What Is IPM? Meaning, Formula, and Mobile Game IPM Benchmarks 2026, Plus How Many Impressions a Creative Test Needs

IPM (installs per mille) is the number of app installs generated per 1,000 ad impressions. You calculate it by multiplying installs by 1,000 and dividing by impressions. Simple as it is, in mobile game UA it decides which creatives win, sets your cost per install (CPI), and even determines how many impressions your campaign gets in an ad network auction. A creative that turns the same impressions into twice the installs halves your CPI and doubles your buying power at the same bid.

The trouble is benchmarks. Gather the public game IPM figures available as of October 2026 and they range from 0.78 on iOS to 47.6 for hypercasual, a spread of more than 60x, and even the "all games average" runs anywhere from about 2 to over 8 depending on the source. The numbers are not wrong. They differ because the channel, format, genre, OS, and the definition of an impression all differ.

The first step of the funnel, click-through rate, is covered in What Is a Good Ad CTR for Mobile Games? 2026 CTR Benchmarks by Channel, Format, and OS, and When a High CTR Misleads. This post covers the next step: the rate at which impressions become installs. It walks through the definition and formula of IPM and its relationship to CPI and eCPM with worked examples, lays out benchmarks by genre, OS, format, and channel with their sources, and then looks at how many impressions a creative test needs before IPM can decide a winner, and when a high IPM still loses money.

What Is IPM, and How Is It Calculated?

IPM is installs per 1,000 impressions, and the formula is IPM = installs × 1,000 ÷ impressions. It lets you compare the install efficiency of creatives and campaigns on the same unit, the ad impression, and combined with CPM it becomes CPI directly.

For example, if a creative was shown 18,000 times and produced 45 installs, its IPM is 45 × 1,000 ÷ 18,000 = 2.5. For every 1,000 impressions, 2.5 people installed. "Mille" is Latin for thousand, so the "installs per mile" spelling found in some sources is a typo.

IPM breaks down into two conversion rates. Multiply the impression-to-click rate (CTR) by the click-to-install rate and by 1,000, and you get IPM. With a 3% CTR and an 8% click-to-install rate, IPM is 0.03 × 0.08 × 1,000 = 2.4. This decomposition is how you find the cause when IPM is low. If CTR is normal but IPM is low, the problem is more likely the gap between what the creative promised and what the store page shows than the creative's ability to get attention.

Move to cost and one more formula applies: CPI = CPM ÷ IPM. With a $12 CPM and an IPM of 2.4, CPI is 12 ÷ 2.4 = $5.00. Swap the creative at the same CPM and raise IPM to 3.0, and CPI falls to $4.00. A 25% rise in IPM means a 20% drop in CPI. Conversely, a 20% fall in IPM raises CPI by 25%. Because IPM and CPI are inversely related, an IPM decline hits cost harder than an IPM gain of the same size helps it.

The relationship shows up in real market data. Singular's Q2 2026 Quarterly Trends Report found that IPM fell 10.5% while CPM barely moved, and global CPI rose 12.0% to $1.24. With CPM fixed, a 10.5% IPM decline should push CPI up by 1 ÷ 0.895 ≈ 1.117, or about 11.7%, which nearly matches the reported figure. Working backward, if you know IPM and CPI you can infer CPM. Multiplying the 2025 game IPM of 8.62 and the global game CPI of $0.56 from Adjust's 2026 gaming app report gives an implied CPM of about $4.83 for that dataset (a value calculated from the published figures, not stated in the report).

The price side is covered in What Are 1,000 Ad Impressions Worth? 2026 Mobile Game eCPM Benchmarks by Format, Region, and OS, and the Q4 Price Window and Mobile Game CPI Benchmarks 2026: Why One Average Is Useless. IPM is the bridge between those two numbers.

What Are the 2026 Mobile Game IPM Benchmarks by Genre and OS?

Based on public data, game IPM ranges from around 3 to over 40 depending on genre, and within the same genre Android runs several times higher than iOS. Absolute values differ by source, but the order of genres and the direction of the OS gap are consistent across sources.

