If you search for mobile game retention benchmarks by genre in 2026, you will find the same table on a dozen sites, with the same seven genres and the same two-decimal precision. Nearly all of it traces back to a single AppsFlyer dataset from the third quarter of 2022.
The version of this page I published in March carried that table. I have rewritten it, because the numbers in it were not what the source said, and because presenting four-year-old data under a 2026 heading is the specific failure this blog got caught doing.
Here is what the measurement actually supports, what it does not, and the ten things I still change first when a studio brings me a retention problem.
Key Takeaways
- The measured median is lower than the folklore. GameAnalytics put D1 at 22% across 11,600 games in 2025. Top quartile is 25-27% on Android and 31-33% on iOS. A 40% D1 target matches nothing in the published data
- Genre-level retention has not been broadly republished since 2022. Every 2026 genre table you find is an anchor from four years ago wearing a current date
- Retention is a multiplier on acquisition spend. At $1.50 to $3.50 for a US casual Android install, moving D30 from 3.5% to 5% takes the cost of a surviving player from $43-$100 down to $30-$70
- Paid user acquisition needs a D7 far above the top quartile. Tier 1 economics start working around a D7 of 18%, against a measured top quartile of 7-8%. Most games are structurally organic-first and do not know it
- Push is now a permission on both platforms. Android 13 killed the 95% Android opt-in figure in 2022. Reach is earned from the player, and the grant rate is a metric you should be watching
- Three stages, three different problems. Day 1-3 is first-session clarity, day 3-30 is habit, day 30+ is content depth. Blending them into one number is why most retention fixes fail
What the Measured Data Actually Says
Two datasets carry almost all of the defensible retention measurement available to a small studio in 2026. They disagree, and the disagreement is informative.
| Segment | D1 retention | D7 retention | D30 retention | Source and vintage |
|---|---|---|---|---|
| All genres, median | 22% | 3.4-3.9% | Not published | GameAnalytics, 2025, 11,600 games and 1.48Bn MAU |
| All genres, top quartile | 25-27% Android, 31-33% iOS | 7-8% | Not published | GameAnalytics, 2025 |
| Casual and puzzle, Android | 28-32% | 9-12% | 3.5-5% | AppsFlyer genre data, Q3 2022, with the Q3 2023 platform split |
The genre row sits above the market median, and that is not a contradiction. GameAnalytics measures games running its SDK, which includes a long tail of small titles that never see a marketing budget. The AppsFlyer genre sample is drawn from apps using its attribution product, which skews towards games with money behind them. Two honest measurements of two different populations.
What matters for planning is the shape rather than the decimal. Roughly three quarters of installs never open the game a second day. Under one in ten are still there at day 7. At day 30 you are working with single-digit percentages of the cohort you paid for, and that is true at the top of the market as much as at the median.
The 40% D1 and 20% D7 figures that circulate as excellence targets have no measured backing. A D7 of 20% sits at roughly three times the top quartile of the entire market. Presenting it as something a good game should clear is a category error rather than ambition, and if your soft launch dashboard is being judged against those numbers, it is being judged against fiction.
Why the 2026 Genre Tables Are Really 2022 Data
This is worth understanding, because it changes how you should read every retention article including this one.
Genre-level retention breakdowns were last published broadly by AppsFlyer for Q3 2022. Since then, two things happened. Attribution granularity on iOS collapsed under ATT and SKAdNetwork, which made clean genre cohorts harder to assemble and less comparable year to year. And the measurement vendors moved their benchmark reports behind lead-capture forms, so what remains openly readable is mostly aggregator content recycling the last open dataset.
The recycling has a second-order effect. SEO aggregators cite each other, several now cite this site, and a figure can appear independently corroborated across five pages while tracing to one unverified origin. When I audited this blog in July 2026, searching for some of its own numbers returned its own sentences as the answer.
The practical rule: if a retention table does not name a primary dataset with a date and a population, it is folklore. That includes tables with two decimal places, which is usually the tell rather than the reassurance.
What Retention Is Worth in Acquisition Terms
Retention arguments get abstract fast. Converting them into what a surviving player costs ends the argument.
