Written by Ar.Bhavesh Panse, AI app growth marketer | ZuAI: 10K → 2M users at $0.02 CAC | $300k/mo ad spend managed
AppGrowth Marketer

App retention: rate formula, benchmarks and how to improve it Nothing counts unless they came back.

App retention rate explained: the day N formula, classic vs rolling, sourced benchmarks with caveats, how to read a retention curve and 10 ways to improve.

app retention rate is the share of users from one install cohort who open your app again on a given day. the classic formula is users active on day N divided by users who installed on day 0, times 100. read it by weekly cohort, watch where the curve flattens, and fix the first session before anything else.

for how retention sits next to CPI, CAC and LTV, read app growth metrics that matter. this page goes deeper on retention only.

What is app retention rate?

app retention rate measures how many people who installed your app in a given period are still using it after a set number of days. it is usually reported as day 1, day 7 and day 30. it answers one question: did the people you acquired find enough value to come back on their own?

retention is not active users. daily active users mixes new and old users, so a big install week can hide a product nobody returns to. retention isolates one cohort at a time.

App retention rate formula: classic vs rolling

classic day N retention is the number of cohort users who opened the app on exactly day N, divided by the cohort size, times 100. rolling retention counts users who opened the app on day N or any later day. rolling is always higher than classic, so never compare a rolling number with a classic benchmark.

  • classic (day N) retention: users active on day N, divided by users who installed on day 0, times 100. strict, and the most common version.
  • rolling (unbounded) retention: users active on day N or any day after, divided by the cohort size, times 100. useful for weekly apps such as budgeting.
  • bracket retention: users active at any point in a window, such as days 7 to 13, divided by the cohort size. a good middle ground for weekly habits.

pick one definition and use it for every cohort.

A worked example (illustrative numbers)

here is one cohort for an imaginary app, so the formulas are concrete. these are illustrative numbers, not from a real account and not a benchmark. 1,000 people install on day 0. the table shows how classic and rolling retention tell different stories about the same users on the same day.

measureuserscalculationresult
cohort (installs on day 0)1,000
opened on day 1280280 / 1,00028% classic day 1
opened on exactly day 79090 / 1,0009% classic day 7
opened on day 7 or any later day160160 / 1,00016% rolling day 7
opened on exactly day 304040 / 1,0004% classic day 30

the gap between 9% and 16% is people who use the app, just not every day. if your product is weekly by nature, classic day 7 undersells it. if it is meant to be daily, the rolling number flatters it.

App retention rate benchmarks (and their caveats)

published benchmarks are a sanity check, not a target. each report uses its own sample, platform mix and definition, and subscription reports measure paying subscribers, not all users. i do not quote a single average for day 1, 7 and 30 here, because the figures repeated online often trace back to older reports with different methods.

here is what recent, dated reports actually say:

source and datewhat it measuresfigure
RevenueCat State of Subscription Apps 2026 (2025 data, 115,000+ apps)share of free trials started on install day80 to 89% in most categories
same reportmedian day 35 download-to-paid10.7% hard paywall vs 2.1% freemium
RevenueCat State of Subscription Apps 2025 (published March 2025)annual subscriptions cancelled in the first monthnearly 30%
same reportpaid subscriptions retained after year 144.1% yearly, 17.0% monthly, 3.4% weekly plans
RevenueCat AI app retention study (August 2026)AI app paid subscriptions still active after one year21.1% annual, 6.1% monthly plans
AppsFlyer app uninstall report, 2025 edition (2024 data, Android only)installs uninstalled within 30 days46.1%

caveats: these measure different things (subscriber retention, trials and uninstalls are not day N user retention), and your category and platform mix matter more than any global number. the install day trial figure is the real lesson: most of the decision happens in the first session.

How to read a retention curve

a retention curve plots the share of a cohort still active on each day after install. every app’s curve drops steeply at first. what matters is whether it flattens. a curve that levels off at a stable share means a group of people found lasting value. a curve that slides toward zero means you have not found that group yet.

read your curves like this:

  1. steep drop from day 0 to day 1: the first session failed. look at onboarding, the ad promise and time to first value.
  2. slow slide with no floor: people liked the first try but have no reason to return. look at the habit, not the funnel.
  3. flattening at a stable level: that is product fit for that cohort. now find more people like them and protect what they love.
  4. newer cohorts flatten higher than older ones: your changes are working. this is the chart to show investors.

plot cohorts by week and by source. a curve that flattens for organic users but not for one campaign means that campaign pulls the wrong audience.

