Half of all live mobile games lose four players out of five within a day of install. That is the middle of the market. The bottom half is worse.
Mobile game KPIs only carry information when you know which population a number describes. Most benchmark tables published for 2026, including the version of this page I shipped in March, print a genre column and a target column without saying whether the figure is a median, a top quartile, or the best hundredth of the market. I have rewritten this page because those columns contradicted each other, and because several other articles on this blog send readers here for the numbers.
Here are the 20 gaming benchmarks worth tracking in 2026. Each one carries its population, its percentile and its vintage. Where a figure does not exist in a primary dataset I could reach, I say so instead of filling the cell.
Key Takeaways
- The market distribution matters more than any genre target. Median D1 is around 22%, median D7 is 3.4% to 3.9%, and median D30 sits between 0.68% and 0.79%. Top quartile D7 is 7% to 8%. The best 1 in 100 games reach 25% D7 and above
- Genre retention data is four years old. The last broadly published genre breakdown is AppsFlyer’s Q3 2022 data. A 2026 article printing a full genre grid is usually reprinting that dataset without dating it
- ARPDAU, blended: roughly $0.01 to $0.05 for ads-only casual, $0.03 to $0.08 for hypercasual, $0.15 to $0.50 for hybrid casual. No primary source I trust publishes mid-core or RPG ARPDAU
- ROAS: Liftoff measured D30 return on ad spend for casual titles at 15% on Android and 47% on iOS, across 2.4 billion installs. The 40% to 60% figure that circulates on benchmark pages has no dataset behind it
- CPI is unreadable without LTV. Plan for a cost per install at 30% to 70% of projected lifetime value, which inverts to an LTV-to-CPI ratio of roughly 1.4x to 3.3x
- Technical KPIs have published thresholds, unlike most of the others. Google Play flags an app once 1.09% of daily users hit a user-perceived crash, and treats a cold start of 5 seconds or longer as slow
How to Read Any KPI Benchmark, Including This One
Three questions decide whether a published number is usable. Ask them before you compare your game to anything.
Which population? GameAnalytics measures live games with at least 1,000 monthly active users, which includes a very long tail of small titles. AppsFlyer measures apps running its attribution SDK, which skews toward studios buying paid traffic. Liftoff measures campaigns on Liftoff inventory. All three describe their sample honestly and get quoted as if they described the industry.
Which percentile? A median and a top quartile of the same metric can sit a factor of two apart. A top 1% figure sits a factor of six above the median. Presenting any of them as a genre target produces the contradiction this page used to carry, where entire genres were benchmarked above the top quartile of the whole market.
Which vintage? Retention by genre stopped being broadly republished after 2022. Rewarded eCPM on casual Android has drifted down across three consecutive halves. A benchmark with no date attached is folklore.
The Market Distribution: Where Your Game Actually Sits
This is the single most useful table on the page, because it maps the whole market instead of listing targets. Find your metric, find your band.
| Band | D1 | D7 | D30 |
|---|---|---|---|
| Median (P50) | ~22% | 3.4-3.9% | 0.68-0.79% |
| Top quartile (P75) | 25-27% on Android, 31-33% on iOS | 7-8% | not published |
| Top decile (P90) | ~40% | not published | not published |
| Top 1 in 100 (P99) | 64-68% | 25% and above | 13-15% |
The median and top quartile rows for D1 and D7 come from GameAnalytics 2025, covering 11,600 games and 1.48 billion monthly active users. The D30 column, the top decile and the top 1 in 100 rows come from the 2026 cut of the same dataset, covering 16K+ live mobile games with at least 1,000 MAU on iOS and Android across nine regions, which states a global median D30 of 0.68% to 0.79%. Cells marked as not published are gaps in the source, and I have left them empty instead of interpolating.
Two consequences. D1 has been trending down year on year, so a flat D1 across releases is quietly an improvement. And a D30 median of well under 1% means the median mobile game has no live audience by the end of the first month. Business viability starts several bands above the middle of this distribution.
Why there is no genre grid on this page any more
The previous version of this article published a seven-genre retention grid. Six of its rows had no traceable source, and one of them benchmarked a genre at three times the measured top quartile of the entire market. I removed it.
What survives verification is a single genre row. AppsFlyer’s Q3 2022 genre data, combined with its Q3 2023 platform split, gives casual and puzzle on Android D1 of 28% to 32%, D7 of 9% to 12% and D30 of 3.5% to 5%. That band reads higher than the whole-market distribution above because it measures a different population, namely apps running an attribution SDK and buying traffic. Read it as a peer group, and remember it is 2022 data.
