Translation is not localisation. The same product claim can be a requirement in one market and a nice-to-have in another.
The usual way to take a campaign international is to run the winner everywhere and translate the copy. It is fast, it is cheap, and it quietly wastes a large share of the budget — because a product claim does not mean the same thing in every market it lands in.
Take an alcohol-free formulation. In the Gulf that answers a requirement, and it belongs in the first line. In the US and Europe the same fact reads as a sensitivity and wellness benefit, competing against a hundred other wellness claims, and leading with it wastes the impression. Same product, same sentence, two entirely different jobs.
So the unit of localisation is not language. It is the reason someone buys. Work out what the product competes against in each market and let that decide the angle, the hero product and the objection you handle first. Translation happens afterwards, to copy that was already right.
The second half is measurement. Platform-reported returns will tell you every market is working, because every platform is grading its own homework. Reconcile against actual revenue before deciding where the next budget goes, and judge on contribution margin rather than blended return — the customer who buys a concentrated oil is not worth what the one who buys a lip balm is, and a blended number hides exactly that.
The upside of doing this properly is that markets stop competing for one global creative and start compounding independently.