I localized into 21 languages before I had real revenue. I still cannot tell you if that was discipline or procrastination.

Two of those languages are right to left, which means Arabic and Persian users get a fully mirrored interface rather than English text shoved into a layout facing the wrong way. That work is not a translation file. It is layout, icon direction, text alignment and a long tail of small things that only break for people you will never watch using the product.

The case against doing it early is obvious. I had almost no revenue. Every hour spent on mirroring was an hour not spent on distribution, which is still my actual constraint. A rational allocator would have shipped English, waited for demand, and localized on evidence.

The case for it is that the product is built for people who work in a language that is not their first, and shipping that product English only would have been a contradiction sitting at the centre of the pitch. Also, retrofitting right to left into a mature interface is genuinely painful, while building it in early is merely annoying.

What I cannot claim is that it paid. I do not have the numbers to prove those languages brought users I would not otherwise have had, and I am suspicious of my own reasoning here, because the work was interesting and interesting work is very easy to justify after the fact.

So a real question rather than a rhetorical one: has anyone localized early and been able to show it mattered? Not a feeling, a number.