App Store product page optimization (PPO) is Apple's built-in A/B test for your listing: in App Store Connect you create up to three alternate versions of your icon, screenshots and app previews, choose how much traffic sees them, and Apple reports which version converts more visitors into downloads.
Google Play's equivalent is store listing experiments in Play Console, which tests up to two variants against your current listing and can include description text.
This guide covers what each store lets you test, how to read the results without fooling yourself, which ideas are worth testing first, and the mistakes that waste a 90-day test.
Apple product page optimization: what you can test, limits and traffic split
Apple's product page optimization tests three kinds of assets: the app icon, screenshots and app preview videos.
Text fields such as the name, subtitle and description can't be tested with PPO.
The rules
- Up to three treatments against your original page, in one test at a time per app.
- Up to 90 days per test. You can stop it earlier by hand.
- App Review applies. All new assets in a treatment must be approved before the test goes live.
- Alternate icons must ship in the binary. Every icon variant has to be included in the published app build. Tests without icon changes can be submitted on their own, without a new app version.
- iOS 15 and later only. Apple shows treatments to users on iOS 15 and iPadOS 15 or later, and the app must be in the Ready for Distribution state.
- Not for custom product pages. PPO isn't available for custom product pages or for Apple Watch and iMessage product pages.
How the traffic split works
You pick a share of App Store traffic for the test, and Apple divides it evenly between treatments.
Apple's own example: put 40% of traffic into a test with two treatments, and each treatment gets 20% while 60% still sees the original.
The same user keeps seeing the same treatment for the whole test.
Treatments appear on the product page, in search results and on the Today, Games and Apps tabs, so a new icon or first screenshot gets tested where most people actually decide.
Each treatment can be localized into all of your supported languages or only some.
Apple warns that more localizations can mean a longer wait to reach significance, because traffic is spread thinner.
Where to start in App Store Connect
- Open your app in App Store Connect and go to Product Page Optimization.
- Create a test, name it, pick the number of treatments and the traffic share.
- Choose the localizations to include.
- Upload the new screenshots, previews or icon (icon variants must already be in the live build).
- Submit for review, then start the test once approved.
You need the Account Holder, Admin, App Manager or Marketing role to create tests.
For asset specs, see our screenshot size reference.
Google Play store listing experiments
Google Play's store listing experiments run in Play Console.
They test up to two variants against your current listing, with visitors split equally across variants.
Two kinds of experiment
- Default graphics experiments test graphics in your default language: the icon, feature graphic and screenshots. You can run one at a time.
- Localized experiments target up to five languages and can also test your short and full descriptions. You can run up to five at once.
That means Play is the only one of the two stores where you can A/B test listing text natively.
If your subtitle-style messaging is the big unknown, test it on Play first and carry the lesson to iOS.
Settings that matter
Play Console lets you choose a target metric: unique user install clicks (users who installed and didn't already have the app on another device), unique user open clicks, or pre-registration clicks for unreleased apps.
Opens are the stricter measure, because they ignore people who install and never launch.
Under advanced settings you can change the audience percentage, the confidence level and the minimum detectable effect.
Play then estimates how long the experiment will need and how many installs or opens it takes to reach a result.
Experiments stop automatically after six months. Google recommends changing one asset at a time so you know what caused the result.
Apple vs Google at a glance
| Feature | Apple PPO | Google Play experiments |
|---|---|---|
| Variants vs original | Up to 3 | Up to 2 |
| Testable assets | Icon, screenshots, app previews | Icon, feature graphic, screenshots; descriptions in localized tests |
| Tests at once | 1 per app | 1 default graphics or up to 5 localized |
| Maximum length | 90 days | Stops after 6 months |
| Success metric | Conversion rate (downloads or pre-orders per unique impression) | Install clicks, open clicks or pre-registrations |
| Statistics | Bayesian, 90% confidence threshold | Adjustable confidence level and minimum detectable effect |
Sources: Apple PPO, App Store Connect help, App Analytics help, Play Console help.
If you want more control than either native tool gives, such as testing text on iOS or running tests on paid traffic, third-party options are compared in our app A/B testing tools roundup.
Reading results: confidence and duration
Both stores do the statistics for you, but they report it differently, and both are easy to misread.
Apple's result labels
Apple uses Bayesian methods and updates results daily.
A test shows up in App Analytics once at least five first-time downloads are attributed to it.
Each treatment then gets one of four labels:
- Performing Better: beats the baseline with at least 90% confidence.
- Performing Worse: loses to the baseline with at least 90% confidence.
- Collecting Data: not enough data yet.
- Likely to be Inconclusive: on current results, unlikely to reach 90% confidence within 90 days.
Alongside the label, Apple shows an estimated conversion rate, an estimated relative lift and a 90% credible interval.
Look at the interval, not just the headline lift.
A treatment showing +12% with an interval that runs from +1% to +23% is a real but uncertain win.
For the difference between Bayesian and classic significance, see our Bayesian vs frequentist guide.
Apple's conversion rate estimate blends test data with existing data, so early numbers can look more stable than they are.
Applying a treatment ends the test, and Apple recommends waiting until you reach 90% confidence first.
Google's result labels
Play shows which variant performed best and can recommend applying it.
If there isn't enough data yet it says "More data needed," and a "Draw" means the variants performed about the same.