Take a US casual or puzzle title on Android, where installs run roughly $1.50 to $3.50 depending on targeting and season. The cost of a player still active at day 30 is simply the install cost divided by D30.
| Tier 1 Android install cost | Cost per player active at day 30, D30 of 3.5% | Same, D30 of 5% |
|---|---|---|
| $1.50 | $43 | $30 |
| $2.50 | $71 | $50 |
| $3.50 | $100 | $70 |
That is arithmetic on two ranges rather than a measurement, and it is worth exactly as much as its two inputs. The structure holds regardless of which CPI you plug in. Moving D30 from 3.5% to 5% does the same work as cutting your CPI by 30%, and unlike a CPI reduction it costs no media budget to attempt.
The corroboration from the revenue side is harsher. Liftoff measured D30 return on ad spend for casual games on Android at 15%, across 2.4 billion installs. A rewarded-led casual title recovers around a seventh of its acquisition spend in the first month. Our CPI benchmarks guide works through why that gap is structural rather than a targeting failure.
Which produces the least comfortable arithmetic in mobile, and the reason most retention content avoids putting these two numbers on the same page. Paid acquisition at Tier 1 prices starts to work somewhere around a D7 of 18% for hybrid casual. The measured top quartile of the whole market is 7-8%.
That 18% is a business-model threshold rather than a performance benchmark, which is exactly why the gap matters. Most games are structurally organic-first and never find that out, because the soft launch question they ask is how to lower CPI when the prior question is whether paid acquisition is available to this title at all.
Retention by Stage: Three Problems, Not One
Retention behaves as three distinct problems unfolding in sequence. Each has different levers, different failure modes and different kill signals. Treating them as one blended number is the root cause of most failed retention work, because the fix gets applied to the wrong stage.
| Stage | Days | Primary goal | Key levers | Kill signal |
|---|---|---|---|---|
| Hook and onboard | 0-3 | Deliver the core loop promise | First-session speed, first-win timing, tutorial framing, device performance | D1 under 22% on Android puts you below the market median |
| Habit formation | 3-30 | Earn the second and third week | Progressive rewards, push, social features, meta layer exposure | D7 at or under 4%, the market median, means the loop is not forming |
| LiveOps depth | 30+ | Convert engaged players into a paying habit | Seasonal events, battle pass, content cadence, monetisation integration | D30 under 3.5% on casual Android means depth is missing |
Stage 1: day 1 to 3
Everything here is first impression. With a median D1 of 22%, roughly three of every four installs decide against you inside 24 hours, and most of that decision happens in the first session.
- Get to playable inside 60 seconds. No account creation, no settings, no extended tutorial before the first gameplay moment
- Design the tutorial as gameplay. The player should feel competent. Contextual hints beat guided walkthroughs
- Land the first meaningful reward inside the first session, before the player has to decide whether to open the app again. The exact timing is a design decision to test, and I have seen 45 seconds and 4 minutes both win in different genres. Anyone quoting a universal number for this is guessing
- Watch session replays for sessions 1, 2 and 3. Drop-off between session 1 and session 2 is usually a content gap rather than an onboarding problem, and the two get confused constantly
Stage 2: day 3 to 30
By day 3 the player knows whether they like the core loop. The job now is a reason to come back tomorrow, and this is where most teams underinvest.
- Progressive reward ladder. Make the day 7 reward materially more valuable than days 1 to 6 combined, so skipping a day has a cost
- Push with restraint. A small number of sends per week, timed to the player’s own play window, always carrying value. Empty “come back” pings buy opt-outs
- Meta layer unlocks by day 3 to 5. The map, collection or progression system should become the reason to relaunch, taking over from the core mechanic alone
- Social hook. Guild invites, leaderboard position or cooperative objectives belong here. Players with at least one connection in the game churn noticeably slower in every title I have worked on, though I have never seen a defensible published coefficient for it
Stage 3: day 30 and beyond
Players who reach day 30 are the cohort your revenue model depends on.
