Localisation was my cheapest feature and my worst growth bet

Localisation was the cheapest feature I ever shipped and the worst growth bet I have made so far. The app runs in 21 interface languages, two of them right to left. The translation files were the easy part. The real cost was Arabic and Persian: mirroring the whole layout properly took longer than the other nineteen languages combined, and I did it anyway because I work in a second language myself and I refuse to ship a tool built for people like me in English only.

Here is the uncomfortable result. I cannot point to sales that the languages produced on their own. The reason is obvious in hindsight. A translated interface is a product feature, but discovery still happens in one language. My listings, my posts and my search presence are all English. Someone searching in Indonesian or Turkish never finds the product, so they never learn that the interface would have greeted them in their own language. Localisation converts users you already reached. It does not reach anyone.

If I did it again I would still ship all 21, but I would pair every language I claim to support with at least one distribution channel in that language, even something as small as one directory listing or one forum where that language actually lives. Support without distribution is a museum exhibit.

For those of you who localised: did any single language ever pay for itself, and if it did, how did those users find you in the first place?