When I submitted an app to the store for the first time, I uploaded eight screenshots straight from my phone: the interface was real, every feature was there, and review did not block me. The real problem showed up in the analytics: only 1.9% of product page visits turned into downloads, and the first three images held people for less than 4 seconds on average. When I rebuilt the screenshots the second time, with the same product and the same copy slots, conversion rose to 4.6%. What I changed was not the product but whether the screenshots were packaged: a device mockup, a background layer, and one headline people could read. With those three in place, visitors understand they are looking at an app rather than a few web page captures.
1. A promo screenshot is not a beautified screenshot
A bare screenshot lacks context. It assumes the reader already knows what they want, but people in a store scan from left to right, and the first image answers one question: does this have anything to do with me. A promo screenshot carries a judgment made within three seconds, so it needs three things: a phone mockup to hint that this is something you can install, one headline that states the value, and a background in the same color family as the app icon so the eight images read as one set. I have since seen people turn promo screenshots into posters, packing in five lines of copy and three decorative elements, and that fails in the other direction: the interface inside the mockup is squeezed until it cannot be read.
2. How I allocated the eight images
| Position | What goes there | Headline approach | Mistake I made |
|---|---|---|---|
| Image 1 | The screen for the core use case | One value line, 12 characters max | I first wrote "efficient to-do management", which was too vague, then changed it to a concrete result |
| Images 2-3 | The main flow split into two steps | Start with a verb, say what the user does | Both used the same screen, so readers thought it was a repeat |
| Image 4 | Differentiating features (reminders, sync) | The subhead adds a limitation | I left out "sync limited to 2 devices" and got bad reviews after launch |
| Image 5 | Stats or report page | Use real numbers, not fake data | The original image showed a client name and I only remembered to handle it before submitting |
| Image 6 | Themes, dark mode | One line is enough | Dark interface on a light background, not enough contrast, the text went muddy |
| Image 7 | Small scenarios (offline, import) | State input and output clearly | Two features squeezed into one image with no visible focus |
| Image 8 | Closing, brand color across the whole frame | One line of attitude | The background was in a different color family from the first seven, so it jumped when swiping |
One figure in my data is even uglier: only 23% of visitors reached the fifth image. So I revised the first three through three rounds and barely touched the rest.
3. How I actually build one set
- First I take real screenshots on the phone. For the ones that show real usernames, I use Image Mosaic to box out the client name and the phone number, export a PNG and move on. Do not wait until the shot is inside the mockup to notice.
- Then I open App Store Screenshot Maker and pick a device mockup: iPhone 6.5 inch, 5.5 inch, mainstream Android models and custom sizes are all available. My habit is to produce the 6.5 inch version first, then run the same set through 5.5 inch just to see whether the headline gets too small to read, because that is the screen most people actually hold.
- For the background I used the brand color
#2F6BFF, as a gradient from dark to light with a faint noise layer on top. Flat color looks dull and patterns look heavy; I tried the polka dot option and it crushed the text inside the mockup. - Headline and subhead: among the Chinese fonts on offer I picked one with heavier strokes, kept the headline within 12 characters and the subhead within 20, placed them a little above center, and never let them cover a key button in the interface.
- I made the device frame one notch thinner. The default border width ate too much of the frame on the 5.5 inch version.
- I export 1080 x 1920 PNGs, about 12MB for all eight. If one file goes over, reduce the size or compress it once; Image Compressor usually takes about half off a PNG.
- Before uploading I open my own product page on a second physical phone to check the thumbnails. That step caught the thin headline twice, and the problem is invisible on a computer screen.
4. Details that kept going wrong
- Status bar time has to match: my eight images mixed 9:41, 10:26 and 23:07, which looked like they were taken on different days, so I shot them again.
- Fake data backfires: a report page showing "10,000+ completed" gets taken seriously and readers ask about it, which is more trouble. I run a test account until it produces data at a realistic scale.
- Avoid absolute claims: my first headline said "number one for efficiency" and I removed it myself. Beyond how it looks, that kind of wording can cause problems in review.
- Keep competitor names and comparison tables out of the images: this is one of the highest review risks. I describe differences in text instead of naming competitors in a chart.
- One point per image: this was my biggest change this round. Eight images map to eight points, with no merging.
5. The nine checks I run before submitting
After that round I wrote the checks into a list and stuck it to the edge of my monitor. Running eight images through these nine items takes about twelve minutes, far faster than starting over after a rejection.
- Headline length: 12 characters max per image. If it runs over, cut it to 12; do not shrink the type.
- Minimum type size: if the main headline's cap height is under 40 pixels in the 5.5 inch mockup, rewrite the copy rather than change the font.
- Interface overlap: title, subhead and device mockup overlap by zero as far as I can see.
- Status bar: time, battery and signal all consistent. I capture everything in the same session.
- Sensitive data: names, phone numbers, amounts and QR codes checked one image at a time at 200% zoom.
- Background consistency: all eight use the same gradient settings, so the color does not jump when swiping.
- Third-party content: competitor names, other people's screenshots and stock system app icons all removed.
- Absolute wording: search for words like "first", "best" and "national level" and change any hit.
- Size and count: under 1.5MB per image, about 12MB in total, compress first if it goes over.
6. Common questions
Q: Do all app stores require the same screenshot sizes?
No, and the rules change. I produce one visual set at 1080 x 1920 as a working draft, then check the size and count of each image against the prompts in the store console before uploading. I do not rely on memory.
Q: How far can I get without a design background?
Device mockup, background and one headline: with those three you are at seventy out of a hundred. The rest of the gap is copy, not visuals, and that is the part I needed three passes to get right.
Q: Should the headline name the feature or the result?
Name the result. "Automatically turns notes into a weekly report" beats "supports export". The first tells readers within three seconds what they save; the second makes them work out how you organize your features.
Q: Can real user data appear in the screenshots?
No. Either switch to a test account or redact first. I masked the client name on the first report page, and before exporting I zoom out for one more look to catch anything I missed.
Q: Are the screenshots uploaded when I build promo images?
No, the whole generation runs locally in the browser. Before a product ships, that is a precondition for me, not a bonus feature.
4.6% is not the end either. After the second version I revised the copy twice more, and the latest pass changed the value line on the first image from "sort out your to-dos" to "a list generated automatically before you leave work each day". What really changed my mind was one reader comment: I had assumed people were comparing feature counts, when in the first three seconds they only wanted to know whether this had anything to do with them.