The KPI Framework: Four Pillars
Every mobile game KPI falls into one of four pillars: engagement, monetisation, acquisition, technical. The dangerous mistake is optimising one while neglecting the others. Having launched 50+ games across F2P, premium and cloud, and managed €12M+ in P&L, I have never seen a title recover from a retention problem through monetisation work.
Engagement KPIs (1-8)
1. D1 Retention
The share of new players who return the day after install. It is the earliest signal of product-market fit and the cheapest to measure.
Locate yourself in the distribution table above. Below 25% on Android, iterate before you spend anything on acquisition. The soft launch process playbook covers the iteration cycles and the go, extend or kill decision framework that turn a D1 reading into a launch decision.
One statistical constraint that saves money: reading a D1 near 30% to within three points at 95% confidence requires 896 installs. Budget roughly 1,000 purchased installs to allow for the drop between install and first session. A single readable cohort of 900 beats nine unreadable cells of 100.
2. D7 Retention
Median across all games is 3.4% to 3.9%. Top quartile is 7% to 8%. The best 1 in 100 reach 25% and above.
D7 reveals whether the core loop sustains interest past the tutorial. It is also where the acquisition maths bites: paid user acquisition in Tier 1 on hybrid casual starts to work around 18% D7, which is more than double the market’s top quartile. That single comparison explains why so many technically competent games never find a profitable channel.
3. D30 Retention
Median across all games is 0.68% to 0.79%. The best 1 in 100 reach 13% to 15%. For casual and puzzle on Android specifically, the 2022 AppsFlyer band is 3.5% to 5%.
D30 is where business viability becomes visible, because it sets the denominator of your cost per retained user. Moving D30 from 3% to 10% cuts the acquisition cost of a retained player by roughly two thirds at any given CPI. For what actually moves it, see our guide on proven retention strategies for mobile games.
4. DAU / MAU (Stickiness)
Formula: DAU divided by MAU, times 100.
I could not find a primary gaming dataset that publishes a DAU/MAU distribution, which means the widely repeated 20% and 30% thresholds have no measured population behind them. Treat the ratio as arithmetic. It approximates the average number of days a monthly player opens the game divided by the days in the month, so 20% means the average monthly player shows up six days out of thirty. Derive it from your own active-days data and track its direction across releases.
5, 6 and 7. Session Length, Sessions Per Day, Daily Playtime
These describe the same behaviour from different angles, so read them together.
| Metric | Median (P50) | Top 1 in 100 (P99) |
|---|---|---|
| Session length | 3.1-3.5 minutes | 22-24+ minutes |
| Sessions per day | 3.8-3.9 | 12-14+ |
| Daily playtime | ~12 minutes | 94-99+ minutes |
A warning on the P99 column, because the previous version of this page got it wrong. Do not multiply the top 1% session length by the top 1% session count and expect the top 1% playtime. Each figure is the 99th percentile of its own metric, and the games behind each column are different games. Titles with the longest sessions are rarely the titles opened most often, which is why 22 minutes times 12 sessions lands nowhere near 94 minutes.
Regionally, Africa stands out for habitual play at a median of 5.48 sessions per day, published at that precision by GameAnalytics. Session length tracks design intent more than quality, so compare yourself to games with a similar loop instead of to the market.
8. Churn Rate
Formula: players lost in the period divided by players at the start, times 100. The inverse of retention, and mostly redundant with it at the aggregate level. Churn earns its place when segmented by acquisition source, geography and spending tier, because that is where you find out which channel is buying players who never come back.
Monetisation KPIs (9-14)
9. ARPDAU
| Monetisation model | Blended ARPDAU | Population |
|---|---|---|
| Casual, ads only | $0.01-$0.05 | Casual titles where rewarded video carries the economy |
| Hypercasual, blended | $0.03-$0.08 | High-volume, low-depth titles |
| Hybrid casual, blended | $0.15-$0.50 | Ads plus IAP, with IAP at 40-60% of revenue |
The $0.03 to $0.08 band gets quoted constantly as a casual benchmark. It is a hypercasual figure, and applying it to a casual title with a real meta layer will make a healthy game look broken.
