How to Read AppsFlyer's Annual Gaming Benchmarks Report

AppsFlyer's annual gaming app marketing report landed this month, its usual January fixture, with a fresh set of CPI and retention benchmarks plus fraud figures, broken out by genre and region. The AppsFlyer gaming benchmarks report is one of the more widely cited documents a UA team will see all year. It's also one of the most commonly misread, because a benchmark built from an aggregate of AppsFlyer's own customer base isn't the same thing as a number that applies to any one studio's specific title or geo mix or channel blend.

What the report is actually measuring

AppsFlyer's benchmarks come from attribution data flowing through its own platform, which means the sample is real installs and real spend, not a survey, but it's also a sample skewed toward AppsFlyer's own customer mix, which leans toward mid-size and enterprise publishers with the budget to run an MMP in the first place.

That isn't a flaw. Every vendor benchmark has a sampling bias baked into its own book of business. It does mean a small studio or a hyper-casual publisher with thin MMP coverage should expect its own numbers to sit somewhat outside the reported range, in either direction, without that being a signal of underperformance.

Genre and region breakdowns help here, since comparing within a genre segment narrows the sampling problem meaningfully compared with comparing against a single blended headline figure. If strategy is the category in question, pull the strategy CPI and retention benchmarks, not the report's overall average.

Reading the CPI figures without over-fitting

CPI benchmarks in a report like this are best read as a range and a direction, not a target. If the report shows CPI for a given genre and region roughly stable or drifting up year on year, that direction is more actionable than the specific figure, because CPI benchmarks are inherently backward-looking and blend a wide spread of buying strategies, spanning broad prospecting and narrow lookalike targeting, that any single campaign won't replicate exactly. As we noted in "Detecting Creative Fatigue Before It Shows Up in CPI," CPI itself is a lagging signal even inside a single campaign's own data, and a report aggregating thousands of campaigns lags further still.

The useful exercise is comparing the report's year-on-year direction against a team's own year-on-year direction for the same genre and region. Suppose a benchmark shows CPI up roughly 10 to 15% and a team's own blended CPI moved less than that. The team is either buying more efficiently than the aggregate or has a channel mix the benchmark's sample doesn't fully reflect, and either explanation deserves investigation rather than an assumption that the benchmark is simply wrong.

Retention and fraud: treat differently

Retention benchmarks travel better across studios than CPI benchmarks do. Day-one retention (and day-seven and day-thirty retention with it) is primarily a function of product and onboarding quality rather than buying strategy, so a genre-level retention benchmark is a more honest yardstick for where a title's own onboarding sits relative to its category, and a title tracking meaningfully below a genre's reported retention benchmark has a product or onboarding problem worth investigating before spending more on acquisition into that funnel.

Fraud figures in AppsFlyer's report deserve the most caution of any figure in the document, since fraud rates vary enormously by channel and geo, and again by campaign type, in ways a blended genre figure will smooth over entirely. A studio buying heavily in geos or channels with known higher fraud exposure should expect its own exposure to sit above a blended benchmark. The operative number for any given campaign is the studio's own MMP fraud dashboard, not the industry report.

Using the report in a vendor or leadership conversation

The AppsFlyer gaming benchmarks report is also useful outside the planning meeting, as a reference point in conversations with ad networks and with leadership, and precision matters there too. When a network account manager cites the report to argue a campaign's CPI is "in line with the market," ask which genre and region cut they're referencing, since a blended figure can make an underperforming campaign look acceptable if the comparison genre is a cheaper one than the title actually competes in. The same caution applies in reverse. A network is unlikely to volunteer a narrower cut that makes its own performance look worse, so the burden sits with the buying team to ask for the specific segment before accepting a benchmark comparison at face value.

With leadership, the report is more useful for calibrating expectations than for justifying a specific number. A finance stakeholder asking why CPI rose this quarter is better served by a comparison against the report's year-on-year genre direction, framed as "the market moved by roughly this much and our number moved by this much," than by a defence built entirely around the team's own internal data, which a stakeholder outside the UA function has no external reference point to judge.

A worked comparison, hypothetically

Take a hypothetical mid-core strategy title with a blended CPI that rose 18% year on year. If AppsFlyer's report shows strategy CPI in the relevant region up in the 10 to 15% range, the title's own rise sits meaningfully above the reported direction, which deserves a specific investigation rather than a note filed under "in line with the market rising." Start with channel mix. A shift toward a higher-cost channel can explain a blended rise larger than a stable-mix comparison would predict, and that possibility needs ruling out before anyone assumes the team's buying efficiency itself has deteriorated relative to the market.

Using the AppsFlyer gaming benchmarks report this week

Pull the genre and region cut that matches the team's own title mix, not the headline figures, and set them beside the team's own trailing twelve-month CPI and retention numbers for the same segment. Where the team's own direction diverges from the benchmark's direction, rather than only the level, that divergence is the finding worth taking into the coming planning meeting. Sampling bias can explain a level gap. A direction gap usually means something specific changed in the team's own channel mix or creative or fraud exposure, something the benchmark cannot see.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• The UA reading list: reports and benchmarks that matter
• Liftoff's Mobile Game CPI Benchmarks 2025, Read Correctly