Skip to content

Mobile game retention benchmarks (2026)

The median mobile game keeps about 22% of new players to day 1, just under 4% to day 7 and under 1% to day 30, according to GameAnalytics' 2026 report on 16,262 games. Puzzle games with paid-for players do much better in AppsFlyer's Q3 2026 data: around 29% to 34% on day 1 and 11% to 14% on day 7 in the United States. The two sets of numbers measure different games and different players, so pick the benchmark that matches how your own tool counts.

By Futigo · Updated · 10 min read

Key points

  • 22%, just under 4% and 0.68% to 0.79% are the median day 1, day 7 and day 30 retention across all genres in GameAnalytics' data for 2025.
  • The top 10% of games in the same data hold about 40% on day 1 and 11% to 12% on day 7, so the top and the middle are far apart.
  • Puzzle games on Android keep 32.55% on day 1 and 11.17% on day 7 in India, against 29.20% and 11.12% in the United States, in AppsFlyer's Q3 2026 paid-user data.
  • A median session lasts 3.1 to 3.5 minutes in GameAnalytics' data, while Adjust reports an average of about 30 minutes: the tools define a session differently.
  • On Android 13 and higher, notifications are off by default for a newly installed app until the player grants permission.

Retention benchmarks by genre

Three publishers put current figures in the open. GameAnalytics reports percentiles across all genres. AppsFlyer reports by genre, country and platform, for players acquired through paid media. Adjust reports day 1 by genre.

Day 1, day 7 and day 30 retention for mobile games, by genre and publisher
Genre and segmentDay 1Day 7Day 30Publisher, period and sample
All genres, global, medianAbout 22%Just under 4%0.68% to 0.79%GameAnalytics 2026 report, calendar year 2025, 16,262 games with at least 1,000 monthly active users, iOS and Android
All genres, global, top 25%Just above 30%6% to 7%1.6% to 1.8%GameAnalytics, same sample
All genres, global, top 10%About 40%11% to 12%2.5% to 5.1% depending on regionGameAnalytics, same sample
All genres, Asia, median17.50%2.68%0.53%GameAnalytics, 2025, yearly average of weekly data
All genres, North America, median23.28%4.97%1.18%GameAnalytics, 2025, yearly average of weekly data
Puzzle, India, Android32.55%11.17%Not publishedAppsFlyer, Q3 2026, paid media cohort, from a dataset of 100,000+ apps. Mean or median not stated.
Puzzle, United States29.20% Android, 34.08% iOS11.12% Android, 14.06% iOSNot publishedAppsFlyer, Q3 2026, paid media cohort
Match, United States29.90% Android, 33.34% iOS11.62% Android, 13.96% iOSNot publishedAppsFlyer, Q3 2026, paid media cohort
Match, India, Android27.15%9.17%Not publishedAppsFlyer, Q3 2026, paid media cohort
Hypercasual, United States22.67% Android, 28.22% iOS5.02% Android, 6.51% iOSNot publishedAppsFlyer, Q3 2026, paid media cohort
Hypercasual, India27.83% Android, 34.89% iOS4.59% Android, 7.87% iOSNot publishedAppsFlyer, Q3 2026, paid media cohort
Simulation, United States23.19% Android, 31.95% iOS5.91% Android, 9.87% iOSNot publishedAppsFlyer, Q3 2026, paid media cohort
All games, global27%Not in the reportNot in the reportAdjust 2026 report, calendar year 2025, its top 5,000 apps mixed with all apps it tracks
Hyper casual and hybrid casual, global27% eachNot in the reportNot in the reportAdjust, 2025
Puzzle, word and casual, globalAbout 20%, 21% and 17%Not in the reportNot in the reportAdjust, 2025, read from the report's bar chart

Read from each report on 15 November 2026. GameAnalytics global figures are ranges across 2025 as described in the report text, and its regional figures are yearly averages. AppsFlyer's page does not say whether its figures are means or medians, or how many apps sit behind each cell. Adjust's puzzle, word and casual values were read from a chart and are approximate.

What good means, and why the top and the median differ

There is no single good number. GameAnalytics counts every live game above 1,000 monthly active users, including many small ones, so its median is low. AppsFlyer counts players that studios paid to acquire, which skews towards games with marketing budgets. Adjust's dataset leans on its 5,000 largest apps. In GameAnalytics' data the top 1% of games hold 64% to 68% on day 1 and 13% to 15% on day 30, roughly twenty times the median at day 30.

The sources also disagree with each other. Adjust shows hyper casual with the highest day 1 retention of any genre and puzzle at about 20%. AppsFlyer shows puzzle ahead of hypercasual on Android in every country it lists. Adjust's own report gives 27% for all games and 19% to 21% for each region, and does not explain the gap. We show all of them and do not pick a winner.

