I shipped my app in English, watched it do fine in the US and UK, and assumed the rest of the world would catch up once word got around.

It didn't. The install rate in non-English markets was roughly a third of what I saw in English ones, even where I had decent traffic. So I dug into
it, and the answer was embarrassingly simple: people couldn't read my store listing.

Not the description. They can machine-translate that in their head, or the store does it for them. The problem was the screenshots.

Store screenshots are the part nobody localizes

Here's the thing about app store screenshots. They're not really screenshots. They're marketing posters with a phone in the middle and a headline on
top. "Track every workout in one tap." "Your money, finally organized." That headline is doing most of the selling.

When someone in Tokyo or São Paulo lands on your listing, those headlines are still in English. The store will translate your text description, but
it will never touch the words baked into your screenshot images. So your strongest sales pitch shows up in a language a big chunk of your visitors
don't read.

I checked the top apps in a few categories across regions and the good ones all do this. Their screenshots are fully localized, headline and all.
The ones that don't bother are mostly small indie apps. I was in the second group.

Machine translation gets you 80% there, and that's the problem

My first attempt was the obvious one. Run my English headlines through a translation API, drop the output back into my design template, export.

It looked fine to me. It was not fine to native speakers.

A few issues I hit:

• Length blows up your layout. German headlines ran 40% longer than English and overflowed the safe area. A line that fit perfectly in English wrapped into three lines and collided with the phone frame.
• Tone goes flat. "Crush your goals" translated literally into something closer to "destroy your objectives" in two languages. Technically correct, reads like a robot wrote it.
• Marketing copy isn't literal copy. A good headline in French is not the word-for-word translation of a good headline in English. It's a different sentence that lands the same way.

So a pure translate-and-paste pipeline produced screenshots that were worse than leaving them in English, because now they looked low effort instead
of just foreign.

** What actually worked**

The version that moved my numbers treated each language as its own piece of copywriting, not a string swap:

1. Write the headline natively per locale, not translated. Short, punchy, sized to fit the layout. If it doesn't fit, rewrite it shorter rather than shrinking the font.
2. Re-check the layout per language. Longer languages need a different break point or a tighter headline. You can't reuse one template blindly.
3. Keep the visual system identical across locales. Same colors, same device frame, same composition. Only the words change. This is what makes a set look intentional instead of random.

Doing this by hand in Figma for one language is a half day. For eight, it's a week I didn't have.

The shortcut I ended up building around

I do this with SnapMonk now because the localization step is built into the generation. I upload my raw screenshots, pick
the output language, and it writes the on-image copy natively in that language instead of translating mine, then lays it out so the longer strings
still fit the frame. I run it once per locale and get a matching set.

To be clear about the bias: I work on SnapMonk, so take the plug for what it is. But the workflow above is the real lesson, and you can run it
manually if you'd rather. The tool just collapsed a week of layout babysitting into an afternoon, which is the only reason my listing is localized
at all.

If you do one thing

Localize your first screenshot in your two biggest non-English markets and watch the install rate for two weeks. That single image is the
most-viewed asset in your entire listing. It's the cheapest high-leverage change I've found, and almost nobody does it.

I'd genuinely like to hear what your numbers do if you try it. Drop them in the comments.

*A few notes: *

• The referral link is a single contextual SnapMonk link plus a disclosure that you work on it. dev.to readers are allergic to undisclosed promotion, and an honest "I built this, here's the bias" reads better and survives moderation. One link is also safer than several for spam filters.
• canonical_url is left blank on purpose. Leave it blank if this is original to dev.to so dev.to keeps the SEO weight and links out to you. Only set it (to a snapmonk.com/blog URL) if you publish the same post on your own site first and want to avoid duplicate-content dilution.
• I kept it in the approved blog voice: "I" for the personal story, "we"/product for SnapMonk, no em-dashes, no slop tells.