A draw is still an answer: the change you tested doesn't matter much, so move on to a bigger idea.
How long to run
Apple says plainly that larger improvements, such as a 30% lift, need fewer weeks of data than small ones, such as 1%.
Store traffic converts at low rates, so small effects take a long time to detect.
Apple's own figure for the average conversion rate on default product pages is 1.6%; check where you stand with our app store conversion rate benchmarks.
Before you start, put your daily product page views and current conversion rate into our sample size calculator.
If the answer is "four months to detect a 5% lift," you need a bolder variant, not a longer wait.
Whatever the math says, run at least one full week so weekday and weekend visitors are both counted.
Test ideas ranked by measured lift
A note on what we have: our database doesn't yet hold store-listing tests with clean, published numbers.
When we add them, they'll appear on the store listing tests page.
What we do have are measured tests of the same ideas one screen later, on paywalls and onboarding, plus ecommerce product-page tests on image order and proof.
The lifts below come from those surfaces, so treat them as a guide to which listing ideas deserve a slot, not as the lift you'll see in PPO.
| Listing idea | Closest measured test | Reported result |
|---|---|---|
| Localized screenshot captions | Daily Horoscope: paywalls localized by country | +183% trial start rate (bundled, vendor-reported) |
| Preview video or motion in first frame | Stardust: video background paywall | +67% paywall conversion; +12.5% proceeds per user (vendor-reported) |
| Screenshots that show the real product | Unchaind: paywall that previews premium features | Over +50% conversion; +64% estimated ARPU (vendor-reported) |
| Lead with the outcome the user gets | Clear 30: paywall that previews the program | 20% to 30% conversion (+50%), before-after |
| Localized text for big non-English markets | Photoroom: localized upsell paywalls | +50% upsell screen conversion (vendor-reported) |
| Rating or proof in an early frame | Ecommerce: star rating and review count above the title | +8.14% conversion |
| Change which screenshot leads | Ecommerce: second photo moved to first in the gallery | Revenue per visitor £3.32 to £3.60 |
1. Localize the first screenshots
Daily Horoscope reported a +183% trial start rate from paywalls localized by country, and Photoroom saw +50% upsell screen conversion from localized upsell paywalls.
Both are vendor-reported. On the store, this means translated captions and culturally matched imagery in your biggest non-English markets.
Play's localized experiments and Apple's per-locale treatments are built for exactly this.
2. Test a preview video, both ways
Stardust's video background paywall lifted paywall conversion 67%, and Kovalee found that a 10-second video preview before PepTalk's paywall beat the control in a multivariate test.
But video isn't automatically better: an ecommerce store that swapped its hero video for a static image with a benefit callout saw +52.67% revenue per visitor.
On iOS, previews autoplay muted, so the first seconds must work without sound.
Our preview video guide covers what to put in them.
3. Show the product, not a slogan
Unchaind's paywall that previewed premium features got over +50% conversion, and Clear 30's program preview took conversion from 20% to 30%.
Screenshots do the same job before the install.
Test a frame that shows a real screen with a one-line benefit against a branded, illustration-led frame. See our screenshot guide for patterns.
4. Reorder before you redesign
The cheapest test is moving your best frame to first place.
In ecommerce, simply moving the second product photo to first raised revenue per visitor from £3.32 to £3.60, and adding benefit images to a product gallery was worth a reported +$10K a month.
Apple shows the first one to three screenshots in search results, so order matters more on a store than on most web pages.
5. Test the icon last, and boldly
The icon shows everywhere, so it has the widest reach, but it needs a new binary on iOS and small icon tweaks rarely reach confidence.
Test a clearly different color or symbol, not a shade change. Our app icon guide has examples.
Common mistakes
Optimizing installs and ignoring revenue
Both stores measure installs or opens, not trials or revenue.
A listing that promises more than the app delivers can win the test and still lose money.
BoldVoice saw this inside the app: a hard paywall with no close button brought more trial starts, but refunds rose by more than the gain.
After applying a winner, compare trial and paid conversion for the weeks before and after.
Testing tiny changes
A slightly different gradient on screenshot four won't reach 90% confidence on typical store traffic.
Make each treatment a different idea: a new lead message, a new first frame, a video versus none.
Changing the page mid-test
Apple warns that releasing a new app version during a test can affect results if the version changes assets you're testing.
Paid campaigns can also skew a test: a burst of ad traffic brings visitors who behave differently from organic browsers.
Plan releases and big campaigns around your test window.
Calling it early
Day-three results swing a lot. Wait for Apple's Performing Better label or Play's recommendation, and run at least a full week.
If Apple says "Likely to be Inconclusive," stop and test something bolder.
Assuming every test should win
Plenty of reasonable ideas lose. Stardust tried a yearly-only paywall and lost on both conversion and proceeds per user.
Across experiments in general, most variants don't beat the control; see our app A/B test win rate research.
A loss that stops you shipping a bad page is worth having.
Confusing PPO with custom product pages
PPO tests your default page against alternates for a random split of visitors.
Custom product pages are separate listings you send specific traffic to, such as an ad campaign, and PPO can't run on them.
Use PPO to find your best default, then use custom pages to match different audiences.
For the wider picture of keywords plus conversion, see our ASO guide.