Translating the ad got me a worse result than not translating it
The setup. Push offer expanding from US to Germany and Brazil. I ran the English winner machine-translated into German and Portuguese. Illustrative numbers.
The move. Translation felt like sufficient localization. Same creative, same angle, just in-language. Scaled all three geos in parallel.
The numbers. US (original): 1.5 ROI. Machine-translated DE: 0.7 — the translation was stiff and the cultural angle (a very American urgency hook) read as spammy. Native-rewritten DE by a local: 1.3. Brazil was worse: the literal translation of a US idiom was nonsense, 0.5 ROI, until a local rewrote it around a different emotional hook entirely — then 1.6, BETTER than the US.
The lesson. Translation moves words; localization moves the angle. A converting hook is built on a cultural trigger, and triggers don't translate. Machine-translating a winner often performs WORSE than leaving it in English, because broken-native reads as a scam.
What I'd do differently. Budget for native rewrites, not translations, before entering a geo — or don't enter it. The $40 I saved skipping a native copywriter cost me weeks of 0.5-ROI spend. The Brazil rewrite outperforming the US original is the whole lesson: the right local angle isn't a tax, it's an edge.
Arb Files
@ArbFiles