Genre makes the biggest difference. In AppsFlyer's 2024 creative report, hypercasual IPM on ad networks was 47.6 against 3.1 for RPG, a 15x gap. Liftoff and Singular's 2025 Casual Gaming Apps Report likewise found that casual games, including hypercasual, consistently post higher average IPM than casino and RPG. A game anyone understands at a glance makes it easy to decide on an install from one ad; genres with a higher barrier to entry produce fewer installs from the same impressions. Comparing an RPG's or strategy game's IPM to a hypercasual baseline is meaningless.

The order of genres holds fairly steady across sources. Racing ranked first in Adjust's data (16.2) and in Singular's Q2 2024 Android data (26.71). The values differ, but the rank is the same, and that is the point. What external benchmarks can give you is relative position more than absolute value.

The OS direction is consistent too. In Singular's Q2 2024 data, game IPM was 0.78 on iOS and 4.58 on Android, a gap of roughly 6x, and its Q3 2025 report found that Android ads drive almost three times as many installs per impression as iOS. Liftoff and Singular's report also found Android IPM considerably higher than iOS across every genre studied. Even kids games, the genre with the highest iOS IPM (4.3), fell short of Android (6.1). Android tends to have less friction between the click and the install in the store, and on iOS, where SKAdNetwork installs are reported in aggregate and with a delay, there is also room for IPM to be measured lower than it really is.

By region, only the direction is clear. According to Adjust, game IPM fell in 2025 in Europe, LATAM, MENA, and North America, and rose only in APAC. Singular's Q2 2025 report put average IPM by country at around 3–5, with China a much lower outlier. The trend continued into 2026: as noted above, Singular's Q2 report showed IPM down 10.5% from the previous quarter. With the same impressions producing fewer installs, using benchmarks from one or two years ago as targets may set the bar too high.

Why Do IPM Benchmarks Differ So Much Between Sources?

Game IPM benchmarks differ between sources for five main reasons: channel type, ad format, genre mix, OS mix, and the way impressions are counted. Two numbers that do not hold these five fixed cannot be compared.

Line up just the numbers published under the name "game IPM" and the gap is obvious. The IPM calculated in our CTR post from the Liftoff × Singular in-app network sample totals is about 2.2, Singular's country average is 3–5, Adjust's all-games IPM is 8.62, and AppsFlyer's ad network hypercasual figure is 47.6. Apple Ads runs higher still. Calculated from the sample totals SplitMetrics published, IPM for its 2024 data (4.4 billion impressions, 253.1 million downloads) is about 57.5, and for its 2026 edition sample (6.3 billion impressions, 315.6 million downloads) about 50.1 (both values are calculated from totals and include non-game categories).

The first reason is channel type. Search ads are shown after the user has typed a query themselves, so installs per 1,000 impressions reach into the dozens. In-app ad networks often reach people who are already playing, through interstitial and rewarded formats, so they tend to run more efficient than social feeds. For social channels such as Meta and TikTok, we could not find, as of October 2026, a neutral source that publishes game IPM by platform along with its dataset. Each dataset has a different channel mix, so an average from data weighted toward search and one weighted toward social are different numbers from the start.

The second is ad format. There is practically no neutral source that publishes absolute game IPM by format with its dataset; what is public is the ratio between formats.

The order is clear. Playables, which let the user try the game, are highest; banners at the edge of the screen are lowest; video sits in between. But Liftoff's 8x and 16x figures pool results from campaigns run in each format rather than testing the same creative across formats, so they should not be read as "build a playable and your IPM goes up 8x." When a playable's production cost pays off is covered in Playable Ads: The Conditions That Justify a $5,000 Production Bill. Rewarded video is a format with many chances to install per impression, since players watch to the end for a reward and the end card leads to the store, but on iOS in 2025 that same structure was identified as a cause of clicks inflated by accidental taps. That is why a high IPM on rewarded inventory should be checked against the click-to-install rate and post-install retention.

