How often do the top 10 apps change their screenshots?

We watched 610 iOS apps holding top-ten positions and compared their full screenshot sets capture by capture. In a median 17-day window, 27.5% changed them. Most of those swaps kept the same number of images, which is why almost every other count of this is too low.

Why this number is usually wrong

Most attempts to measure creative churn count screenshots: six last week, six this week, nothing happened. That misses the commonest kind of change by a wide margin, because iterating on a screenshot set usually means replacing image three, not adding a seventh.

We store the full ordered list of screenshot URLs on every capture, not just the count. Apple's asset URLs carry a per-upload identifier, so when the list changes something was re-uploaded. In our data 84% of confirmed changes kept the count identical. Count-only measurement sees one change in six.

Method, including what we had to throw away

We took every iOS app observed holding a top-ten position in our corpus, kept those with at least three captures spanning a week or more, and compared consecutive screenshot lists. That leaves 610 apps with a median observation window of 17 days.

A change counts only when it persists into the following capture. That rule exists because a single odd reading is otherwise indistinguishable from a real swap, and it costs us the ability to detect a change in the final capture of each series, which we accept as the cheaper error.

The result

183 confirmed changes across 610 apps. The mean works out at about five changes per app per year, but the mean is doing something misleading here: the median app made zero. Among the apps that changed at all, the rate is roughly 15 a year, or a swap every three or four weeks.

So the honest summary is not “top apps iterate constantly” and not “top apps never touch their creative”. It is that a minority iterate hard and the majority leave the set alone, which matches what we found when we looked at icons: the behaviour is concentrated in a small group rather than spread across the top of the store.

What bounds the rate: the test, not the design

Apple documents Product Page Optimization plainly. A test “can include up to three treatments with alternate app icons, screenshots, and app previews”, shown to a randomly selected share of visitors, and the dashboard reports “the number of impressions, conversion rate, percent improvement, and confidence level for each treatment compared to the chosen baseline”.

That design is what sets the ceiling on sensible iteration. A test needs enough impressions to reach a confidence level worth acting on, and an app without much traffic will wait weeks for that. An app changing its screenshots every three days is not running out of ideas faster than everyone else; it is either enormous, or it is not testing.

Google Play offers store listing experiments with a similar shape. The practical consequence is the same on both stores: your traffic, not your design team, decides how often you can honestly change anything.

What the research says about visual assets

There is real academic work on whether store visuals move behaviour, and it is more careful than the industry case studies. Wang and Li, in Electronic Commerce Research in 2016, extracted colour, visual complexity and symmetry features from Android app icons and found measurable relationships between those properties and download behaviour — evidence that visual choices are not cosmetic, established on features you can compute rather than on taste.

That is about icons, not screenshots, and we are not going to pretend otherwise. What it supports is the general proposition that the creative is a measurable variable. What it does not supply, and what nobody has published at scale, is a causal estimate of what a screenshot swap does to conversion across a population of apps. The vendor case studies that claim one are almost always a single app with no control.

The same caution applies here. We measured how often the set changes. We did not measure whether changing it helped, and nothing in this study should be read as saying it does.

How to use this

Blog · Free ASO tools · Products · AsoTheory