Should I translate Appray into other languages?

I live in Rijeka. English is not my first language, and Appray is built entirely in it, including every piece of copy a user sees during the interview. So when the subject of localisation comes up, people assume I must have thought about it already and have a plan. I had thought about it. I did not have a plan, and after a weekend of actually trying, I still don't, on purpose.

What the numbers actually say

Appray has no analytics inside the app, that part hasn't changed. But App Store Connect still tells me which country each download came from, because Apple collects that whether I ask for it or not, and it's the same aggregate report every developer gets. The United States is 51% of downloads. Germany is second at 9%, ahead of the UK at 7%. Croatia barely registers, under 1%, which stung a little the first time I looked.

So Germany is a real market and nobody there has asked me for a German version. What they have done is write support emails in careful, slightly formal English, the kind you write when you've rehearsed it. Twelve of the 26 were German, eight Italian, six Croatian. Every one answered a question about the interview, never about the language it was in.

The job isn't "read it in my language"

This is the part I had backwards at first. I assumed the job, if there was one, would be comprehension, that someone in Munich is struggling to parse an English sentence and wants it swapped for a German one. That's not what the emails describe. The emails describe people who read English fine and are using Appray to interview themselves honestly about a business decision, answering questions like "who specifically asked for this" in their own words.

That's a different job. The job a translation would actually be doing is letting someone think in their first language while they're being that honest with themselves, instead of running the thought through English on the way in and out. It's closer to what a therapist means when they say the language you dream in matters. Nobody in my inbox has said this directly, because nobody describes their own cognition to a support address. But it's the only version of "translate the app" that explains why fluent English speakers would still want it.

Where it actually broke

I treated this as a string file problem for about a day, which is the mistake everyone warns you about and I made anyway. Appray's interview isn't a fixed script. It generates follow-up questions by inserting the feature name you typed into a sentence template, so "Who specifically asked for {feature}?" becomes "Who specifically asked for the CSV export?" using whatever words you gave it. In English that's one swap and the sentence still works, because English barely inflects nouns.

Croatian has seven grammatical cases. The ending on "the CSV export" changes depending on whether it follows "for" or "about" or "instead of", and it changes by gender too, which I don't know in advance because the user typed the feature name, not me. German is gentler, three genders and four cases rather than seven, but you still need to know whether "the export" is der, die or das before you can put the right article in front of it, and an English phrase a user typed doesn't come with that information attached.

I spent most of a Saturday trying to dodge this by rewriting the templates so nothing ever needed to decline, things like "Regarding {feature}, who asked for it first?" It technically works. It also reads like a form letter in every language I tried it in, including English, which defeats the entire point of an interview that's supposed to feel like it's actually talking to you. A tool whose whole value is making people answer honestly cannot sound like a government form while doing it.

What I'm doing instead

I'm not translating the app, not yet. What I did translate, last week, was the App Store listing itself, the description, the screenshot captions, into German. That part doesn't touch the interview engine at all, it's static text I write once, and the general pattern holds up in everything I read while deciding this: localised store listings reliably move downloads in that language's market, sometimes by a lot, because the App Store itself surfaces apps differently once there's local-language metadata attached to them.

That's the Prove version of this decision. If German downloads move after a translated listing with an otherwise-English app, that tells me Germans will tolerate English inside the product in exchange for finding it in the first place, which is useful and cheap to learn. If nothing moves, that's informative too, and it costs me a translated App Store page rather than a rebuilt interview engine. Either way I'm not committing to the expensive version, full in-app localisation with a template system that can handle grammatical case, until I know whether the cheap version even matters.

I'll check the German numbers again in a month. If I ever do build the in-app version, it won't be a string file and a translator. It'll need the interview engine to know the gender and case of whatever a stranger just typed into a text field, which is a real piece of engineering I don't have yet and don't want to fake with something that sounds like a form.