How retention is measured

Classic retention (also called N-day or strict) counts players who came back on exactly day N. Rolling retention (also called unbounded) counts players who came back on day N or any later day, so it is always equal or higher. Amplitude's documentation describes both, and GameAnalytics states that it uses the strict form.

How analytics tools define retention, days and sessions, from their own documentation
ToolRetentionHow a day is countedSession
Google Analytics for FirebasePercentage of new users who return each day in their first 42 days, with day 1, day 7 and day 30 lines per acquisition-day cohortReports follow the property's reporting time zoneStarts when the app opens in the foreground with no active session. Times out after 30 minutes of inactivity by default, and the timeout can be changed.
Google Play Console (statistics)2, 3, 7, 14 and 28-day retention, for users or devices: opened the app on the Nth day after first openInstallation data uses Pacific TimeNo session metric. Reports average app usage time per daily active user.
Play Games Services (statistics)Share of players still playing 7 or 28 days after installing, counted on the exact dayNot stated on the pageNot stated. Engagement data covers players signed in with a Google Account.
GameAnalyticsStrict: installed on a given day and returned on exactly day N. Activity on other days is ignored.UTC datesFrom a session start event to a session end event. Playtime per session is the mean length per session end event.
AppsFlyer (benchmarks)Percentage of users who opened the app 1, 7 or 30 days after installing. Classic or rolling is not stated.Not stated on the definitions pageThe total number of times the app is opened
TenjinUsers retained on day N divided by the day 0 install cohortUTC by default: day 1 starts at the first UTC midnight after install. The setting can be changed.Not covered on the page we opened
devtodevOffers both methods24-hour intervals from install by default, calendar days as an optionNot covered on the page we opened
AdjustNot confirmedNot confirmedNot confirmed

Read from each provider's help or documentation pages on 15 November 2026. Adjust's glossary pages returned errors on that date and AppsFlyer's retention dashboard help page refused the request, so those cells are limited to what we could open.

Why two tools disagree about the same game

  • Calendar days against 24-hour windows. devtodev explains that with calendar days, day 1 can be anything from 0 to 48 hours after install. With 24-hour windows it is 24 to 48 hours. It says the hourly figure is always the lower one, by 4 to 5 percentage points in some projects.
  • Time zones. A player in India who installs at 9pm and plays again at 1am has crossed midnight locally but not in UTC. One tool counts a day 1 return and another does not.
  • What starts the clock. Play Console counts from first open. GameAnalytics counts from the first session it records. An attribution tool counts from the install it attributed.
  • Who is counted. Play Games Services statistics cover signed-in players. AppsFlyer's benchmarks cover paid installs. Play Console reports users and devices separately.
  • Estimation. Google Analytics says its session and user metrics are calculated through an estimation.

Session length and sessions per day

Session length, sessions per day and daily playtime for mobile games, by publisher
SegmentSession lengthSessions per player per dayPublisher and period
All genres, global, median3.1 to 3.5 minutes3.8 to 3.9 (about 12 minutes of play a day)GameAnalytics, 2025, 16,262 games
All genres, global, top 25%About 5.2 minutes5.3 to 5.7 (22 to 24 minutes a day)GameAnalytics, 2025
All genres, global, top 10%Just over 8 minutesUp to 9.6 (40 to 42 minutes a day)GameAnalytics, 2025
All genres, Asia, median3.12 minutes4.19 (11.79 minutes a day)GameAnalytics, 2025, yearly average
All genres, North America, median3.64 minutes4.23 (14.45 minutes a day)GameAnalytics, 2025, yearly average
All games, global, averageAbout 30 minutes1.65 on the day of installAdjust, 2025
Casual and puzzle, global, averageAbout 26 minutes eachNot broken outAdjust, 2025, read from the report's bar chart
Hybrid casual, hyper casual and word, global, averageAbout 23, 22 and 23 minutesNot broken outAdjust, 2025, read from the report's bar chart

Read on 15 November 2026. GameAnalytics does not split these metrics by genre in its 2026 report. Adjust's genre values were read from a chart and are approximate. Its report text gives exact figures only for strategy (37.51 minutes), racing (16.33) and action (43.8).

A three-minute median and a thirty-minute average cannot describe the same thing. A tool that ends a session when the game leaves the screen records many short sessions. A tool that keeps a session open until 30 minutes of inactivity, as Google Analytics does by default, merges several visits into one long session. Compare your game only with a benchmark from the tool you use.

The first session and install-to-play

GameAnalytics' report puts the loss early: with a median day 1 retention of about 22%, roughly 78% of new players do not come back the next day. Its commentators argue the decision is made in the first 5 to 15 minutes. That is opinion from practitioners, not a measured figure. We found no published 2025 or 2026 benchmark for tutorial completion, first-level failure rate or the effect of the first advert, so we give none.