- Seasonal events on a predictable cadence, every two to four weeks. Each one is a re-engagement moment for dormant players as much as content for active ones
- Battle pass timing. Introduce it once players have a reason to commit rather than at launch. Early exposure tends to hurt conversion
- Monetisation at natural decision points, such as level completion or near-miss moments. Intrusive monetisation at this stage is the fastest way to lose an otherwise healthy cohort
- Plan the first major content update before players hit the wall. A content ceiling with no visible signal that more is coming empties the day 30 cohort within a couple of weeks
Where to Spend First, by Genre
Genre-level retention numbers are not currently sourceable, but genre-level failure patterns are stable enough to plan against. The table below is judgement from 50+ launches rather than measurement. Use it to pick which hypothesis to test first.
| Genre | Where players usually leave | First place to spend |
|---|---|---|
| Match and puzzle | Day 7 to 14, at the content wall | Meta layer depth, collection systems, event frequency |
| RPG | Day 3 to 7, on core loop complexity | Onboarding simplification, early progression clarity, guild hook |
| Casino and tabletop | Day 1, sessions too short to form a habit | First-session reward timing, daily bonus stacking, social competition |
| Strategy | Day 14 to 30, at the progression plateau | Content update cadence, alliance mechanics, PvP ladder |
| Hypercasual | Day 1, by design | Monetisation speed, second mechanic injection, genre graduation |
| Hybrid casual | Day 3 to 10, on meta hook timing | Meta unlock speed, early battle pass exposure, LiveOps trigger in the first week |
The rule I apply: fix the genre-specific drop-off before investing in LiveOps infrastructure. A match game losing players at day 7 to 14 needs meta depth. A seasonal event engine will not save it.
Strategy 1: Nail the First Session
Onboarding is where most games lose half their audience, and it is the cheapest thing on this list to change. Get players into core gameplay in the first minute rather than into tutorials, settings screens or account creation.
At Gameloft, the titles that performed best on D1 were the ones that let players feel the fun before asking them to learn the systems. Show, do not tell, and let the tutorial emerge from play.
Key actions:
- Skip forced tutorials and use contextual hints
- Delay account creation until after the first win
- Front-load the most satisfying mechanic
- Instrument time-to-first-reward as a tracked metric, then test against it
I have deliberately not put a percentage on what this is worth. The widely quoted figures for onboarding lift come from single-title case studies presented as market norms. Measure it on your own cohort, with a cohort large enough to read.
Strategy 2: Build Progressive Reward Systems
Flat daily login bonuses do very little. Reward curves that build value over consecutive days create a cost to skipping, which is the mechanism you actually want.
Structure the calendar so the day 7 reward is worth more than days 1 to 6 combined. That single anchor drives more D7 movement than the rest of the ladder. Progressive rewards are also among the easiest systems to A/B test during soft launch, provided your cohorts are large enough to read the result.
Strategy 3: Personalise, Then Verify
Machine learning applied to difficulty curves, offer timing and content recommendation is standard practice at the top of the market. Segment by behaviour rather than by demographics: a player who logs in daily needs different handling from one who plays three times a week.
The caution is that personalisation systems are very good at producing metrics that look like improvement. Hold out a control group permanently, and check that lifted engagement is showing up in day 30 cohorts rather than only in day 3 ones. For where AI genuinely pays across the value chain, see our guide on AI in mobile game development.
Strategy 4: Design Social Loops That Work Solo
Guild systems, cooperative objectives and competitive leaderboards correlate with longer sessions and slower churn. The difficulty is that most social features assume a population density the game does not have yet.
The mechanics that survive contact with a small player base are the ones that let a solo player benefit from other people’s activity without needing them online: asynchronous challenges, shared community milestones, leaderboards that populate from cohorts rather than friends.
Strategy 5: Treat Push as a Permission You Earn
The 95% Android opt-in figure quoted across retention content, including the earlier version of this page, is obsolete. Google announced the change in March 2022: “Android 13 introduces a new runtime permission for sending notifications from an app: POST_NOTIFICATIONS. Apps targeting Android 13 will now need to request the notification permission from the user before posting notifications.”
Both platforms now gate push behind an explicit grant. That changes the work:
- Ask after a success moment, never on the launch screen. The prompt is a single shot on most devices
- Track grant rate as a first-class metric, split by platform and by where in the flow you ask
- Personalise on last-session state. A broadcast schedule treats every player as the same player
- Cap frequency and always carry value. Rewards, event launches, social updates. A notification that says nothing costs you the channel permanently
Strategy 6: Make Content Cadence a Retention Plan
Your content roadmap is your retention roadmap past day 30. Plan updates on a rhythm players can anticipate: new levels, seasonal events, limited-time challenges, battle pass content.
