Plan a screenshot A/B test you can interpret
Define one screenshot hypothesis, prepare controlled variants, and record platform results without treating an early percentage change as a proven win.
A screenshot experiment compares alternatives for your app's audience. Skir can help prepare the static artwork; Apple or Google runs the store experiment and reports its results. A more attractive design, an exported file or a vendor case study is not evidence that your own conversion rate improved.
Start with one falsifiable idea
For a fictional planner, an initial hypothesis could be: ‘Showing today's task list first will help visitors understand the core use case better than opening with the weekly calendar.’ Keep the headline style, color palette and remaining screenshot order consistent. The main change is the leading screen and its matching explanation.
Write down what you will change, which audience will see it, which platform metric you will use and what would count as an inconclusive result. If you change the icon, screenshot story, pricing and acquisition campaign together, you lose a clear explanation for any difference.
Choose the experiment in the store console
Apple supports an original page and up to three treatments, with tests running for at most 90 days. Google recommends changing a single asset and covering at least a full week. These are platform capabilities and guidance, not a promise that your app will have enough traffic to reach a decision. Use the current console's setup and result definitions.
| Platform | Where the experiment runs | What to review |
|---|---|---|
| Apple App Store | Product page optimization in App Store Connect | Treatment allocation, localization, review status and reported confidence |
| Google Play | Store listing experiments in Play Console | Selected audience, asset variation, acquisition and retention results |
Prepare comparable artwork
- Archive the original images and record their order. Use clear local filenames such as control-01 and treatment-a-01.
- Prepare the alternative in Skir from the same template and the same release-build captures. Keep platform and export dimensions identical.
- Review the images side by side at store-preview size. Check that only the intended story element changed.
- Export and inspect each file before uploading it to the corresponding treatment. Confirm the console preview and image order before starting the experiment.
Read the result with its denominator
Record the date range, traffic allocation, eligible audience, conversion definition, sample counts and uncertainty reported by the platform. A percentage without its audience or denominator cannot support a useful decision. Avoid repeatedly stopping a test as soon as the headline number looks favorable; decide your review rule before launch.
For arithmetic illustration only: moving from a 10% conversion rate to 11% is a one-percentage-point absolute change and a 10% relative change. Those numbers do not establish statistical confidence. They are not Skir customer results, and two periods with different traffic sources are not automatically a controlled experiment.
Keep an inconclusive result useful
If the console needs more data, record that status. Low traffic, a small effect or a changing audience can prevent a clear answer. Keep the original unless you have a documented reason to choose another treatment; do not rename ‘inconclusive’ to ‘no difference.’ Save the actual report, the tested files and a short explanation of your next decision.
The worksheet below has room for the hypothesis, asset changes, audience, stopping rule and outcome. It deliberately contains no conversion forecast. Skir's site analytics measures consented website activity; it does not replace App Store Connect or Play Console experiment data.
Download the editable Markdown worksheet. Open it in any text editor and fill it with your own release evidence.
Sources and related guides
Platform documentation checked September 10, 2026. The workflow advice is Skir’s editorial guidance; it does not certify a particular app or predict conversion results.