How to A/B Test a Game's Store Page in 2026: Google Play Store Listing Experiments and Custom Store Listings vs Apple Product Page Optimization and Custom Product Pages

The store page is the last screen that every player crosses, whether they arrived from a paid ad or from search. It decides how many of the visitors you paid for actually tap install, so a few percentage points of store conversion move your effective CPI with them. According to AppTweak, an ASO tool vendor that aggregated 2025 US store data, the average share of page visitors who downloaded was 16.15% on Google Play and 8.56% on the App Store, and the lowest Google Play category was strategy games at 6.6%. Games, and midcore genres in particular, are among the categories that lose the most visitors on the store page.

Both stores give you free tools to experiment with that rate. Google Play has store listing experiments and custom store listings. The App Store has Product Page Optimization (PPO) and Custom Product Pages (CPP). The names are similar, but what you can test, how a winner is declared, and how each tool connects to advertising differ quite a bit. Based on both stores' official documentation as of October 2026, this post compares the four tools side by side and covers the problems game studios actually hit when designing a test: sample size, novelty effects, and shifts in traffic mix.

We covered the principles of ASO in ASO for Mobile Games: The Most Efficient Way to Drive Installs Without Paid Ads, and what to expect when you outsource the work in What Do App Store Optimization Services Actually Deliver for Games? Agency vs Tool vs In-House, 2026 Pricing, RFP Questions, and a 90-Day Pilot. This post focuses on running the tests themselves.

What can a store A/B test change, and what can it not?

Store-native tests mainly change graphic assets (icon, screenshots, feature graphic, video), and on Google Play, localized experiments can include descriptions as well. Elements tied directly to search ranking, such as the app name, category, and rating, are not testable.

This distinction matters because a store test only measures the stretch from "shown to a visitor" to "installed." Part of the effect of a new icon on search result taps is captured, but an increase in impressions caused by better keyword rankings cannot be seen in a test. Put the other way, the winner of a store test is the page that turns the same visitors into more installs, not the page that brings in more visitors.

Align definitions before you set a baseline. In its custom product page guidance, Apple gives 1.6% as the average conversion rate of default product pages, which is far from AppTweak's App Store average of 8.56% (downloads divided by page views). Apple does not disclose the period or formula behind its figure, while AppTweak calculated downloads over page views. Because App Store users can download straight from search results without opening the page, the result changes several-fold depending on the denominator. Use outside numbers only for direction, and judge tests by comparing the same metric inside your own console.

How do Google Play store listing experiments work?

Google Play store listing experiments show up to two variants against your current listing to a share of visitors and declare a winner based on installs, opens, or pre-registrations. According to Play Console Help, an app can run one default graphics experiment or up to five localized experiments at the same time.

There are two experiment types. Default graphics experiments test the icon, feature graphic, and screenshots, and if your listing has only one language, they are shown to all users. If you have added localized graphics for a language, users viewing in that language are excluded. Localized experiments let you pick up to five languages and test descriptions in addition to the icon, feature graphic, and screenshots. Google recommends changing one asset at a time.

The metric choice deserves attention from game studios. Google Play offers three. Unique user install clicks count users who installed and did not have the app on another device at the time. Unique user open clicks count users who installed and then actually opened the app. Unique user pre-registration clicks serve games that have not launched yet. A variant that lifts installs but not opens, for example flashy screenshots far from the real gameplay, is more likely to be filtered out when you judge by opens. We discussed refining the listing during pre-registration in How App Store Pre-Orders and Google Play Pre-Registration Differ for a Mobile Game Launch.

There are two statistical settings. The minimum detectable effect is the smallest difference needed to declare a winner, and anything smaller is treated as a draw. Raising the confidence level reduces the chance of a false positive but takes longer to reach a result. The experiment setup screen estimates the time and the number of installs, opens, or pre-registrations needed for a statistically significant result, so you can adjust the number of variants or the audience percentage based on that estimate. Experiments stop automatically after six months, and traffic returns to the current listing. Results appear as "more data needed," a draw, or a recommended variant.

How is App Store Product Page Optimization different?

App Store Product Page Optimization (PPO) tests up to three treatments against the default product page for up to 90 days, and it covers the icon, screenshots, and app preview videos. You can create only one test at a time, and a test cannot be changed once it has started.

