Paywall A/B test calculator
Compare paywall variants on what they earn per view. Enter the purchases of each plan, and the calculator weighs a yearly sale against a weekly one for you.
Ship Yearly plan first
Yearly plan first has a 96% chance of earning the most per view. If it is not the best, you would lose about $0.0011 per view on average.
- Current paywall converts more people, but Yearly plan first earns more per view. Buyers are choosing different plans, so judge the test on revenue, not conversion.
| Variant | Revenue per view | Chance it earns most | Conversion |
|---|---|---|---|
| Current paywall | $0.613Control$0.524 to $0.717 | 4% | 3.60%303 of 8,420 |
| Yearly plan first | $0.749+21.8% vs control$0.643 to $0.869 | 96% | 2.99%251 of 8,388 |
How much more each variant earns per view
Each bar is a range of possible lifts, sized by how often it came up in 20,000 simulations. Bars left of 0% mean the variant earns less than Current paywall.
Revenue uses prices after the 15% store fee. Bayesian model with a flat prior on conversion. Frequentist checks for reference: Yearly plan first: p = 0.074 on revenue, 0.028 on conversion. Traffic split check passed (p = 0.805). How this works
Questions people ask
Why judge a paywall test on revenue per view?
A paywall that nudges people toward the weekly plan can win on conversion and still lose money, and the reverse happens just as often. Revenue per view combines how many people buy with which plan they buy, so it answers the question you actually care about.
What counts as a view?
Use whatever your paywall tool counts as an exposure: one person who saw the paywall at least once during the test. Count each person once. If you only have installs, use installs, as long as every variant is counted the same way.
What value should I enter for each plan?
The price works for a quick read. If your variants push plans that renew very differently, such as weekly against yearly, enter each plan's expected revenue per subscriber over 12 months instead. The LTV calculator gives you that number.
What do the chance and the range mean?
Chance it earns most is the probability, given your data and a neutral starting assumption, that the variant's true revenue per view is the highest. The range under each number holds the true value with 95% probability.
When should I stop the test?
Decide the length up front with the sample size calculator and stop on that day. We recommend shipping when a variant has at least a 95% chance of earning the most and the expected cost of being wrong is under 1% of revenue per view.
What about free trials?
If most of your purchases start as trials, wait until the trials in the test have ended and count conversions to paid. Or test on trial starts and track the trial-to-paid rate separately for each variant.
Get your test results in your inbox every Monday
We're building a weekly report for subscription apps. You upload your RevenueCat Experiments export, and each Monday you get, for every running paywall test:
- The chance the new paywall earns more per view, with a likely range
- A warning when the traffic split is off and the result can't be trusted
- About how many more days the test needs
- One concrete test to run next
RevenueCat already emails each variant's conversion and realized LTV against control. This report tells you whether the difference is real and what to do about it.
Planned price: $49 a month, or $99 for teams.