App Store screenshot localization: the ZiMAD case, and what a solo developer should copy

A casual game publisher called ZiMAD ran an experiment on the Japanese App Store that is worth more to a solo developer than most ASO advice floating around. They rebuilt the screenshots for one of their puzzle games with the Japanese market in mind, and measured the result afterwards: conversion rate up 36 percent. The figures come from iTunes Analytics and were published as a case study with SplitMetrics. The interesting part is not the percentage. It is what they actually changed, and what their own write-up admits went wrong.

Read the case closely and you notice they did not translate their screenshots, they reconstructed them. The team studied how Japanese publishers design their own store images: dense compositions, lots of small elements, effects and emotional markers packed into the frame. To a European eye it looks busy, almost chaotic. To a Japanese user it looks familiar, and familiarity is the product here. The case study borrows a term from psychology for this, processing fluency: the brain processes familiar images faster and rewards that ease with a pleasant feeling. Your screenshots are not an illustration of your app. They are a first impression built from visual codes, and those codes are local.

Why should that matter to someone shipping one app alone at night? Because of how short the first impression is. SplitMetrics analysed close to 200 apps and more than 3,500 A/B tests across both stores for their ASO Benchmarks and Mobile Trends Report 2024, and the most useful line in it is brutal: nearly half of app store visitors look at only the first two screenshots, and only 7 percent ever reach the fifth. Your description is read by a minority. The first two images are the entire pitch.

Now the uncomfortable parts of the case, because they are the parts nobody quotes. First, ZiMAD's own rule of thumb: a screenshot variant that wins an A/B test by around one percent does not reliably hold up in the real store and can even turn negative, so they only shipped changes with a visible margin, at least two percent. If you are one developer who cannot run hundreds of tests a month, take the warning seriously. Move the structure, not the pixels. Second, their admission that their user acquisition interstitials were never aligned with the new screenshots, so the ads kept bringing a different audience to a page built for someone else. The ad creative and the store images are one funnel, not two projects.

None of this is as quiet as the default of the whole ecosystem. Most listings are written in English (US) and then never touched again. Translate the text but keep the images, and Google's own Play Console help says it plainly: text translations without localized graphic assets mean the listing keeps showing graphics from the default language. Apple structures screenshots the same way, as per-localization assets, and a storefront with no images of its own falls back to the next best available language. Nobody warns you. Your listing simply shows an English screenshot to a Brazilian user, and nothing in your dashboard tells you it happened.

That gap has a price, and I know it because I paid it. I have spent on Meta and TikTok ads and learned that ads rent you an audience: the day you stop paying, the downloads stop with them, and I have watched a launch go quiet that way. Organic placement does not work like that, which is why I care about the boring side of ASO. Downloads are concentrated outside the United States, and languages do not behave identically. On my football app, French and Brazilian Portuguese brought the volume, while German users converted to paid subscriptions at a much higher rate. Volume in one language, revenue in another, and both of them invisible from an English-only listing. The niche decides which markets are worth the effort: a football app should fight for Brazil, Indonesia, Turkey and France before it fights for the United States.

Here is what I would copy from the Japanese experiment with one app and no budget. Pick one market where your niche is strong. Spend an evening as a user there: search your main terms in that language, open the top five competitors, and note the visual grammar they share. Rebuild your first two screenshots for that market, captions in the local language, a few words each. Leave the rest of the set alone for now, almost nobody will see it. Then let the store re-index, and give it a month before judging anything.

The honest blocker is the mechanical work. Every language needs its own captions, its own exports in the sizes each store demands, and a manual upload pass every time you change a screen. That iteration cost is what stopped me for years, and it is what AsoFast exists for. It generates App Store and Google Play metadata with AI, translates listings into more than 80 languages, writes screenshot captions in the target language and frames your images at the sizes the stores require, then publishes directly to App Store Connect and Google Play. It runs entirely on your machine, credentials included, open source and free. Put that next to the usual stack, where ASO platforms run from roughly $30 to $80 a month, screenshot design tools sit near $19, and the enterprise intelligence suites start around $500.

Start with one language this week. The screenshots are the cheapest part of your listing to fix, and the ZiMAD case is a reminder that the fix is visual, not only textual. The code is at https://github.com/ediaStudio/asofast.