Each update should give returning players a reason to reopen and active players something to wait for. This is much easier to build if the cadence is designed during the soft launch phase rather than retrofitted after the day 30 cohort has already gone.
Strategy 7: Monetise Without Spending the Cohort
Monetisation is a retention system. When purchases feel like progression, paying players stay longer, and the economy becomes an engagement mechanism rather than a tax on it.
Avoid aggressive monetisation in the first 48 hours and let attachment build first. Rewarded video remains the format that damages retention least, because the player chooses to watch in exchange for something clear. Studios moving beyond programmatic rewarded video into structured brand work should read our in-game advertising and brand partnerships guide, and our F2P monetization model comparison breaks down how each model interacts with the retention curve.
Struggling with retention metrics? With 20+ years of gaming expertise, I help studios diagnose and fix retention problems at every stage. Book a consultation to review your game’s retention funnel.
Strategy 8: Read Regional Retention Against Regional Economics
Retention varies by region. The specific geo splits circulating in 2026 come from the same 2022 dataset as the genre tables, and I am not republishing them here because I cannot source them at a granularity I would defend.
What is currently sourceable is the cost side, and it is enough to make the decision. US casual and puzzle Android installs run roughly $1.50 to $3.50, against roughly $0.15 to $0.60 in LATAM. Rewarded eCPM falls in the same direction, and the ratio improves less than most people assume, which is the part studios miss when they see cheap installs and conclude they have found an arbitrage.
The practical consequence for retention work: a retention curve measured in a cheap market does not transfer to a Tier 1 launch, because the players are not the same and neither is the monetisation ceiling. Our soft launch guide covers which markets give a readable signal for which decision. The same asymmetry shapes cloud gaming telco partnerships and B2B2C distribution models, where subscriber engagement varies as much by market as by product.
Strategy 9: Acquisition Decides Half of Your Retention
Retention is partly a property of who you bought. AppsFlyer’s 2024 gaming report found day 30 retention peaked at 7.5% when user-generated content was used in creatives, which is a statement about acquisition creative rather than about game design. The same build retains differently depending on the creative that recruited the player.
That has a direct operational consequence. Optimise UA towards retained users rather than installs, and read D7 by creative and by channel before you touch the game. If your CPI looks good and your D7 does not, the problem is upstream of the product.
Studios without senior marketing leadership to bridge that gap often find a fractional gaming CMO can build the alignment in weeks, and our consultant versus in-house comparison helps evaluate which model fits your stage. For the full operating framework connecting UA, retention, LiveOps and monetisation, see the mobile game growth strategy guide.
Strategy 10: Measure at a Size You Can Read
Most retention arguments in soft launch are arguments about noise. Reading a D1 of around 30% to within 3 points at 95% confidence requires 896 installs. Budget roughly 1,000 purchased installs to cover the drop between install and first session, and roughly double that per arm when comparing two variants.
A cohort of 120 cannot distinguish a D1 of 24% from one of 34%. Two people can look at that dashboard and reach opposite conclusions, both defensibly. One readable cohort of 900 beats nine unreadable cells of 100.
| Metric | What it tells you | Where to worry |
|---|---|---|
| D1 retention | First-session quality | Under 22% on Android, below the measured market median |
| D7 retention | Core loop strength | At or under 4%, the market median. Clearing 7-8% puts you in the top quartile |
| D30 retention | Depth and game-market fit | Under 3.5% on casual Android |
| D7 to D1 ratio | Where the curve breaks | No published threshold exists. Track your own trend across cohorts |
| Churn by cohort | When players leave | A spike at a specific level is a difficulty problem, not a retention problem |
| Push grant rate | Whether re-engagement is available at all | Falling after a prompt-placement change |
Track these at cohort level rather than blended. A single underperforming acquisition channel can drag the blended figure below target while the organic cohort is healthy, and the fix in that case is in targeting rather than in the game. The 20 mobile game KPIs that matter covers how these connect to the rest of the system, and our free F2P audit checklist covers the full framework from soft launch through scale.
Sequencing, Not Ranking
Retention content loves ranked tables of tactics with a percentage-point lift attached to each one. I removed that table from this article, because the lifts in it did not survive being added up: applying half of them produced a D30 above the D1 of the same game.
