For ad-monetised games, 89 percent of the money is already in by day 7.

That is AppsFlyer’s State of App Monetization 2026, across 7.2 billion dollars of verified in-app advertising revenue: “By Day 7, in-app advertising has generated 89% of its full Day 60 total, making it the fastest-maturing revenue stream across all monetization models.” For hypercasual it is more extreme still, 63 percent of day 60 ad revenue landing on day 1 alone.

Now hold that against the metric most UA teams optimise toward. GameAnalytics, measuring 16,262 games across calendar 2025, puts median day 30 retention at 0.68 to 0.79 percent.

So the standard funnel review optimises a signal that arrives after nine tenths of the revenue has been banked, using a metric that for the median game rounds to less than 1 percent.

That is not a retention problem. It is a sequencing problem: the funnel is being read in the wrong order, on the wrong clock.

The benchmark everyone quotes and the benchmark that exists

The industry quotes roughly 5 percent as the day 30 retention benchmark for mobile games. GameAnalytics measures 0.68 to 0.79 percent for the median game. Both are honest, and the gap between them is the single most common source of false comfort in a UA deck.

The difference is the weighting basis. GameAnalytics takes the median of 16,262 games, so a small title counts the same as a hit. MMP benchmarks are usually user-weighted across a client base that skews toward funded apps running paid UA. One answers “what does a typical game do”, the other answers “what does a typical funded install cohort do”. Mixing them in one slide produces a 7x error and a confident decision.

Only the top 1 percent of games reach 13 to 15 percent day 30 retention, in the same GameAnalytics dataset. Median day 1 sits around 22 percent, median day 7 just under 4 percent.

The largest drop-off is not where the chart puts it

Most funnel diagrams go impression, click, install, day 1, day 7, day 30, with a tidy percentage on each arrow. The real discontinuity sits in a gap the chart does not have: between the install you paid for and the first meaningful session.

AppsFlyer’s uninstall research puts it plainly: “over 1 in every 2 gaming apps are uninstalled within 30 days”, with the all-category Android uninstall rate at 46.1 percent in 2024 against 46.9 percent in 2023, and the majority of uninstalls arriving on day one. Note the limitation in the same source: iOS uninstall tracking is limited after iOS 15, so every uninstall benchmark in circulation is Android-only.

A day one uninstall is not a retention failure. It is a promise failure, and it is a creative and store-listing diagnostic, not a product one.

Revenue curves differ more by genre than retention does

The same AppsFlyer monetisation dataset shows how differently the clock runs by model:

1. Hypercasual banks 63 percent of its day 60 ad revenue on day 1.
2. Casino in-app purchase is the slowest curve in gaming, at 23 percent of its total by day 1, because players deliberate before spending.
3. Day 90 ARPU splits the same way: casino at 2.43 dollars, midcore at 2.13, casual at 1.34.

Platform then doubles the spread. Liftoff measured day 30 ROAS for casual games at 47 percent on iOS against 15 percent on Android, a gap wider than most genre-to-genre differences, on a dataset of 1.1 trillion impressions and 2.4 billion installs.

If your ROAS gate is the same number on both platforms, you are running two different businesses through one rule.

Five things worth doing with this

1. Compute, on your own cohorts, what share of day 60 revenue you have realised by day 3, day 7 and day 14, split by in-app advertising and in-app purchase. That number, not an industry benchmark, sets where your ROAS gate belongs.
2. If your advertising curve resembles the 89 percent by day 7 shape, move the gate to day 7 and compress your optimisation loop from a month to a week. If you run casino or midcore purchase revenue, the 23 percent by day 1 curve says the opposite, and killing cohorts early is the expensive mistake.
3. Split “install” into store install, first open and survived-to-second-session, and instrument uninstall separately on Android. The biggest leak usually sits between the event you pay for and the first real session.
4. Fix one measurement basis, game-weighted or user-weighted, and never mix them in a single report. Then segment by platform before genre, because the 47 against 15 percent ROAS gap is bigger than the genre spread.
5. Check the cohort size before you read a retention difference. At a true median day 30 of 0.7 percent, a 1,000 install cohort leaves about 7 retained users, so any creative comparison at that depth is noise wearing a decimal point.

Where the buying model changes the funnel

Every stage above is a place where an advertiser can pay for something that never became a user. On a CPM or CPI contract, that cost is entirely the advertiser’s: the impression is billed whether or not it reached a person, and the install is billed whether or not the app was ever opened.

For example, at SpinX the model is CPA with MMP-verified attribution, on owned and operated supply plus direct publisher relationships. The billable event sits after the part of the funnel where most of the leakage happens, and it is confirmed by the advertiser’s own measurement partner rather than by us. That does not make the funnel shorter. It moves the risk of the early stages to the side that controls the placement.

If your day 7 numbers look nothing like your install numbers, that gap is worth pricing before your next budget cycle, at spinx.io.