10 ways to improve app retention

the best retention gains come in order: first session, then first week, then long term. fix the moment people decide whether your app is worth a second open before you invest in reminders or loyalty features. each idea below is grouped by the stage where it does its work.

first session

  1. match the first screen to the promise. if the ad or store page promised one outcome, the first screen should move toward exactly that outcome.
  2. deliver value before signup. let people try the core action first, then ask for an account to save the result. my full checklist is in app onboarding best practices.
  3. cut every step that does not change the first result. each extra question is another place to drop off.
  4. ask for permissions in context. Apple and Google both advise asking when a feature needs it, not at launch. on Android 13 and later, new installs start with notifications off.

first week

  1. give people a reason to return tomorrow. a plan, a reminder they set themselves, a result that updates overnight.
  2. send only notifications people asked for. on iOS, provisional authorization delivers notifications quietly so people can judge them before allowing alerts.
  3. follow up on day 1 to 3 drop offs with one next step. an email or message that points to the single action that made returning users stick.
  4. find the step where each cohort leaves. compare the events of users who came back with those who did not, and fix the biggest gap first.

long term

  1. make the app better with use. saved history, personal progress and preferences that sharpen recommendations give people something to lose by leaving.
  2. plan for the first month of subscribers. RevenueCat’s 2025 report found nearly 30% of annual subscriptions are cancelled in the first month, so show new value during those weeks instead of reaching for discounts.

The rule that keeps retention honest

on ZuAI, the AI study app that grew from 10K to 2M users in eleven months at $0.02 blended CAC, the working rule was simple: nothing counts unless they came back. an install, a signup or a trial was a hint. a user was someone who returned on their own, and every channel was judged on that.

a creative with a cheap CPI that brought people who never returned got cut, however good its dashboard looked. spend divided by users who came back is the CAC that matters, as explained in CPI vs CAC.

What not to do

do not buy retention with tricks. dark patterns and notification spam can lift a chart for a week, then cost you uninstalls, bad reviews and trust with the platforms. if a tactic only works because users cannot find the exit or cannot tell what they agreed to, it is a liability, not a retention strategy.

  • no dark patterns. no hidden cancel buttons, guilt-trip copy on the “no” option or fake countdown timers. Apple’s App Review Guidelines require you to clearly describe what a subscriber gets before asking them to subscribe.
  • no notification spam. Android lets users see how many notifications your app sends each day and revoke permission at any time. Apple’s guideline 4.5.4 says push should not be used for promotions unless people explicitly opted in.
  • no forced permissions. Apple’s guidelines say apps must not manipulate, trick or force people to consent to unnecessary data access.
  • no incentivised opens or fake engagement. paying people to open the app inflates the curve and teaches you nothing about real users.

for how retention fits the whole role, read what an AI app growth marketer does, or browse more guides in the learn hub.

Frequently asked questions

How do you calculate app retention rate?

take the users who installed in one period, your cohort, and count how many opened the app again on day N. divide that by the cohort size and multiply by 100. 1,000 installs with 250 back on day 1 is 25% day 1 retention. those numbers are illustrative. always say whether you used classic or rolling retention.

What is a good app retention rate?

it depends on category, platform and how you measure, so compare against your own earlier cohorts first. published benchmarks from RevenueCat, AppsFlyer and other analytics vendors use different definitions and samples. the most useful sign is a curve that flattens: a stable share of each cohort that keeps coming back months later.

How can I improve app retention quickly?

start with the first session. make sure the first screen delivers what the ad or store page promised, get people to one meaningful result before signup, and ask for permissions only when a feature needs them. these fixes usually move day 1 retention faster than anything you do with notifications or emails later.

free second opinion

Want this applied to your app? Send me your numbers.

tell me about your app, what you are testing and where it is stuck. i'll reply myself with what i would try next. no pitch, not a sales call.

your details go only to Ar.Bhavesh Panse, to reply to you. see privacy.

keep reading

More from the playbooks