App Store product page optimization: Apple's A/B test

Product page optimization (PPO) is the A/B test built into App Store Connect. You compare up to three variants of your default product page against the original, changing the icon, screenshots or app previews, on a share of traffic you pick, for up to 90 days. Apple marks a variant as better or worse once it reaches 90% confidence. You can then apply the winner to your page.

I publish my own iOS apps, including Solys and TradaZen, with product pages in 38 languages. In October 2026, I set up a PPO test for TradaZen. This guide is based on the App Store Connect help, reread on October 15, 2026.

App Store product page optimization: what you can test

According to Apple's page on product page optimization, each variant, which Apple calls a treatment, can change three things:

• the icon, as long as it's already in the binary of the live version;
• the screenshots;
• the app previews.

You can't test the name, subtitle or keywords here. The Analytics help also mentions the description. But the treatment setup page sticks to the icon, screenshots and previews, plus a Header and Search Results tab for the header visuals that arrived in October 2026.

The requirements, according to the help overview:

• your app is on sale on the App Store (Ready for Distribution status);
• the test runs on your default product page, never on a custom product page;
• variants show on iOS 15 and iPadOS 15 or later.

Two details matter. A visitor sees the same variant for the whole test. And if they download the app from a variant with a different icon, that icon also shows on their iPhone. Variant visuals can appear in search and in the Today tab.

Creating a test: variants, traffic and languages

In App Store Connect, the Product Page Optimization section is where you create a test. According to the help on creating a test, you set five things:

1. A reference name, 64 characters max, visible only to you.
2. The number of variants, up to three. The more you have, the longer the test runs.
3. The traffic share, split evenly between variants. Apple's example: at 30% with three variants, each one gets 10% of visitors.
4. The languages. A visitor whose language is excluded isn't part of the test.
5. The duration estimate, optional. You pick the improvement you're aiming for, and Apple calculates the duration and impressions needed from your numbers.

A test lasts 90 days max, and you can't edit it once it's running. If you stop it, it won't restart: you have to create a new one.

Each variant starts as a copy of your page. Whatever you don't change stays identical to the original, screen size by screen size. Dimensions are in App Store screenshot sizes.

Variant review

According to the help on treatments, a test's visuals must be approved before it launches:

• New screenshots or new previews: a review, no new app version needed.
• Just a new order, or icon only: already approved, no review.
• New icon: it must be in the binary of the live version, at 1024 x 1024 pixels, built with Xcode 13 or later. So you need one more version before the test, with a new build number.

When you start, the help on running a test offers to add the test to a submission. It's reviewed with your latest iOS version. Typical timelines are in App Store review time.

Apple also warns: releasing a new version during the test can skew results if it changes a tested element.

Reading the results

Results are in Analytics, under Acquisition, then Product Pages. According to the App Store Connect Analytics help, they show up after the first 5 downloads attributed to the test, then update daily. Apple uses a Bayesian method, and each variant gets a status:

Apple status	Meaning	What to do
Collecting Data	No verdict yet	Wait
Performing Better	Better at 90% or more	Apply it
Performing Worse	Worse at 90% or more	Keep the original
Likely to be Inconclusive	No verdict within 90 days	Test something bolder

For each variant, the table shows unique impressions, estimated conversion rate, improvement over the baseline, and confidence. A big improvement gets detected faster than a small one. Apple asks you not to stop a test before the verdict. You can also change the baseline at any time to compare two variants against each other.

To apply the winner, the help on applying a treatment sets three rules:

• one variant per test, and the action is final;
• only screenshots and previews get applied: a winning icon has to become the default icon in your next version;
• applying a variant stops the test.

My rule: no test under 1,000 impressions

I hold all my apps to one rule: no A/B test until what you're testing gets at least 1,000 impressions a month. Below that, chance weighs more than the variant. It's not an Apple threshold, it's a floor. The App Store Connect duration estimate then tells you how many impressions you really need.

If your traffic is low:

• just one variant, so you don't split already thin traffic;
• a clear difference: a different first screenshot, not a color tweak;
• the languages that get traffic, not all of them;
• a high traffic share, to go faster.

Below the floor, change the page all at once and compare before and after over several weeks, noting the date, just like when you track your app's ranking. It's less rigorous, but honest about what your numbers can tell you.

Before launching a test: my checklist

This is the list I go through before launching a test:

• A hypothesis written in one sentence: "showing X first will get more downloads."
• Only one element changed per variant, so you know what made the difference.
• At least 1,000 impressions a month in the tested languages.
• An Apple duration estimate under 90 days, otherwise fewer variants or more traffic.
• The alternate icon already in the live binary, if you're testing it.
• Each language's visuals checked in the variant, not just English.
• No planned version touching the tested elements during the test.
• A decision written down in advance for each possible status.

My case: the "duel first" test on TradaZen

On TradaZen, my trading simulator, the duel against another player is the mode I want to highlight. Its story is in my case study. My hypothesis: showing it in the very first screenshot will get more downloads than the current opening. Here's the setup, with no results yet.

• One variant, "Duel first," against the original.
• 50% of traffic, to go faster with a single variant.
• Ten languages, the ones with traffic, not the page's 38.
• Screenshots replaced per language. Each language in the variant inherits the version's screenshots. Adding more on top went over the limit of 10 per size: I had to replace the whole set.
• No submission until the version in review is accepted. The test is reviewed with the latest iOS version, and I don't want a rejection of one to delay the other.

Once it's running, I'll only decide at Performing Better or Performing Worse.

PPO vs. Custom Product Pages: the difference

People often mix up the two tools. Here's the gist, according to the help on custom product pages:

Criterion	PPO	Custom Product Pages
Goal	Test your default page	One page per audience
Number	3 variants per test	70 pages per app
Who sees it	Randomly selected visitors	Dedicated link or keywords
Icon	Testable if in the binary	The app's icon
Promotional text	No	Yes, per page
Duration	90 days max	No limit
Own URL	No	Yes

PPO tells you which page converts best for everyone. A custom page shows a different page to a specific audience. You can't test a custom page with PPO. Details are in the Custom Product Pages guide, and for a time-bound moment, read the one on in-app events.

To prepare your variants

The App Store Screenshots Kit brings together the 4 screenshot styles from my apps and the skill Claude uses to build yours, in every size and every language. Enough to prepare a variant without design software. For the words on your page, read App Store keywords.

Getting ready to submit? The free diagnostic sends you my Apple anti-rejection checklist and tells you what's left to do. The full method for your product page is on the App Store ASO page.

Frequently asked questions

What is product page optimization on the App Store?

It's App Store Connect's A/B testing tool. You compare up to three variants of your default product page against the original, changing the icon, screenshots or app previews, on a share of traffic you pick.

How long does a PPO test last?

90 days max, or until you stop it. A stopped test won't restart. Apple recommends waiting until a variant is declared better or worse at 90% confidence.

Can you test the name or keywords with PPO?

No. Variants change the icon, screenshots and previews. The name, subtitle and keywords change with a new app version.

Do you need a new version to test an icon?

Yes, if the icon isn't already in your app: tested icons must be in the binary of the live version. For screenshots or previews, a review is enough.

How many visitors do you need for a PPO test?

Apple doesn't give a minimum. Results show up after the first 5 downloads, and the duration estimate tells you how many impressions to aim for. My personal rule: no test under 1,000 impressions a month.

Can you run several PPO tests at the same time?

Apple's page on product page optimization talks about one test at a time. The Analytics help mentions simultaneous tests on different elements. For a small app, one test at a time is still the easiest to read.