  • Install-to-play. We found no published benchmark for the share of installs that are ever opened. Play Console reports first opens next to install events, so a developer can work out the ratio for their own game.
  • Day of install. Adjust counts 1.65 sessions per user on the day of install across all games in 2025. Most new players open a game once or twice on the first day.
  • Ad timing. AdMob's placement rules, covered in how casual mobile games make money, keep interstitials out of app load and out of play.
  • Permission prompts. Android's guidance is to let new users get familiar with the app before asking for any permission.

Daily puzzles, streaks and notifications

A daily puzzle or a streak gives a player a reason to return on a particular day, and a notification is the usual reminder. We found no published study that isolates how much a daily puzzle or a streak changes retention, so treat claims about either with care. The rules on reminders are clear.

  • Permission is required. Android 13 (API level 33) and higher has a runtime permission, POST_NOTIFICATIONS. On those devices a newly installed app's notifications are off by default until the user grants it.
  • Ask late and in context. Android suggests asking after a user action, or waiting until the third or fourth launch. If the user selects "Don't allow", the app cannot send notifications.
  • Users can see and revoke. Android notes that users can see how many notifications an app sends each day and can revoke the permission at any time.
  • No adverts in notifications. Google Play's policy on imitating system functionality says system notifications may only be used for an app's integral features. It lists ads delivered through a notification as a violation, and says not to use notifications to promote a different app.
  • Many players decline. In Airship's benchmarks for 2025, the median Android opt-in rate across all app types was 53.3%, and 48.85% on iOS. For gambling and gaming apps, the lowest tenth on Android sat at 21.35%.

What a small game can learn from its own numbers

Small samples are noisy. With 200 installs and a measured day 1 retention of 25%, the 95% margin of error is about 6 percentage points either way, so the true figure could sit anywhere from 19% to 31%. With 1,000 installs it narrows to under 3 points. At day 30 the problem is worse: 1% of 200 installs is two players. These are our own calculations from the standard formula for a proportion.

  • The benchmarks may not include you. GameAnalytics excludes games under 1,000 monthly active users. AppsFlyer's figures cover paid installs, which a game with no advertising budget does not have.
  • Compare against yourself. The same tool, the same definition and a long enough period show whether a change helped. A single week does not.
  • Read day 1 and day 7 first. They fill up with players sooner than day 30.
  • Play Console may show no peers. Its peer comparison appears only when enough apps match the chosen peer group and metric.

What we do at Futigo

We make calm casual games. Several have no timers and no lives, and some offer a daily puzzle, for example Looplet, which has a new kolam puzzle every day.

What the benchmarks do not tell us

  • Day 30 by genre. AppsFlyer's Q3 2026 page stops at day 7 and GameAnalytics does not split by genre. We found no current, primary day 30 figure for puzzle games.
  • Organic players. AppsFlyer measures paid cohorts only. None of the three reports separates players who found a game by themselves.
  • Means or medians. Only GameAnalytics states its percentiles. AppsFlyer and Adjust do not say how their figures are averaged.
  • Very small games and premium games. Each sample is one vendor's customers, and none reports paid, ad-free games separately.
  • Cause. A benchmark shows where games sit. It does not show what moved them there.
  • India on iOS. AppsFlyer lists no iOS figure for puzzle or match games in India.

Questions

What is a good day 1 retention rate for a mobile game in 2026?
The median across all genres is about 22% in GameAnalytics' data for 2025, the top 25% of games hold just above 30% and the top 10% about 40%. Puzzle games with paid-acquired players hold 29% to 34% in the United States in AppsFlyer's Q3 2026 data.
What is the average day 7 and day 30 retention for mobile games?
GameAnalytics' 2026 report puts the median at just under 4% on day 7 and 0.68% to 0.79% on day 30, across 16,262 games. The top 10% of games reach 11% to 12% on day 7.
What is the retention rate for puzzle games in India?
On Android, 32.55% on day 1 and 11.17% on day 7, according to AppsFlyer's Q3 2026 benchmarks for players acquired through paid media. AppsFlyer publishes no iOS or day 30 figure for puzzle games in India.
What is the difference between classic and rolling retention?
Classic retention counts players who return on exactly day N. Rolling, or unbounded, retention counts players who return on day N or any later day, so it is always equal or higher. GameAnalytics uses the classic, strict form.
How long is an average mobile game session?
It depends on the tool. GameAnalytics reports a median of 3.1 to 3.5 minutes per session in 2025, while Adjust reports an average of about 30 minutes across all games. Google Analytics keeps a session open until 30 minutes of inactivity by default.
Do Android games need permission to send notifications?
Yes. On Android 13 (API level 33) and higher, a newly installed app's notifications are off by default until the user grants the POST_NOTIFICATIONS permission. Google Play policy also bars using notifications to deliver adverts.