ASO20 September 20264 min read

Are your App Store screenshots costing you installs?

A diagnostic you can run in ten minutes, on your own listing, without any tools. Six specific failures that leak installs from traffic you already earned — and which ones are worth fixing first.

A row of smoked glass panels with one cracked and dimmed

Screenshots do not bring you traffic. They decide what happens to the traffic you already have — organic search, paid clicks, referrals, press. Which makes a weak set expensive in a quiet way: you paid for the visit and lost it at the last step, and nothing in your analytics is labelled "bad screenshots".

Here is a diagnostic you can run right now, on your own listing, with nothing but a phone.

Run it properly

Open your listing on a phone, not on a desktop browser. Hold it at normal reading distance. If you can, go outside, or at least away from a dark room. This matters more than it sounds: you designed at 100% zoom on a large bright display, and your customer is looking at a thumbnail in daylight.

Now work through these.

1. The two-second test

Look at your first screenshot for two seconds, then look away and say what the app does.

If you cannot answer — and you built it — a stranger has no chance. This is the single highest-leverage fix available to almost every listing, and it is usually failing because frame one is a splash screen, a logo lockup, or an onboarding slide. All three feel like the natural opening because they are the natural opening inside the app. Neither is the natural opening to a stranger.

Three smoked glass panels, the first much larger and brighter than the two behind it — Everyone who reaches your listing sees frame one. Only a minority swipe. The set is not eight equal panels.

2. Can you read it at actual size?

Screenshots appear in search results at roughly a quarter of the size you designed them. Look at your captions on the phone, at arm's length. If you are squinting, they are decorative rather than functional — and a caption nobody reads is worse than no caption, because it took space from something that would have been read.

3. Does the set make one promise or five?

Read your captions in order, out loud. If they name five unrelated features, your listing reads as five different apps and a browsing user has to assemble your value proposition themselves. They will not.

The strongest sets make one promise and then spend the remaining panels proving it.

4. Would you know it was yours?

Open your three closest competitors' listings in other tabs. Look at all four sets together.

If yours is interchangeable with theirs, that is not a small aesthetic complaint — it means the visual work is doing nothing to distinguish you, and your install decision comes down entirely to rating, rank and icon. This is what a template library does at scale: what a browsing user recognises stops being your app and starts being the tool everyone used.

5. Is anything in there not true?

Look inside every device screen. Any rating, review quote, download count, streak, revenue figure or user total that is not real?

6. The boring technical pass

Less interesting, still costs you:

Are all screenshots within a display class exactly the same pixel size?
Is anything visibly upscaled from a smaller capture, and therefore soft?
Does the status bar change between panels — different times, different battery?
On Google Play, is your feature graphic a stretched screenshot?
Are you missing a required size, so the store is scaling something for you?

If you are already staring at an upload error, the dimensions guide covers the exact-pixel rule that causes most of them.

Fix them in this order

Not all of these are worth the same.

Frame one (test 1). Largest audience, largest effect, usually the cheapest change.
Anything untrue (test 5). Not a conversion issue — an existential one.
Legibility (test 2). Cheap, mechanical, immediately better.
One promise (test 3). Requires deciding what you are, which is why it gets skipped.
Distinctiveness (test 4). The expensive one, and the one that compounds.
Technical (test 6). Boring, fast, do it while you are in there.

Then stop guessing

Everything above is a prior, including our opinion about it. Apple's Product Page Optimization and Google's Store Listing Experiments both run a variant against your live listing with real traffic, and they are the only honest referee available. Change one thing at a time, start with frame one, and let the test reach its own confidence threshold before you believe it.

More on running those properly in product page optimization, and if you would rather someone else looked, we do a free review.

ASO