The middle of the
app market disappeared.

The top quartile of subscription apps grew monthly recurring revenue by 80% or more in the year to 2026. The bottom quartile lost more than 33%. The gap between them is 113 points.¹

The report that publishes those figures - 115,000 apps, more than $16 billion in revenue - describes what sits between them as a middle ground that is evaporating.¹

Figure 01

The middle went missing

Year-on-year monthly recurring revenue growth, subscription apps, by quartile.

Source: RevenueCat, State of Subscription Apps 2026, 19 March 2026. The top and bottom figures are thresholds, not averages - 80% is the floor of the top quartile and −33% the ceiling of the bottom, so the true spread is wider than 113 points. The median figure needs care: the report presents median growth as a band of roughly 5–17% that varies by the developer's headquarters country, and secondary coverage reports a global median near 5.3%. We plot the global figure and flag that it is one point drawn from a geographic distribution.

Alongside the growth spread sits a revenue spread that has been widening for two years: the top 5% of apps earn 400 times what the bottom quartile earns, up from 200 times in 2024.¹

And into that market, more than 14,000 new apps arrive every month.¹ Feature 46 of this edition records that total installs across both stores grew 0.8% over the same period.

Supply expanding at 14,000 products a month, demand for installs expanding at under one per cent a year, and a revenue distribution stretching from 200-to-1 to 400-to-1 in twenty-four months.

The industry is moving away from its own evidence

The most interesting number in the report is not about winners. It is about a decision the whole market is getting wrong in the same direction, on purpose.

Free trials of 17 to 32 days convert to paid at a median of 42.5%. Trials of under four days convert at 25.5% - meaning the long trial converts roughly 70% better.¹

Over the same period, the share of apps using trials of under four days rose from 42.1% to 46.5%.¹

Figure 02

The conversion the market is walking away from

Median trial-to-paid conversion by trial duration.

Source: RevenueCat, State of Subscription Apps 2026. This is observational, not experimental. Apps that choose long trials differ from apps that choose short ones in ways this comparison does not control for - category, price point, product complexity and how confident the team is in its onboarding. A team switching trial length should not expect to move from 25.5% to 42.5% by changing one setting.

The report's own explanation, quoted from its authors, is that developers run three-day trials for cash flow - money in three days rather than thirty - and to get conversion data back fast enough to keep running onboarding and paywall experiments.¹

That is not ignorance. It is a working-capital decision that costs conversion, made by teams who need the cash sooner than they need the higher rate. Which is exactly what you would expect from a market where the bottom quartile is shrinking 33% a year.

A three-day trial is a one-session trial

55.4% of all cancellations on three-day trials happen on day zero. By the end of day one, 84% of them have happened. A year earlier the day-zero figure was around 51%.¹

So the modal user subscribes to get past the paywall, evaluates the product in a single session, and turns off auto-renew before closing the app. The trial length is nominal. The real evaluation window is the first session, and it is getting shorter.

This reframes the paywall argument that sits next to it in the same report. Apps with hard paywalls show a median day-35 trial-to-paid conversion of 10.7% against 2.1% for freemium, and generate $3.09 revenue per install at day 60 against $0.38.¹

But retention a year later is 27% for hard-paywall apps and 28% for freemium - a difference the report itself calls negligible.¹ And hard-paywall conversion fell from 12.1% the year before, while freemium held flat at 2.1%.

Annual subscriptions stopped meaning a year

72% of annual subscribers cancelled auto-renewal during Year 1, up from around 56% the previous year. 35% of all annual cancellations occur in Month 1. Cancellations then fall to 3–10% a month through the middle of the year before spiking again ahead of renewal.¹

The user is not buying twelve months of a relationship. They are buying one year of a product and closing the loop immediately so it does not recur. The developer's Year 2 revenue is decided in Week 1, and in most cases decided against them.

A third of Android churn is a billing bug

31% of subscription cancellations on Google Play are involuntary billing failures - expired cards, declines, retries that did not recover. On the App Store the figure is 14%.¹

Both moved: Google Play worsened from 28.2% the previous year, the App Store improved from 15.1%.¹

Nearly a third of Android churn, then, is not a product judgement. It is a payment that failed and was not recovered. That is the cheapest revenue in this entire feature, and it is recovered by dunning logic and grace periods rather than by acquiring anyone.

What to do about it

Fix billing recovery before you fix anything else. If you have a meaningful Android base, roughly a third of your churn is involuntary. Retry logic and grace periods recover revenue from users who never decided to leave, at no acquisition cost.

Know why your trial is three days. If the answer is cash flow or experiment velocity, that is a defensible trade and you should write down what it costs you in conversion. If the answer is that everyone does it, the data says everyone is wrong by roughly 70%.

Design the first session, not the trial. With 55% of three-day trial cancellations landing on day zero, the trial length is a billing setting and the first session is the actual product decision.

Treat Week 1 of an annual subscription as the renewal campaign. Month 1 accounts for 35% of annual cancellations. The Year 2 conversation happens in Week 1 whether you participate in it or not.

Do not read the paywall comparison as causal. Hard paywalls collect earlier. The one-point retention difference at twelve months says they do not obviously collect from better people. Pull revenue forward if you need it forward - but book it as a cash-flow decision, not a quality one.