For ad-monetised titles, do not benchmark ARPDAU at all until you have decomposed it:
ARPDAU_ads = (rewarded impressions per daily user) x eCPM / 1000Both inputs are measurable in your own data within a week, and the decomposition tells you which one to fix. How much ARPDAU you can expect depends on your monetisation model, whether IAP, ads, hybrid or subscription. For studios building in the fastest-growing category on mobile, our 2026 guide to hybrid casual games covers the retention and monetisation signals that indicate product-market fit.
10. LTV (Lifetime Value)
LTV = (area under your retention curve) x ARPDAUThe ratio that matters is CPI against projected LTV, and the rule practitioners use is that CPI should land at 30% to 70% of projected lifetime value. The tighter end applies when you need fast payback or carry little risk tolerance. Inverted, that is an LTV-to-CPI ratio of roughly 1.4x to 3.3x. The 3:1 target that circulates in planning decks sits at the conservative end of that band, so it is a reasonable place to aim once you add the rest of your acquisition cost stack on top of media. It stops being reasonable when it is quoted as a law without a measured LTV underneath it. For how LTV, contribution margin and the cost stack connect in a real game financial model, see our mobile game unit economics guide.
11. Conversion Rate (Free to Paying)
I searched for a reliable published conversion rate for casual puzzle and did not find one, so this article does not print a target. Track time-to-first-purchase alongside the rate itself. If players convert late, the problem sits in your early monetisation touchpoints, upstream of pricing. Your economy design, including currencies, paywall placement and progression pacing, is the primary lever: our game economy design guide covers how to balance sinks and sources to accelerate that first purchase.
12. ARPPU (Average Revenue Per Paying User)
ARPDAU spreads revenue across everyone, ARPPU isolates the payers. The relationship between conversion rate and ARPPU describes your monetisation shape. Low conversion with high ARPPU means whale-dependent revenue, which is fragile to a handful of churning accounts. Higher conversion with moderate ARPPU indicates a broader base. Neither has a published benchmark worth quoting.
13. eCPM (Effective Cost Per Mille)
Rewarded video eCPM for casual titles on Android, by region, from TopOn’s H1 2025 monetisation report:
| Region | Rewarded eCPM, casual Android |
|---|---|
| EU and North America | $8.90 |
| Hong Kong, Macao, Taiwan | $7.42 |
| Japan and Korea | $6.87 |
| LATAM | $2.18 |
| Southeast Asia | $1.64 |
| South Asia and India | $0.10 |
TopOn publishes these to two decimals, so they are reproduced here at source precision. Three things in this table change how you plan.
The Tier 1 to Tier 3 spread is nearly 90x, far wider than the CPI spread across the same markets, which is why buying cheap installs in low-eCPM markets rarely improves unit economics as much as expected. The Android to iOS gap widens sharply as you go down the tiers, with Android rewarded eCPM at 73% of iOS in EU and North America but only 3.6% of iOS in India. And in India the usual format hierarchy inverts entirely: Android interstitial earns $0.47, roughly five times rewarded at $0.10.
The direction of travel matters too. Casual Android rewarded eCPM overall went from $3.60 in H1 2023 to $3.02 in H1 2025, so a flat eCPM year on year is a relative gain.
14. IAP Revenue per Paying Session
Uncommon and highly diagnostic. It tells you whether paying players are spending more or less per session over time, which is a leading indicator of monetisation fatigue well before ARPPU moves.
Acquisition KPIs (15-18)
15. CPI (Cost Per Install)
There is no single useful CPI number, and any source quoting one to the cent is hiding its methodology. Plan by market. These are ranges for casual and puzzle on Android, the segment most small studios are planning for:
| Market | Android CPI, casual and puzzle | Confidence |
|---|---|---|
| United States | $1.50 - $3.50 | Moderate to high |
| UK, Canada, Australia | $1.00 - $2.50 | Moderate |
| Western Europe | $0.60 - $1.50 | Moderate |
| LATAM | $0.15 - $0.60 | Moderate |
| Southeast Asia | $0.20 - $0.60 | Low to moderate |
| India | $0.08 - $0.30 | Low to moderate |
For iOS, apply a 3x to 4x multiplier within the same market and genre. Published iOS-to-Android ratios span 1.5x to 10x, which reveals inconsistent methodology rather than real variance.
Never place a blended global figure and a geo-level figure in the same table. A volume-weighted global average is dominated by markets where installs cost pennies, and it can differ from a Tier 1 estimate by more than a factor of ten while both are accurate. For the full treatment of why published CPI benchmarks disagree so violently, see our UA cost benchmarks by genre and platform. Studios adapting measurement to privacy changes should also review our guide to SKAN and Privacy Sandbox attribution.
