Paywall A/B testing for mobile apps

To A/B test a mobile app paywall, randomly assign new installs to two or more paywall versions and judge them on revenue per install over a fixed window, usually 30 to 60 days, after refunds. Trial starts and paywall conversion are tempting because they arrive fast, but they reward versions that push people toward the cheapest option or a free trial that never converts.

Test the things that change what people pay first: which plans you offer, which one is pre-selected, whether there is a trial, and when the paywall appears. Every version has to follow the store rules. Apple requires the billed amount to be the most prominent price on the screen, and both Apple and Google require clear trial and renewal terms. A version that only wins by hiding the price is not a winner you can keep.

What to measure

Revenue per install is the main number. It is the money collected from everyone who installed and was assigned to a version, within the window, minus refunds, divided by installs. Store fees apply at the same rate to both versions during a test, so you can compare gross revenue within a test. Use net proceeds if you compare across tests or against ad spend.

Secondary numbers help you understand why a version won:

• Paywall view rate, since a paywall later in the app is seen by fewer people.
• Trial start rate and trial-to-paid rate.
• Plan mix between weekly, monthly and annual.
• Refund rate in the first billing period.
• Renewals, read again at 90 days and after the first annual renewal if you can wait.

What to test, in order

The order follows how much each change can move revenue. RevenueCat's State of Subscription Apps 2026 found that apps with a hard paywall had a median download-to-paid rate of 10.7% by day 35, against 2.1% for freemium apps. Differences between apps explain part of that, but it shows how much the paywall model can matter compared with a button color.

Weekly vs annual plans

Weekly and annual plans make money on different schedules. A weekly plan charges often and early. An annual plan collects a year up front, then most people don't renew.

RevenueCat's retention study of 10,000+ apps found a median of 3% of weekly subscribers, 11% of monthly and 28% of annual still subscribed after a year. The plan mix also varies by category. In the 2026 report, 82% of gaming subscriptions sold were weekly, while productivity apps sold 77% annual.

Neither fact tells you which will earn more for your app. Weekly plans can win on revenue per install when people use the app heavily for a short period. An annual plan makes more sense for an app people use for months. Test the lineup, and judge it over a window long enough to see weekly churn.

A worked example

This is a made-up example. An app offers annual at $39.99 with a 3-day trial and weekly at $4.99 with no trial. Version A pre-selects annual. Version B pre-selects weekly. Each gets 20,000 installs.

Here is the maths. A collects 360 × $39.99 = $14,396 from annual plus 400 × $4.99 = $1,996 from weekly, for $16,392. B collects 90 × $39.99 = $3,599 plus 2,800 × $4.99 = $13,972, for $17,571. Divide by 20,000 installs.

A team judging on trial starts would pick A by four to one. On 60-day revenue per install, B is 7% ahead. The gap after a year depends on how long weekly buyers keep paying. If they average 10 weekly payments in the first year, B earns $3,599 + 7,000 × $4.99 = $38,529, or $1.93 per install, against A's $14,396 + 1,000 × $4.99 = $19,386, or $0.97. If weekly buyers leave faster, the gap shrinks. Pull the real number from your own weekly retention curve before you project anything.

A 7% difference at 60 days also needs a lot of installs to confirm. Use the sample size calculator with your own revenue per install before you launch, and read the sample size guide if the result surprises you.

App store rules to follow

These rules apply to every version you test. They are summarized in general terms here, so read the current text before you build.

• Apple's App Store Review Guidelines, section 3.1.2, require auto-renewing subscriptions to last at least seven days and to give ongoing value. Apps that trick users into subscribing under false pretenses or use bait-and-switch practices can be removed. Before asking someone to subscribe, the app should clearly describe what they get for the price.
• Apple's subscription guidance says the billed amount must be the most prominent pricing element, with any monthly equivalent or savings shown smaller. For a free trial, the purchase flow should show how long the trial lasts and the price billed after it. The sign-up screen needs the full renewal price and a way to restore purchases.
• Google Play's subscriptions policy requires clear disclosure of the offer terms, cost, billing frequency and whether a subscription is needed to use the app. It says not to name an auto-renewing subscription "Free trial", and apps must offer an easy online way to cancel.

In practice, this rules out tests like "$0.99 a week" in large type with the annual charge hidden, or a close button that doesn't close. Test real choices presented clearly.

How to run a paywall test

1. Assign the version at install or first launch, by app user ID, and keep it for the life of the test.
2. Serve paywall versions from a remote config or paywall tool so you can change them without an app release.
3. Split results by platform. iOS and Android users can behave differently, and the store rules differ.
4. Check the split is even with the sample ratio mismatch checker. Paywall tests break when one version fails to load.
5. Run until each version has the installs your sample size calls for, then wait for the revenue window to close for the last install.
6. Read renewals again at 90 days, and after a year for annual plans.

Outtest connects to RevenueCat read-only for revenue data. If you also run tests on your website, it holds winners to the same kind of bar described here. By default, a winner needs at least 7 days, a 90% chance of beating the original and a revenue per visitor lift of at least 10%.

Tests to try, and what to measure

1. Add a weekly plan next to annual. Measure revenue per install at 60 days and plan mix.
2. Pre-select annual vs weekly. Measure revenue per install at 60 days and projected 12-month revenue.
3. No trial vs a 3-day trial vs a 7-day trial. Measure revenue per install in a window covering the longest trial plus one billing cycle. The free trial guide covers trial length.
4. Paywall at the end of onboarding vs after the first completed task. Measure revenue per install and day-7 retention. The onboarding guide covers the first session.
5. Hard paywall vs a free tier. Measure revenue per install at 60 days.
6. One plan shown vs a choice of three. Measure revenue per install and plan mix.
7. An outcome headline ("Sleep better in 7 nights") vs a feature list. Measure revenue per install. Only use claims you can support.
8. Annual discount shown as a percentage vs a weekly equivalent in smaller type under the annual price. Measure revenue per install and refunds.

For how the plan period changes revenue over time, see annual vs monthly pricing.