Where these numbers come from.
Every figure below comes from RevenueCat’s State of Subscription Apps 2026, which analysed more than 115,000 apps and around $16 billion of tracked revenue. It is the largest openly published dataset on app subscription behaviour, and it is their research, not ours — the full report is linked at the end.
What follows is what these numbers imply for anyone building or fixing a subscription product.
The AI paradox: converts well, churns fast.
AI app launches have risen sevenfold since 2022, and roughly one in four new apps is now AI-powered. On the money, they look excellent:
| Measure | AI apps | Non-AI apps |
|---|---|---|
| Trial-to-paid conversion | 8.5% | 5.6% |
| Revenue per payer | $30.16 | $21.37 |
| Annual retention | 21.1% | 30.7% |
So AI apps convert about 52% better and earn about 41% more per payer — then churn roughly 30% faster.
The reading that fits: novelty sells the first subscription, and habit sells the twelfth. An AI feature is a superb acquisition mechanism and a poor retention mechanism on its own. If the product does not become part of someone’s routine, the improved conversion just fills a leakier bucket faster.
Incumbency is the strongest force in the data.
Apps launched before 2020 generate 69% of subscription revenue. Apps launched from 2025 onwards account for about 3%.
And the spread within the market is widening: the top quartile is growing around 80% year on year while the bottom quartile shrinks by about a third.
This is the number to take to a planning meeting. A new subscription app is entering a market where distribution, brand and habit are already allocated. That does not make it impossible — it makes “we will win on features” an insufficient plan.
Hard paywalls beat freemium, and it is not close.
Hard paywalls convert at 10.7% against 2.1% for freemium — about five times better — and produce eight to nine times the revenue per install in the first 60 days.
The objection is always that a hard paywall buys worse subscribers. The data does not support it: one-year retention is 28% for freemium and 27% for hard paywall. Statistically the same.
So the trade-off most teams believe they are making — conversion now versus loyalty later — largely is not real. What a hard paywall genuinely costs is top-of-funnel volume and word of mouth, which is a different argument and worth having explicitly.
Longer trials convert better.
Trials of 17 to 32 days convert at 42.5% against 25.5% for four-day trials — around 70% better.
The mechanism shows up in day-zero cancellations: 55.4% of three-day trials are cancelled on the first day, against 31.1% of 30-day trials. A short trial triggers immediate defensive cancellation. A long one removes the urgency to act, and the user forgets to be defensive — which is worth being honest with yourself about as a design choice.
The Android billing leak.
On Google Play, 32.2% of cancellations are involuntary — failed billing rather than a decision. On the App Store the figure is 15.2%.
Roughly a third of Android churn is therefore an infrastructure problem wearing a product problem’s clothes. Retry logic, grace periods, card-update prompts and dunning are unglamorous and among the highest-return work available to an Android subscription business.
What the data says to do.
- Treat day zero as the business. Over half of conversions happen on install day, and roughly 80% of trial starts. Onboarding is not a preamble to the product; it is where the revenue is decided.
- Push annual plans. They produce about twice the revenue per install of monthly and about five times that of weekly.
- Keep the paywall simple. Around 60% of top performers in Health & Fitness use a two-plan paywall.
- Avoid weekly plans. They show 1–2% annual retention. That is effectively total churn with extra billing.
- Be careful pricing low. Cheaper does not mean stickier — higher-priced yearly plans retained at 23% against 36% for lower-priced ones, so price is doing selection work as well as revenue work.
Common questions.
Should a new app use a hard paywall?
The conversion and revenue-per-install case is strong, and the usual retention objection is not supported by the data. The real cost is reach: fewer installs convert to trials at all, so word of mouth and organic growth suffer. If your growth model depends on volume, test before committing.
Why do AI apps churn so much faster?
The most consistent explanation is that AI features drive trial by novelty rather than by habit. Products that survive the first renewal tend to be the ones where the AI sits inside a workflow the user was doing anyway, rather than being the reason to open the app.
How long should a free trial be?
The data favours somewhere between 17 and 32 days over three or four. Short trials produce a spike of same-day defensive cancellations. Longer trials cost you nothing in unit terms and convert markedly better.
Is it too late to launch a subscription app?
No, but the shape of a winning plan has changed. Pre-2020 apps hold 69% of revenue, so a new entrant needs a distribution advantage or a specific underserved audience. Features alone do not beat habit.
What is the quickest win for an existing app?
On Android, billing recovery — retries, grace periods and card-update prompts — because a third of that churn was never a decision. After that, onboarding on install day, since that is where most conversions happen.
What this means for a buyer.
Two things in this data matter more than the rest: the first day decides most of your revenue, and an AI feature is an acquisition tool rather than a retention one. Both are product decisions, not marketing ones. Simam Digital builds subscription products and the onboarding, paywall and billing infrastructure around them — see why to prototype before the full build.
Sources and further reading
A version of this article was first published in Tech Alchemy, the Simam Digital newsletter on LinkedIn.