The third and fourth are genre and OS mix. As the previous section showed, the gap between genres reaches 15x and the gap between OSes 3–6x. A dataset heavy on hypercasual and Android will inevitably show a high average IPM, and one heavy on RPG and iOS a low one.

The fifth is how impressions are counted. The impression in IPM's denominator is less consistent than it looks. It can be counted when the ad server sends the ad, or only when the ad is actually rendered on screen or a certain share of it is visible. Attribution providers (MMPs) only receive impressions from media sources that share impression data, so the range of impressions captured in the denominator varies with the media mix. On iOS, SKAdNetwork installs arrive aggregated at the campaign level and delayed, which leaves room for IPM to be measured low: impressions are all counted while only some installs are linked. Public reports rarely spell out these measurement definitions, so it is safer to avoid comparing absolute values across sources in the first place.

In short, use external IPM benchmarks only to check whether your game is off the market by an order of magnitude, and set targets from your own data for each combination of channel, format, OS, and country.

How Many Impressions Does a Creative Test Need Before IPM Can Decide?

To judge creatives on IPM, size the sample by installs rather than impressions: 50 installs per variant is enough for a first screen that catches large differences, while telling apart a 20–30% difference takes roughly 200–430 installs per variant. Winners that appear on small install counts are mostly chance.

Unity's creative testing guide, published in October 2026, sets out the same principle. It recommends deciding in advance that the deciding metric is usually IPM for early testing and D7 ROAS for scale, warns that apparent winners on small install counts are often random, and calls deciding within hours a common mistake. As a budget example, it notes that at a $2 CPI and a threshold of 50 installs per variant, the budget comes to roughly $100 per variant.

You can calculate how precise 50 installs actually is. Installs are rare events and can be approximated with a Poisson distribution, and the range in which the true IPM likely sits (95% confidence interval) depends on the number of installs observed.

An IPM of 2.0 measured on 50 installs could really be 1.5 or 2.6. Trying to separate two creatives at 2.0 and 2.3 at this stage is no better than a coin flip. Fifty installs is a first-screen threshold for filtering out creatives at half or double the IPM.

Separating two creatives with statistical confidence takes more installs. Using standard A/B test settings (two-sided 5% significance, 80% power), the installs needed per variant are set almost entirely by the size of the difference you want to detect, regardless of baseline IPM, and the impressions needed are those installs divided by the baseline IPM.

The key point is that the lower a genre's IPM, the more impressions the same verdict requires. An RPG with an IPM around 3 needs close to 150,000 impressions per creative to detect a 20% difference, while a hypercasual game with an IPM in the 40s needs just over 10,000. That is why test budgets and durations should differ by genre.

Early kill rules can be built on the same math. If the IPM of your control creative is established on enough volume, a new creative can be stopped once its impressions reach the point where the control's IPM would have produced 30 installs, if it has 15 or fewer. The probability that a new creative actually as good as the control lands at 15 or fewer by chance is only about 0.2%. With a control IPM of 2.0, that means stopping a new creative with 15 or fewer installs, an IPM of 1.0 or below, at 15,000 impressions. Declaring winners early, on the other hand, is a mistake to avoid. An IPM that starts high often drifts back toward the average as the sample grows.

In practice, four rules cover it. Compare within the same network, placement, OS, country, and period, splitting impressions evenly. Set the installs needed for a verdict before the test, and convert them into impression targets and budget using the expected IPM. Use early kills only for clearly bad creatives as above, and declare winners only after the planned sample is filled. And run creatives that pass on IPM through D1 and D7 retention and D7 ROAS before increasing budget. Monitoring continues after scaling: Unity's guide treats a top creative's IPM sliding while spend holds steady as a sign of fatigue and a cue to prepare a refresh. Creative test design as a whole is covered in Why Creative Testing Is the Highest-Leverage ROAS Driver in Mobile Games.

When Does a High IPM Still Lose Money?

