Paywall Timing: Show Early or at the Limit A/B Test

Sessions that end right at the usage limit, support requests that follow the limit message, or users who near the limit and never return all suggest free users walk away instead of upgrading. If the early prompt is the better option, users upgrade before the limit and activation does not drop; if the limit prompt wins, most early prompts simply get dismissed. If users leave because a message shows remaining usage incorrectly, fix that first, and settle any planned price change before you start, since timing can only be read with the price fixed.

Upgrade rate

Measure paid upgrades per eligible assigned user over the same follow-up period, including users who never see the prompt.

Also track

• Activation rateCompare core product use using the same start event and follow-up period.
• Prompt dismissal rateAmong users shown the modal, report dismissals as a diagnostic; keep the primary analysis on all assigned users.
• Seven-day retentionUse the same definition of active use on day seven after assignment.
• Support requestsTrack questions about the limit or upgrade prompt using the same visitor denominator.

• TimingCompare one predefined early trigger with the actual usage-limit trigger.
• EntitlementsKeep the free usage limit and feature access the same.
• ContextKeep usage information and offer wording accurate and unchanged.
• DismissalKeep the ability to dismiss and continue within existing entitlements consistent.
• EligibilityDefine which users are eligible to upgrade; report trial and paid cohorts separately if relevant.

1. 1Decide on paid upgrades per eligible assigned user; the prompt dismissal rate is measured only among users who saw the prompt, so treat it as a diagnostic. If upgrades rise while activation, seven-day retention or support requests cross the limits you set, do not roll it out.
2. 2The early prompt should stay a prompt that users can close and that lets them keep working within their current entitlements. If it stops work before the limit or reopens again and again, you are measuring pressure, not timing.
3. 3Only this pair of triggers and this offer are covered by the result. If relevant, review trial and paid cohorts separately; a different early trigger needs its own comparison.
4. 4More users may see the prompt in the early arm, but assignment should stay balanced, so when the difference is unclear, first check that arm sizes match the planned split (sample ratio mismatch). Because upgrades are a rare event, use the sample size calculator to find how many eligible users each arm needs at the current upgrade rate and the lift you care about. Keep the test running until each arm reaches that number and the follow-up window closes.