Both stores will A/B test your listing — for free.

Every install starts the same way: someone lands on your product page, looks for about a second, and decides. Apple and Google both ship a free, native tool for testing exactly that moment — and most indie listings have never run a single experiment. Here's how both tools actually work, the traps that make people quit, and the first test worth running.

The cheapest growth lever most listings never pull

Almost everything else in app growth has a bill attached. Ads charge you per view. Keyword rankings are fought over daily and can vanish in an algorithm shuffle. A conversion-rate improvement is different: it applies to every visitor from every source — search, ads, press, word of mouth — from the day you ship it, at no ongoing cost.

It's also the one place in ASO where you don't have to guess. The usual method — change the screenshots, squint at last week's installs — is polluted by seasonality, featuring, and ranking shifts. The stores' native tools split live traffic between versions at the same moment, so the comparison is actually clean. You get a real randomized experiment on real store visitors, run by the store itself.

Apple: Product Page Optimization

Product Page Optimization (PPO) lives in App Store Connect and lets you test app icons, screenshots, and app previews — up to three treatments against your original page, each localizable in any language your app supports. You choose a traffic proportion: Apple's own example is three treatments at 30% traffic, meaning each treatment gets 10% of visitors while 70% still see the original.

The constraints matter. You get one test at a time, it runs up to 90 days (or until you stop it), and you can't edit a test once it starts. Treatments go through App Review before the test can publish — screenshot and preview tests can be submitted on their own, but icon tests require every icon variant to ship inside the binary of your published app, so plan those around a release. One nice touch: if someone installs while seeing an icon treatment, that icon is what appears on their home screen.

Results show up in App Analytics as an estimated conversion lift against your baseline, with a confidence indicator that tells you whether the test needs more data. Treatments display to visitors on iOS and iPadOS 15 or later, and tests don't run on custom product pages.

Google Play: store listing experiments

Play Console's version is store listing experiments, and it's broader: you can test the icon, feature graphic, screenshots, short description, and full description — up to three variants against your current listing. You pick an audience between 1% and 50% of visitors; a classic single-variant test splits traffic 50/50.

Play gives you two success metrics, and the choice is a small strategy decision in itself: first-time installers, or retained installers — people who installed and kept the app for at least a day. Pick retained. A variant that wins installs but loses retention is usually a set of graphics that over-promises; you'd be optimizing for downloads that churn. Google also recommends running experiments for at least seven days, which smooths out the weekday–weekend cycle.

The four traps that kill tests

1. Timid changes on indie traffic. Statistical significance needs either a lot of visitors or a big difference between versions. A slightly warmer background color needs tens of thousands of page views to resolve; a completely different first-frame message can prove itself on far less. If your listing gets modest traffic, test bold swings — different benefit, different concept, different visual system — not shades of blue.

2. Peeking and stopping early. Day-two leads are mostly noise. Both tools tell you when they have enough data — Apple through its confidence indicator, Google through its seven-day minimum and result ranges. Calling a winner before that is just gambling with extra steps.

3. Testing frame five. Visitors decide on the icon and the first two screenshots — the frames that render in search results before your listing is even opened. A test that varies the last frame of the carousel is measuring something almost nobody sees.

4. Winning the wrong metric. The install isn't the goal; the user is. On Play, judge by retained installers. On Apple, sanity-check a winning treatment against your retention numbers in App Analytics before you celebrate — a lift built on over-promising pays itself back in one-star reviews.

The first test worth running

Start with the message on your first screenshot — it's the highest-traffic asset you control and the cheapest to vary. Keep the design system identical and test your current benefit headline against one or two genuinely different value propositions: if your notes app leads with "Capture ideas instantly," test "Find any note in seconds" and "Your notes, everywhere." That's a strategy question — which promise makes people install? — and the delta between real strategies is usually big enough to resolve on indie traffic.

When a winner emerges, promote it to the live listing, then use what you learned: the winning message should shape your subtitle, your description opener, even your ads. One test a quarter — Apple's one-at-a-time, 90-day format practically schedules it for you — compounds quietly into a listing that out-converts categories of apps that never test at all.

The real bottleneck is making the variants

The honest reason most developers skip testing isn't the console — it's that producing three polished, pixel-exact screenshot sets is a day of design work each. That's the part ShotCanvas removes: build the set once, then re-render variants in minutes — swap the headline, the palette, or the whole template — and export exact dimensions for every App Store and Google Play slot. The free tier covers your first sets, no credit card.