IPM shows efficiency up to the install, but not whether the people who install stay or spend. A creative with a high IPM losing on D7 ROAS is common, and when budget follows IPM alone, installs grow while payback slows.

Compare two creatives at the same $10 CPM. Creative A has an IPM of 5.0, so its CPI is $2.00, and with $0.18 in 7-day revenue per install its D7 ROAS is 9%. Creative B has an IPM of 3.0 and a CPI of $3.33, but with $0.60 in 7-day revenue per install its D7 ROAS is 18%. Spend $1,000 and A delivers 500 installs and $90 in 7-day revenue, while B delivers 300 installs and $180. The creative with a 67% higher IPM pays back at half the speed.

There are three typical cases where a high IPM comes with lower install quality. The first is a creative that makes a promise the game cannot keep. Creatives showing scenes or mini-games that are not in the actual game bring in installs well but lose many players on day one. The second is the structure of the genre itself. AppsFlyer reported a 47.6 IPM for hypercasual on ad networks, while cautioning that hypercasual games often struggle with post-install monetization, so a high IPM does not tell the whole story. The third is inventory mixed with accidental taps or measurement artifacts. On placements where clicks are inflated, a larger share of the users who make it to install may have arrived without intent.

So IPM only becomes meaningful when multiplied by or set beside post-install metrics. Multiply IPM by D1 retention and you get "users still around the next day per 1,000 impressions," and comparing creatives on that number separates creatives that only win installs from those that bring players who stay. The decision to scale is made on D7 ROAS. D7 ROAS baselines by genre and monetization model are covered in D7 ROAS Benchmarks: The Targets That Tell You a Campaign Will Pay Back.

How Does IPM Interact With eCPM Bidding on Ad Networks?

On ad networks where you bid on CPI, your campaign's competitiveness is set by the product of your bid and your IPM, that is, eCPM. A creative with a high IPM wins more impressions at the same bid, and when IPM drops, impressions shrink even if the bid stays the same.

Advertiser-side eCPM is calculated as eCPM = IPM × CPI. Unity's glossary gives this formula and explains that the higher the eCPM, the more scale and impressions a campaign can unlock, and that improving IPM boosts eCPM and can raise a campaign's rank in the waterfall. For the publisher selling impressions, showing first the campaign that earns the most per 1,000 impressions is the rational choice. Note that publisher-side eCPM (ad revenue ÷ impressions × 1,000) shares the name but is calculated by a different party.

The numbers make the structure clear. If campaign A bids $3.00 per install with an IPM of 4.0, its eCPM is $12.00. Campaign B bids $4.00 but has an IPM of 2.5, for an eCPM of $10.00. A wins the auction with a 25% lower bid. To beat A, B would need to bid $4.80 or more. Now suppose A's creative fatigues and its IPM drops to 3.0. A's eCPM falls to $9.00 and it loses to B; to hold its position, it would have to raise its bid to $4.00.

Two practical rules follow. First, when impressions fall, check IPM before raising the bid. Raising the bid by 20% and raising IPM by 20% have the same effect on eCPM, but a bid increase raises your cost per install by the full 20%, while an IPM improvement grows volume at the same cost per install. Second, treat creative refresh cadence as part of bidding strategy. The longer a fatigued creative stays in rotation, the higher the bid needed to keep the same impressions.

The principle is similar on social channels with different bidding mechanics. Meta explains that, unlike traditional auctions that pick the winner by highest bid, it weighs several factors to decide which ad to show, using estimated action rates and ad quality to measure relevance. For a campaign optimized for installs, the estimated action rate is effectively predicted install efficiency, so a creative with a high IPM is likely to be shown on better terms for the same budget. But ad quality assessment also takes in low-quality signals such as clickbait and exaggeration, so an IPM inflated by overpromising is unlikely to stay an advantage in the auction. How to choose creatives as AI multiplies creative volume is covered in AI-Generated Ad Creatives in Game UA: Volume Is Easy Now. Judgment Isn't.

