About once a quarter we get the same email: an overseas brand running Japanese GEO traffic, frustrated that their “localized” creative is underperforming, asking us to look at what’s wrong. We open the deck. The headlines are grammatically correct. The CTAs say the right things. The imagery is unchanged from the EU original. The CTR is 0.4% and they can’t figure out why.
The diagnosis is almost always the same: they paid for translation and assumed they were getting localization. The vendor billed appropriately for translation and delivered exactly that. Nobody lied. The brand just bought the wrong product.
This is a working definition of the gap between the two — what you actually get from each, where each is appropriate, and how to read a vendor’s process to figure out which one you’re getting.
What translation is good for
Translation transfers semantic content from one language to another with high fidelity to the source. It’s the right tool for:
- Legal documents, terms of service, privacy policies
- User manuals, product documentation
- Internal communication between offices
- UGC moderation
- Customer service responses (with translator-in-the-loop)
In all these cases, staying faithful to the source is exactly what you want. The source contains the meaning; the translation propagates the meaning into a new language. Any creative reinterpretation by the translator would be a defect, not a feature.
Performance creative is the inverse situation: the source is one expression of the offer in one cultural context, and the goal in the new market is not faithfulness — it’s effectiveness with the new audience. Translation isn’t built for that.
Where translation fails for performance creative
Five places where translation reliably underperforms in performance contexts:
Tone and register. Japanese has explicit politeness levels (keigo / teineigo / casual). Translation pipelines default to polite/formal forms because they’re “safe.” Most performance creative in Japan needs casual or semi-casual register. Translated copy in polite forms reads corporate and untrustworthy; native copy in the right register reads natural.
Cultural references. Western direct-response leans on idioms (“game changer”, “no-brainer”, “the only X you’ll need”). These either don’t translate (idiomatic loss) or translate literally and read absurd. Native copy uses Japanese performance idioms — different ones, with their own working patterns.
Visual-copy coupling. Banner creative is a unit, not a layered transformation. Copy that translates literally often doesn’t fit the visual it was designed for, because Japanese character density differs from Latin. Native writers adjust copy to fit the layout; translators don’t have authority to do so, so the result either overflows or pads.
CTA conventions. “Get started”, “Learn more”, “Sign up free” each have multiple Japanese renderings, and the right choice depends on offer, audience, and placement. Translators default to literal renderings. Native copywriters default to whatever pulls in your specific segment. The difference can be 30–50% on CTR.
Comparative claims. Western performance creative relies heavily on competitive framing (“better than X”, “the X killer”). This reads as attack-on-competitor in Japan and weakens trust in the speaker. Native copy reframes claims as positive (“loved by X users”, “ranked #1 in Y”) that hit the same conversion target without the cultural penalty.
What localization actually involves
Localization, done right, looks more like commissioning new creative than like processing existing creative. The workflow:
- Read the brief, not the source. The brief is the offer, target audience, goal, and constraint. The source creative is one expression of that brief — useful as reference, not as input.
- Draft natively from the brief. Native copywriters write Japanese copy from scratch, optimized for the Japanese audience and the placement spec.
- Adjust visual treatment. Designers tune layout density, color saturation, type weight, and CTA position for Japanese conventions — even when the brand guideline came from the US/EU original.
- Internal review for cultural fit. Native reviewer reads the asset cold and flags anything that reads foreign. Not a copyedit pass; a does-this-land pass.
- A/B test against in-market peers. The right baseline isn’t “the EU original.” It’s “what other brands in your placement are pulling in Japan.” That’s where the real ceiling is.
Step 1 is the biggest divergence from translation workflows. Translators receive the source file and translate; localizers receive the brief and create. If your vendor’s intake form starts with “upload your existing creative” rather than “describe the offer”, you’re getting translation.
The cost and time differential
Quick benchmarks for a typical Japanese performance banner set:
| Translation | Localization | |
|---|---|---|
| Cost per concept | ~$30–60 | ~$200–500 |
| Turnaround | 24–48h | 48–72h |
| Output ownership | Source-derivative | Original-in-target |
| Expected CTR vs in-market peer | 30–50% | 90–110% |
Localization costs more and takes a little longer. It also performs roughly 2× on CTR against Japanese audiences. For any campaign with meaningful media budget, the math is straightforward: paying 4–8× more for creative that performs 2× better is profitable as soon as your media spend exceeds the production cost differential — usually within the first week of testing.
Signs you’re getting translation when you paid for localization
Five red flags from the vendor’s process:
- Intake starts with files. They ask for your existing creative deck before asking about the offer.
- No native reviewer step. Output goes from translator to client with no Japanese-native creative review.
- Single-language deliverable. You get Japanese copy with no explanation of register choices, idiomatic decisions, or cultural adjustments.
- Visual is unchanged. Imagery, layout, color treatment, and CTA position are identical to the source.
- Pricing is per-word. Per-word pricing is the structural tell of translation work. Localization is priced per-concept or per-asset.
Most overseas brands have at least three of these in their current Japanese workflow. The fix isn’t to add more rounds of translation — it’s to change the unit of work.
If you suspect your Japanese creative is translation in localization clothing, the easiest way to find out is to A/B test a properly-localized concept against your current. That’s exactly what a single one-off project from a localization studio is for: a sample big enough to settle the question before you commit to changing your whole creative pipeline.