How to measure paywall conversion

Write the fraction in full.

“Our conversion rate is 8%” leaves most of the useful information out. Eight percent of installs, paywall viewers, or trial starters? Over one day or a month? For a first purchase or a renewal?

For an initial paywall test, one useful starting population is first-time, purchase-eligible users who saw a specific variant. Count each person once in that cohort. Separate returning subscribers and restores from new purchases, and decide how users exposed to multiple variants will be handled.

Record the window too. For example: trial starts within seven days of first exposure, followed by successful first payments from those trials. Wait for every included trial to have had its full conversion opportunity, including the billing follow-up window you chose.

One example produces three valid rates.

Imagine a trial-only offer shown to 1,000 eligible people. Within the chosen start window, 200 begin a trial. After those trials have matured, 80 of the same 200 make their first successful payment.

The first rate helps you inspect the offer and willingness to try. The second helps you inspect what happens during the trial. The third connects the two. In this simplified funnel, 20% × 40% = 8%. Direct purchases, different windows, or different populations would break that simple relationship.

If trial starts rise while trial-to-paid falls, the new paywall may be attracting more curiosity without more paying users. Compare the final paid outcome, then check refunds and net value per acquired user.

The same label can hide a different denominator.

RevenueCat’s Paywall Conversion chart reports trial conversions relative to paywall viewers. Its separate Trial Conversion Rate chart reports converted trials relative to trial starts. Those percentages answer different questions.

The paywall chart also uses its own first-impression cohort rules and initial conversion window. Do not assume that a custom seven-day funnel matches the dashboard simply because both say “conversion.” Read the chart definition before comparing results.

Keep a short metric contract next to the report: numerator, denominator, eligibility, cohort start, observation window, and refund treatment. That is more useful than a screenshot with a percentage and no explanation.

Use the drop-off to choose the next question.

If people reach the paywall but rarely start, investigate whether the benefit, price, or timing is unclear. If they start but rarely pay, investigate whether the product delivers during the trial. These are hypotheses to test, not conclusions the funnel proves.

Change a meaningful variable, use consistent assignment, and compare cohorts with enough time to mature. A few early payments can swing a small sample dramatically. Decide the primary outcome and acceptable uncertainty before interpreting a winner.

Use external benchmarks for context. Use a well-defined comparison within your own app to decide whether a change helped.

Trial length changes what you need to measure.

RevenueCat’s analysis published on 28 September 2026 compares trial conversion and renewal across more than 17,000 subscription apps on the App Store and Google Play. Longer-trial groups often show higher renewal, but conversion does not improve consistently across plans, categories, and subscriber regions. This is an observational comparison, not a randomized test of changing an app’s trial.

The study uses August 2025–July 2026 trial starts for conversion and renewal dates for retention; its combined trial-to-first-renewal measure includes earlier annual cohorts. Rates are app-level medians. Trial conversion uses trial starters as its denominator; first renewal is conditional on a previous payment. Regional and category breakdowns are context, not a country-specific target for your app.

For a small team, the useful next question is whether extra trial time changes the outcome for the same eligible audience. A renewal percentage among people who already paid cannot answer that by itself.

Test trial length without moving the goalposts.

Here is our suggested experiment brief. Use it when trial duration is the uncertainty you want to resolve; it is not a promise that extending a trial will help.

1. Choose one eligible audience. Start with one subscription plan, platform, and market. Confirm store offer eligibility. Randomly assign new eligible users before showing either offer, and keep assignment stable across return visits.
2. Keep the comparison interpretable. Hold price, access, onboarding, and acquisition settings steady. Change the trial duration and the wording needed to describe it accurately. Check that the store purchase sheet matches each displayed offer.
3. Fix the clock before launch. Define an enrollment period, a trial-start window, and one observation horizon from assignment. Allow the longest trial, the paid period, and your billing follow-up window to finish if first renewal is an outcome. Wait for every included user to reach that horizon.
4. Keep everyone assigned in the decision metric. Compare net proceeds minus variable service costs per assigned eligible user at that horizon, including people who never start a trial. Inspect trial starts, first payments, and renewals separately to understand the result.
5. Set a decision rule. Choose the smallest worthwhile improvement, required sample, review date, and acceptable uncertainty before reading results. Track refunds, support issues, and usage costs. If traffic cannot resolve the difference, record an inconclusive test.

In a hypothetical comparison of seven-day and fourteen-day trials starting at assignment, a day-ten payment check gives only the shorter trial time to charge. That apparent lead is a timing artifact. Equal cohort age is necessary, but the chosen age must also let both offers reach the outcome being compared.

RevenueCat’s trial conversion definition still counts a trial that converted and was later refunded. Keep conversion counts separate from refund-adjusted proceeds. Read payback alongside the result: an improvement at one horizon does not establish lifetime value or guarantee cash recovery.