Buying Play Instead of Impressions: The Playio Perspective

IPM is a risk carried by whoever buys the impressions. When you buy on CPM or in an impression-based auction, CPI rises the moment IPM falls; in a structure that prices installs or completed in-game actions, that swing does not pass directly into the advertiser's cost.

As shown above, the trend in 2026 is for the same impressions to produce fewer installs, and creative fatigue and auction competition keep pulling IPM down. Run UA on impression-based channels alone in this environment, and a large share of the effort spent on creative testing and bid adjustment goes into defending against falling IPM. Running a channel that charges on outcomes alongside them separates part of the portfolio's cost from IPM swings.

Playio runs an Android-based gamer community of five million players, analyzing genre preferences, play history, and in-game behavior data to connect games with players whose tastes match. Pricing is CPI by default, and CPE based on reaching a playtime threshold or completing a specific in-game action is also supported. Because what advertisers buy is installs and post-install play rather than impressions, costs stay predictable even when IPM is volatile, and performance can be judged directly on the post-install metrics emphasized above.

You can find more details here. (https://playioadsen.oopy.io/bizdeck)

Key Takeaways

IPM is installs per 1,000 impressions (installs × 1,000 ÷ impressions); it breaks down into CTR × click-to-install rate × 1,000 and connects directly to cost through CPI = CPM ÷ IPM. A 25% rise in IPM cuts CPI by 20%, and when IPM fell 10.5% in Q2 2026, global CPI rose 12.0% even though CPM barely moved. As of October 2026, public benchmarks such as 8.62 for all games (Adjust, 2025), 16.2 for racing, 5.13 for casino, and 47.6 for hypercasual versus 3.1 for RPG on ad networks (AppsFlyer) show a spread of 15x or more by genre, and Android delivers 3–6x more installs per impression than iOS. Numbers differ by source because channel, format, genre, OS, and the definition of an impression differ, so use external figures only as an order-of-magnitude check and set targets from your own data. In creative tests, size the sample by installs rather than impressions: 50 installs for a first screen, 200–430 per variant to decide a 20–30% difference, and stop any creative with 15 or fewer installs at the point where the control would have produced 30. Run IPM winners through D7 ROAS again, and when impressions fall, check IPM before the bid: that is how you protect cost in an eCPM auction.

For inquiries about Playio's advertising solutions, reach out at: [email protected]

Sources

• Unity, Installs per mille glossary (IPM = installs × 1,000 ÷ impressions; improved IPM boosts eCPM and waterfall rank): https://unity.com/glossary/installs-per-mille

• Unity, eCPM glossary (advertiser-side eCPM = IPM × CPI; higher eCPM unlocks more scale and impressions): https://unity.com/glossary/ecpm

• Unity, What is creative testing? A guide for mobile game creatives, 2026-10-15 (IPM for early testing, D7 ROAS for scale; winners on small install counts are often random; $2 CPI × 50 installs ≈ $100 per variant; top creative's IPM sliding at steady spend signals fatigue): https://unity.com/resources/ad-creative-testing-guide

• GameDev Reports, Adjust: Gaming App Insights Report 2026, 2026-04-15 (game IPM 8.62 in 2025, 8.59 in 2024; racing highest at 16.2; casino up 65% to 5.13; IPM fell in Europe, LATAM, MENA, North America, grew in APAC; global game CPI $0.56; January 2024 to January 2026, top 5,000 apps and full Adjust dataset): https://gamedevreports.substack.com/p/adjust-gaming-app-insights-report

• Adjust, Gaming app insights 2026 (original report page for the figures above): https://www.adjust.com/blog/gaming-app-insights-2026/

• Segwise, CPI, IPM, and ROAS Benchmarks for Optimizing Ad Spend, updated July 2026 (cross-check of Adjust figures 8.59 to 8.62, racing 16.2, casino 5.13): https://segwise.ai/blog/cpi-ipm-roas-benchmarks-optimizing-ad-spend/

