on one day, count how many were active again the next day, and divide. 300 returning users out of 1,000 installs is 30% D1 retention.
What is the difference between classic and rolling retention?
Classic retention counts users active on exactly day N. Rolling retention counts users active on day N or any later day, so it is never lower.
What is a good retention rate for an app?
It depends on the category and the countries. Published benchmarks from analytics firms are the best reference; our retention benchmarks guide collects them.
Is retention rate the same as churn rate?
No, they are opposites. For one cohort on one day, churn is 100% minus the retention rate.