It differs from Google Play in three main ways. First, any new metadata in a treatment must be approved by App Review, and to test alternate icons, every icon variant must be included in the binary of your published app. Icon tests therefore have to be tied to your release schedule, while tests without alternate icons can be submitted for review independently of an app version. Second, PPO runs only on the default product page and is not available for custom product pages. Third, a person sees the same treatment throughout the test, and treatments are shown only to users on iOS and iPadOS 15 or later.

You set the traffic allocation yourself. In Apple's own example, if you allocate 40% of traffic to the test and have two treatments, each treatment receives 20% and the original page receives the remaining 60%. Apple recommends waiting until at least one treatment has been declared better or worse than the baseline with at least 90% confidence before applying a treatment or stopping the test, and notes that applying a treatment mid-test ends the test. App Store Connect Help also warns that releasing a new version containing the assets or metadata under test may affect the results.

One case Apple has published is an icon test by the brain training app Peak. It tested three icon treatments over 44 days. A brain-shaped icon improved conversion by 8% with more than 98% confidence, while a robot icon lost 3.5%. The design the team expected to win lost, which shows that the value of a store test lies as much in filtering out the option you were sure about as in finding a winner.

Are custom store listings and custom product pages testing tools or targeting tools?

Custom store listings and custom product pages are not randomized A/B tests. They are targeting tools that show a different page to each traffic source. Because the audience itself differs, a custom page converting better than the default page does not tell you whether the page or the audience made the difference.

On Google Play, you can create up to 50 custom store listings per app, and each one can change the app name, icon, descriptions, and graphic assets. The targeting options are especially useful for running a live game. Beyond country, pre-registration, specific search keywords, and Google Ads traffic, you can target users who uninstalled the app (churned), users who have not opened it in 28 days (lapsed), users who installed but never bought, one-time and repeat buyers, users with no purchase in 180 days, and custom audiences you define. That makes it possible to show returning players what changed since an update, or to show non-buyers how much of the game is free. Churned and lapsed targeting does not apply to new apps, and a country can be targeted by only one custom listing at a time.

Each custom listing can carry a unique URL parameter, so you can send a specific traffic source to a link in the form `https://play.google.com/store/apps/details?id=[packageName]&listing=[parameter]`. To connect Google Ads, set the audience to ads traffic, paste the ad group ID, and after approval select the listing as a landing page in Google Ads. Help states that these ads listings exclude Google Play search ads traffic, so their numbers will never match total app campaign numbers. And as noted above, Google Play lets you run experiments on custom listings too, so you can A/B test inside a page built for a specific country or keyword.

On the App Store, you can publish up to 70 custom product pages per app, and you can change the screenshots, promotional text, and app preview videos. The icon and app name cannot be changed. If you assign keywords, a custom page can appear in search results for those terms instead of the default page, and each keyword combination can belong to only one page. On iOS and iPadOS 18 or later, each page can carry a deep link to a specific screen in your app, and ad networks can pass the page ID through SKOverlay or SKStoreProductViewController to show a custom page from in-app ads. App Analytics shows impressions, downloads, and conversion rate per page, plus retention and average proceeds per paying user, so you can compare post-install quality page by page.

Apple says developers who refer people to a custom product page see a 2.5 percentage point increase in conversion on average. Actual adoption is low, though. In AppTweak's 2025 US Apple Ads data, 74% of games used no custom product pages in their campaigns, and only 7% used them more than five times in the year. In the same data, game ads using custom pages had a tap-to-install rate of 56.8%, 2.8 points higher than those without (54.0%). On Google Play, 26.5% of top US games ran at least one custom store listing. Keep in mind that these comparisons are simple splits by usage, not controlled experiments.

How many visitors does one store test need?

The number of visitors you need rises sharply as the baseline conversion rate falls and as the lift you want to detect shrinks. On a page converting at 16%, detecting a 5% relative difference (16.0% to 16.8%) at 90% confidence takes roughly 26,000 visitors per variant.

The table below gives illustrative estimates from the standard sample size formula for comparing two proportions (90% two-sided confidence, 80% power). Each store uses its own statistical method, so the console's estimate will not match exactly, but the scale is right.

Three practical conclusions follow. First, a game with 1,000 page visitors a day and an 8% conversion page would need close to six months to confirm a 5% relative difference, so at that scale you should compare options with clearly different messages (the genre's core fun, character-led vs combat-led, and so on) rather than small detail changes. Second, more variants mean less traffic per variant and slower results, so low-traffic games are better off starting with a single variant. Apple likewise suggests reducing treatments or increasing allocation if the estimate exceeds 90 days. Third, localized experiments only count visitors in the chosen language, so it is realistic to start in one or two core markets.