What I can offer instead is an order of operations, with the signal that tells you the work landed.
| Stage | Intervention | Effort | Signal it worked |
|---|---|---|---|
| Day 0-1 | First-session speed and time-to-first-reward | Low | D1 moves on a readable cohort within one release |
| Day 1-30 | Progressive reward ladder anchored at day 7 | Low | Day 7 return rate separates from day 6 |
| Day 3-30 | Meta layer visible before day 5 | Medium | Session 3 and session 4 depth increases |
| Day 3-30 | Behavioural push, after the grant rate is fixed | Medium | Reactivation rate by segment, not overall opens |
| Day 3-30 | Social hooks that work at low player density | Medium | Churn gap between connected and solo players |
| Day 14+ | Soft monetisation at decision points | Low | D30 holds while ARPDAU rises |
| Day 30+ | Seasonal event cadence | High | Dormant player reactivation per event |
| Day 30+ | Battle pass, once commitment exists | High | Conversion without a D30 drop in the same cohort |
The pattern that holds across launches is that first-session work and meta-layer timing are the cheapest interventions available, and LiveOps pays most at day 30 but requires infrastructure built during development. Studios that try to add an event engine at day 60 typically need three months to ship the first event, by which point the cohort it was meant to save has gone.
If the diagnosis needs an outside read across retention, LiveOps and monetisation together, a mobile game growth consultant should connect those levers rather than treating retention as an isolated feature backlog.
Common Retention Mistakes
Skipping first-session speed testing. Most studios test difficulty and progression but never time-to-first-reward. There is no published coefficient for what a minute of tutorial costs you, and the ones in circulation are invented. Instrument it, then test it against your own baseline.
Launching monetisation before retention is stable. Adding a shop and IAP events before D7 has settled creates noisy data and spends the trust of your first cohorts. Get a stable D1 first, then introduce meaningful spending moments.
Treating all push notifications alike. Generic pushes buy opt-outs, and on both platforms an opt-out is now close to permanent. Behavioural notifications triggered by last-session state, level position or event timing are a different product from broadcast messaging.
Building LiveOps after launch. Seasonal events, battle passes and content refreshes need to be designed during development. Retrofitting them is slow and it is slow at exactly the moment you cannot afford it.
Measuring blended retention. A single bad UA channel drags the blended D7 below target while the organic cohort is fine. Segment by source, creative and market before diagnosing anything.
Reading unreadable cohorts. The most common retention mistake I see is not a wrong decision, it is a confident decision taken on 150 installs. See Strategy 10.
What We Could Not Verify
This article publishes fewer numbers than the version it replaces, and that is the point. Here is what I looked for and did not publish.
- Genre-level D1, D7 and D30 for 2025 or 2026. Does not exist in openly published form. Every genre table in circulation traces to AppsFlyer Q3 2022. The genre row in this article is labelled with that vintage
- Regional retention splits. Same problem, same 2022 origin. Removed rather than restated
- A D30 figure from GameAnalytics. Its 2025 report covers D1, D7 and D28, but the report is distributed through a lead-capture form rather than an open page, so a reader cannot check the figure without requesting it. The D1 and D7 values here come from that dataset via our internal benchmark reference and carry that limitation
- Percentage-point lift per retention tactic. No source publishes defensible per-tactic lift, and the tables that do are internally impossible. Removed
- The effect of push opt-in on retention post Android 13. The permission change is documented by Google; its retention consequence is not published anywhere I would cite
- Session-length and social-connection coefficients. Widely quoted, sourced nowhere. The directional claims remain in this article, without numbers attached
I would rather leave a gap than fill it with a figure I cannot defend, and after auditing this blog I would rather do that publicly.
The Bottom Line
The honest state of retention benchmarking in 2026 is that the market median is lower than the targets everyone quotes, the genre breakdowns are four years old, and the gap between the two is filled with recycled numbers.
That does not make retention less important. It makes benchmarks less useful than your own cohorts. Use the ranges here to sanity-check whether a plan is plausible, then spend the money to read one clean cohort properly, and set your thresholds from that.
The studios that win on retention are rarely the ones with the cleverest LiveOps. They are the ones who knew which of the three stages was actually broken before they started fixing things.
Want an outside read on your retention curve? Get in touch to discuss how Game Growth Advisor can help, or explore our consulting services.