16. ROAS (Return on Ad Spend)
Liftoff measured D30 return on ad spend for casual titles at 15% on Android and 47% on iOS, across 2.4 billion installs. On category averages, a rewarded-led casual title on Android recovers roughly a seventh of its acquisition spend in the first month.
That figure is far below the 40% to 60% that circulates on benchmark pages, and the gap matters, because planning against the higher number produces a budget that never pays back. If early ROAS looks strong and D30 collapses, you have a retention problem long before you have a channel problem.
17. Organic Multiplier
Formula: total installs divided by paid installs. It reflects store optimisation, word of mouth and brand strength. No primary dataset publishes a distribution for it, so the commonly quoted 1.5x to 2x healthy range is practitioner convention with no measurement behind it. Track it against your own baseline, and treat a multiplier drifting toward 1.0 as a warning that your growth is entirely rented.
18. K-Factor (Virality)
Formula: invites sent per user multiplied by the conversion rate of those invites. Above 1.0 means self-sustaining viral growth, which is rare outside social and social casino titles. Even a K-factor of 0.3 meaningfully reduces effective CPI, since it adds free installs to every paid one.
Technical KPIs (19-20)
These are the only two KPIs on this page with an authoritative published threshold, because the platform owner sets it.
19. Crash Rate
Google Play flags an app for bad behaviour once at least 1.09% of daily users experience a user-perceived crash across all device models, or 8% on a single device model. The equivalent thresholds for application-not-responding events are 0.47% across all models and 8% on a single model.
Those are the points at which store visibility is at risk, so treat them as a ceiling and set your own target well below. Monitor by device model and OS version, because a single bad device can breach the per-model threshold while your global rate looks healthy.
20. Load Time
Android vitals treats a cold start of 5 seconds or longer as slow, a warm start at 2 seconds or longer, and a hot start at 1.5 seconds or longer. Cold start is the one that costs you installs, since it sits between the store and the first session, and it is the number your D1 cohort silently pays for.
Building Your KPI Dashboard
Not every studio needs all 20 from day one. Sequence by stage.
Soft launch: D1 and D7 retention, session length, crash rate, cold start time. Resist adding revenue metrics before the cohorts are readable.
Global launch: add CPI, ROAS, DAU, ARPDAU and organic multiplier.
Live operations: the full set, with LTV, D30, ARPPU, eCPM by region and churn segmented by acquisition source.
AI-powered analytics tooling is making real-time tracking cheaper, and our guide on how AI is reshaping mobile game development covers where it helps and where it adds noise. If you want an outside read on which metrics matter for your specific title, or a full audit of your current setup, explore mobile game consulting.
What I Could Not Verify
The following figures appear routinely on 2026 benchmark pages, including on earlier versions of this one. I searched for a primary dataset behind each and did not find one, so they are not published here.
- ARPDAU for mid-core and RPG titles. No source I would rely on breaks ARPDAU out at that granularity
- eCPM below the casual level. No public source separates puzzle, match or arcade
- Free-to-paying conversion rates for casual puzzle. Widely quoted, never sourced
- A DAU/MAU distribution for mobile games. The 20% and 30% thresholds trace back to general app investing heuristics with no gaming dataset behind them. The stickiness figure frequently attributed to a market-leading match-3 title has never been published by its developer
- Genre-level retention beyond casual and puzzle on Android. The last broad breakdown is 2022
- Distributions for organic multiplier and K-factor. Both are practitioner conventions
- An absolute go or no-go CPI. It does not exist, because the threshold is always relative to LTV
- Rewarded impressions per daily user by genre. The 3 to 5 per day figure in circulation is a design recommendation that nobody has measured
Admitting a gap costs less than filling it. If you find one of these in a published table, check whether the page names a primary dataset with a date. If it does not, the number is being recycled.
Conclusion
The mobile market in 2026 does not lack benchmarks, it lacks benchmarks that state what they measured. Studios that win are the ones who know which population a number describes, which percentile they are comparing against, and how old the data is. Whether you are scaling a mobile title or building a cloud gaming telco partnership, the same discipline applies.
KPIs are most useful inside a broader operating system, and our end-to-end game growth strategy connects acquisition, retention and monetisation around a single north star. If your studio lacks the senior leadership to own that process, a fractional CMO with gaming analytics expertise can establish the framework and coach the team.
Download our free F2P audit checklist to benchmark your own game against the distributions above, or book a call if you want someone who has managed the metrics behind 50+ launches to read your numbers with you.