We covered impression counts for banner and video ad tests in What Is IPM? Meaning, Formula, and Mobile Game IPM Benchmarks 2026, Plus How Many Impressions a Creative Test Needs. If you bring the messages that won your ad tests into your store tests as candidates, you can validate the options with the best odds first, even on limited traffic.

In what order, and under what conditions, should a game store test be designed?

For games, most decisions are made on the icon and the first two or three screenshots (or the video), so test them in that order, and always check novelty effects, seasonal events, and traffic mix changes when reading results.

The icon is the only asset that also appears in search results and recommendation surfaces, so it affects behavior even before the page visit. Apple notes that alternate icons display not only on the App Store but also on the devices of users who downloaded the app, so on iOS an icon test also touches the experience of existing players. Next comes the first screen of screenshots. Most visitors do not scroll, so changing the message of the first two or three screenshots (whether core gameplay, genre signals, or social proof comes first) gives a stronger signal than swapping the whole set. Video costs more to produce and autoplay behavior varies, so it is more efficient to turn the messages that won in the first two steps into video afterward.

Watch for the following when reading results.

The traffic mix problem happens more often than you might expect. Once you start sending App Store ad traffic to custom product pages, organic traffic from search and browse makes up a larger share of the default page's visitors. Because PPO runs only on the default page, the same test can produce different results before and after ad traffic leaves. The same applies on Google Play: creating an ads listing changes the population of experiments on the default listing.

How do you match custom pages to ad creatives?

The key to matching ad creatives with custom pages is to show, on the first screen of the store, the scene the ad promised. Build one page per creative theme and look at each page's conversion together with post-install metrics, and your creative tests and store tests become one funnel.

In practice, start by grouping creatives into three to five themes. For an RPG, that might be "combat action," "character collection," and "guild co-op," with a custom page for each theme whose first screenshot and opening video frame start with the same scene as the ad. On Google Play, separate traffic by Google Ads ad group, and use unique URL parameters for other channels. On iOS, connect through Apple Ads ad variations or page IDs passed by ad networks. This lets you tell apart "creatives with a high CTR that lose people on the page" from "creatives with an average CTR that carry through to installs and opens." We covered how to read CTR at the ad stage in What Is a Good Ad CTR for Mobile Games? 2026 CTR Benchmarks by Channel, Format, and OS, and When a High CTR Misleads.

Note that per-listing reports only break out once a page receives enough visitors (per Google Play Help), so if you split creative themes too finely, no page will collect enough data to judge. Using the sample size table above as a guide, group themes into units that can each gather at least several thousand visitors.

The habit of splitting traffic by source carries over to judging the quality of each channel. On the same store page, visitors whose tastes match the game raise both conversion and open rates, while visitors who came only for a reward may add installs but drag down post-open metrics. Playio is an Android-based community of 5 million gamers that connects games with players whose tastes match, using genre preferences, play history, and in-game behavior data. Its default pricing model is CPI, and it also supports CPE based on in-game actions such as completing the tutorial or reaching a specific level. If you separate each channel's traffic with Google Play custom store listing URLs, you can compare the store conversion and post-open behavior of visitors from every channel, Playio included, on the same basis.

More details are available here. (https://playioadsen.oopy.io/bizdeck)

Key takeaways

Store-native tests find the page that turns the same visitors into more installs, and they cannot measure gains in search impressions. As of October 2026, Google Play store listing experiments run up to two variants against the current listing, with one default graphics experiment or up to five localized experiments at a time, judged by installs, opens, or pre-registrations, and they can run on custom listings too. App Store Product Page Optimization runs up to three treatments for up to 90 days, one test at a time, requires review for new metadata, and works only on the default page. Custom store listings (up to 50) and custom product pages (up to 70) are targeting tools by traffic source, not randomized tests, so a conversion gap should not be read as a page effect. Detecting a 5% relative lift on a 16% baseline takes roughly 26,000 visitors per variant, so the less traffic a game has, the fewer variants it should run and the more distinct its options should be. Read results weekly after at least one to two weeks of novelty, and avoid app releases and changes to ad budgets or traffic routing during a test.

For inquiries about Playio's advertising solutions, reach out at: [email protected]