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

Page: https://alidemirbas.com.tr/en/lab/ab-test-library/tests/paywall-timing-show-early-or-at-the-limit-ab-test

## Hypothesis

An early prompt may introduce the offer while users keep working, while a prompt at the limit arrives when the need is clearest but may interrupt active work. If users make up their minds ahead of time, the early prompt could bring the higher upgrade rate; if they only feel the need at the limit, the limit prompt could.

## When to run this 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.

## Primary metric

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

## Other metrics

- **Activation rate**: Compare core product use using the same start event and follow-up period.
- **Prompt dismissal rate**: Among users shown the modal, report dismissals as a diagnostic; keep the primary analysis on all assigned users.
- **Seven-day retention**: Use the same definition of active use on day seven after assignment.
- **Support requests**: Track questions about the limit or upgrade prompt using the same visitor denominator.

## What to check before and during the test

- **Timing**: Compare one predefined early trigger with the actual usage-limit trigger.
- **Entitlements**: Keep the free usage limit and feature access the same.
- **Context**: Keep usage information and offer wording accurate and unchanged.
- **Dismissal**: Keep the ability to dismiss and continue within existing entitlements consistent.
- **Eligibility**: Define which users are eligible to upgrade; report trial and paid cohorts separately if relevant.

## What to watch when reading the result

- Decide 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.
- The 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.
- Only 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.
- More 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](https://alidemirbas.com.tr/en/calculators/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.

## Frequently Asked Questions

### Does the early-warning group lose any usage rights?

No. Keep free limits, feature access, usage information and offer copy identical. Only warning timing changes. Do not turn the early warning into a premature paywall.

### Why is modal dismissal not the decision metric?

Only exposed users can dismiss it, and exposure populations can differ. Report dismissal diagnostically. Use paid upgrades per eligible assigned user as the primary outcome.

### How can I tell whether the warning harms product usage?

Track activation, retention, core usage and support contacts over a common window. Do not roll out an upgrade lift that exceeds predefined usage-harm limits.

### Can I change the price in the same test?

No. Keep price fixed to interpret timing. Allow dismissal, avoid repeated reopening and show remaining usage accurately. Review relevant trial and paid cohorts separately.