• AppsFlyer, The state of ad creatives in app marketing: 2024 edition (220k creatives, 2,000 apps, 720M non-organic installs; ad network IPM hypercasual 47.6, RPG 3.1; game video with animation 26% higher IPM; caution that hypercasual's high IPM does not capture post-install monetization challenges): https://www.appsflyer.com/resources/reports/creative-optimization-2024-report/

• GameDev Reports, Singular: Mobile Market Trends in Q2'24, 2024-08-13 (game IPM iOS 0.78, Android 4.58; Android racing 26.71, educational 24.17, trivia 13.56; iOS educational 12, trivia 8.7, hypercasual 4.5): https://gamedevreports.substack.com/p/singular-mobile-market-trends-in

• GameDev Reports, Singular: Trends in Mobile Game Marketing in Q2'25, 2025-07-14 (average IPM by country 3–5, China a much lower outlier): https://gamedevreports.substack.com/p/singular-trends-in-mobile-game-marketing

• Singular, Mobile gaming CTR is out of control, 2025-08-13 (Q3 2025 report highlight: Android ads drive almost 3x more installs per impression than iOS): https://www.singular.net/blog/mobile-gaming-ctr/

• Singular, Q2 2026 Quarterly Trends Report landing page (IPM -10.5%; global CPI +12.0% to $1.24; CPM barely moved): https://lp.singular.net/q2-2026-quarterly-trends-report

• Liftoff, 2025 Casual Gaming Apps Report (Singular data, February 2024 to February 2025; kids IPM iOS 4.3, Android 6.1; Android IPM higher than iOS across all genres; casual average IPM higher than casino and RPG): https://liftoff.ai/?p=34267

• MarTech Series, Liftoff Mobile Ad Creative Index (2024), 2024-03-04 (January 2023 to January 2024; game playables about 20x more likely to drive installs than banners): https://martechseries.com/mobile/mobile-advertising/liftoff-mobile-ad-creative-index-reveals-ugc-and-interstitial-ads-are-key-to-unlocking-advertising-roi/

• Liftoff 2025 Mobile Ad Creative Index press release, 2025-07-07 (4.7T impressions, 263B clicks, 1.1B installs; playable impression-to-install rate 8x non-playable for top spenders, 16x for other game advertisers): https://www.newswire.ca/news-releases/liftoff-s-2025-mobile-ad-creative-index-reveals-ugc-is-a-key-differentiator-for-top-apps-837345804.html

• App Developer Magazine, SplitMetrics Apple Ads Search Results Benchmarks Report 2025, 2025-05-30 (January to December 2024; 4.4B impressions, 395.4M taps, 253.1M downloads): https://appdevelopermagazine.com/apple-ads-search-results-benchmarks-report-2025/

• SplitMetrics, Apple Ads Search Results Benchmarks Report 2026 (sample totals 6.3B impressions, 495.6M taps, 315.6M downloads): https://splitmetrics.com/apple-ads-search-results-benchmarks-2026/introduction

• Meta for Business, Ad auction (winner not chosen by highest bid alone; estimated action rates and ad quality used to measure relevance; low-quality signals such as clickbait and exaggeration considered): https://www.facebook.com/business/ads/ad-auction

• Illustrative calculations: IPM 2.5 example (45 × 1,000 ÷ 18,000); 3% CTR × 8% = IPM 2.4; $12 CPM ÷ IPM 2.4 = $5.00, at IPM 3.0 = $4.00; Singular Q2 1 ÷ 0.895 ≈ +11.7%; Adjust implied CPM $0.56 × 8.62 ≈ $4.83; Apple Ads IPM from totals about 57.5 (2024) and about 50.1 (2026 edition); Poisson 95% confidence intervals (20, 50, 100, 400 installs); installs per variant (two-sided 5%, 80% power, two-proportion z-test approximation); early kill rule P(X ≤ 15 | expected 30) ≈ 0.2%; creative A/B D7 ROAS example ($10 CPM); eCPM example ($3 bid × IPM 4.0 vs $4 × 2.5)