The bottom line
Apple's native tools let you A/B test app icons: add every alternate icon in Xcode, ship them in an approved build, then run the test in App Store Connect. This shows which icon wins before you launch, even during pre-order.
Key takeaways
- Every icon you want to test must first sit in the app's asset catalog. 1:21
- Keep the default icon and each alternate in Assets.xcassets, added as a new iOS app icon through the context menu or the plus button.
- Choose "single size" for iOS in the right sidebar, so each icon needs only one image.
- The icons must ship in a build that Apple approves before any test can run. 3:46
- In the main app target's General tab, tick the box to include all icon assets.
- Archive the build and send it through the normal review cycle. Screenshots can be swapped freely, but icons cannot.
- The test itself runs under Product Page Optimization in App Store Connect. 4:36
- Once the app is ready for distribution, press the plus button, name the test and pick the treatment.
- The app icon tab appears only if the icons were set up correctly in the asset catalog.
- With one treatment, traffic splits 50/50 by default. With two treatments, Apple will not let you pick 100%.
- Change only the icon, and make the versions look clearly different. 7:21
- Keep screenshots identical so the icon is the only variable. Use dramatic changes, such as a purple icon against the original.
- Run one treatment at a time unless you have a lot of traffic.
- Results need enough traffic, and the test can start before the app is live. 8:49
- The sample test showed a 40% improvement but a confidence level of zero, because the numbers were too low.
- The demo app was still in pre-order, so the test helps pick the launch icon.
Keep the screenshots the same, change the app icon. You know, for me, in any AB test, I just change one parameter at a time.
Steve P. Young
Chapter summaries
0:00 Setting up the icons in Xcode
A better icon lifts downloads, and testing it starts in the Xcode project, not in the store.
- Icon names are free, but every icon must share one file. 1:46
- Xcode names the default icon app icon on its own.
- The alternate names were made up on the spot.
- No required naming scheme is known. The only rule is one shared file.
- A Fiverr designer made one icon and sent a Figma file. Justin edited it into several versions.
- The default icon is chosen separately in the target settings. 4:10
- The app icon field sits right above the include-all-icons checkbox.
- Changing that field changes which icon is the default.
- Single size is a recent Apple feature, so one image per icon is enough.
- Color, visibility and how fast the icon feels relatable to the target market all shape downloads.
4:30 Creating the test in App Store Connect
The test setup screen estimates how long the test will take and what share of visitors will see it.
- Traffic decides how long the test runs and how many people see it. 5:12
- The estimated duration is calculated from your traffic. More traffic means less time.
- The traffic proportion is the share of visitors who see the test.
- The Product Page Optimization list is empty the first time. Justin already had one test running.
- Justin tried adding a second test and believes Apple allows only one at a time.
- Justin doesn't know how Apple picks which visitors see which icon.
6:16 Reading a test that is already running
A running icon test uses the same screens as a screenshot test, so the same habits apply.
- The demo test used a tiny icon change, which is a weak example. 7:00
- The original icon is Treatment A, and only one treatment was chosen.
- Justin made the change small only so he had something to show.
- He plans to run bolder color changes soon.
- Screenshots stayed identical because only the icon was being tested.
- The screenshot field works like the normal product page. 7:47
- Treatment screenshots use the same field as the normal distribution page.
- Tinting screenshots to match each icon was considered, but it is not needed yet.
- The analytics view is the same as for any other test. 8:40
- The baseline can be changed.
- The simpler icon was ahead, and Justin thinks it is better.
- The tool could not confirm it with so little traffic.
9:09 Creating a new purple icon test
A new test takes a few clicks, and it only goes live after Apple reviews it.
- A treatment is one extra version beyond the original. 9:34
- One treatment is a plain A/B test. Two treatments make it A/B/C.
- Justin stopped his old test and named the new one "Icon 2 test", with a purple icon.
- Most viewers lack the traffic for more than one treatment.
- The icon is swapped in the app icon tab, not uploaded. 10:04
- The tab pre-fills the original icon, and "change" picks the alternate.
- Screenshots are different. Delete all and upload new ones.
- An icon has to ship with the build.
- Starting the test sends it through a review cycle. 11:04
- Justin picked purple against the simple icon, since simple won the earlier test.
- The treatment name can be edited before starting.
- After "start test", he submitted it for review.
In the growth library
Justin Allen and Steve P. Young, Justin Allen is an indie iOS developer; Steve P. Young is founder of App Masters. A step-by-step walkthrough of setting up app icon A/B tests with Xcode and App Store Connect product page optimization, including what to test and how to run it. The useful result is that icons can be tested even while an app is in pre-order, so you can pick the launch icon with real data.
Lessons that draw on this:
- How to win in the App Store
- How to run more experiments faster
- How to run trustworthy experiments
- How to judge test results correctly
More from
App Masters · The whole library, and ask it anything