Translation workflows across marketplaces.

Launching in a new Amazon marketplace is almost never "translate the title and we're done". It is matching titles to the new marketplace's Style Guide, redoing keyword research in the local language (because direct translation rarely matches actual local search behaviour), complying with local certifications and forbidden-word lists, localising imagery where culturally relevant, and — the most-missed risk — not silently breaking the source listing through Amazon's own translation-sync tooling. A multi-marketplace catalogue done well is a force multiplier; done poorly it leaks revenue in every locale simultaneously and the damage compounds because the same mistake repeats across five PDPs at once.

The three translation workflows

• Manual translation. Export the source listing through the reverse feed (Episode 12), hand off to a translator (ideally one who understands Amazon Style Guides and the per-locale forbidden-word lists), receive the translated TSV back, upload to the target marketplace as a PartialUpdate. The highest-quality path, the most expensive in money and time, and the most reliable in terms of preserving the work. The right answer for hero SKUs, brand-defining catalogues, and any product where conversion lift on the translated listing materially exceeds the translator fee.
• Build International Listings (BIL). Amazon's own auto-sync tool that mirrors a source listing into target marketplaces with machine translation. Free to use, fast to set up, mediocre translation quality, and dangerous: BIL can silently overwrite manually-translated bullets every time the source listing changes. The right answer for long-tail catalogues where you'd rather have mediocre translation than no translation, and where you commit to BIL as the canonical sync mechanism (never half-on).
• Translation Memory (TM) tools. Smartling, Lokalise, memoQ, Phrase. Higher quality than BIL with better consistency than ad-hoc human translation, full traceability for compliance audits, and reusable translation memory that compounds in value over time. Requires integration work (typically API-based) and a recurring SaaS fee. The right answer for catalogues above ~100 SKUs across 3+ marketplaces where translation work is a continuous operational rhythm rather than a one-off project.

What changes per marketplace

• Style Guide. Title formula, byte limits, capitalisation rules, allowed punctuation differ per locale. A DE Style Guide may require sentence-case titles where the US Style Guide requires title case. The IT Style Guide may forbid colons where the FR one allows them. Always pull the destination-marketplace Style Guide before translating.
• Forbidden words and claims. Wellness, medical, environmental, and cosmetic claims vary dramatically across locales. "Anti-aging" is acceptable on US Beauty listings with substantiation; the same phrase is forbidden in FR and IT regardless of substantiation. "Boosts immunity" is fine in some US categories; banned in EU under the Nutrition and Health Claims Regulation. Maintain a per-marketplace forbidden-word list as a translator-facing brief.
• Keywords. Direct translation almost never matches local search behaviour. The German for "yoga mat" is "Yogamatte", but the high-volume head term in DE Beauty also includes "Gymnastikmatte" and "Sportmatte" — synonyms with overlapping intent that English-only research never surfaces. Re-run keyword research in the local language with native-speaker validation for every marketplace launch. Module 6 of the Listing course covers this in depth; for the translation workflow, the rule is "translator drafts the prose, native-speaker keyword researcher reviews the title and bullets for keyword presence".
• Certifications and compliance. CE / RoHS / REACH for EU, FCC for US, PSE for JP, RCM for AU, NOM for MX. Required certifications must be visible on the listing as attributes and frequently on the product imagery itself.
• Currency and pricing-display conventions. "$" vs "€" vs "£" vs "¥" handled by Amazon at the offer level, but any pricing references embedded in image copy or A+ Content need re-rendering per locale.
• Measurement units. Imperial vs metric — every dimension, weight, capacity, and temperature reference in the listing copy and the imagery has to translate as well as the language. A US apparel listing showing "32 inches" needs to become "81 cm" in EU markets; failing to convert is a frequent return-driver.
• Cultural localisation of imagery. Hero lifestyle shots that work in the US (a American suburban kitchen) read as foreign in DE; on-model apparel imagery often needs reshooting per region for cultural fit. The level of investment here scales with category — fashion and beauty demand local imagery; commodity hardware rarely does.

Build International Listings — the BIL trap in detail

BIL is the most-used and most-misunderstood translation tool in Amazon. The mechanics:

1. You select a source marketplace (typically your home marketplace) and target marketplaces.
2. BIL syncs the source listing's attributes (title, bullets, description, images) to the target marketplaces, running them through Amazon's machine-translation pipeline on the way.
3. BIL also syncs pricing (with FX conversion and configurable margin uplift) and inventory (when paired with multi-marketplace FBA).
4. Every time the source listing changes, BIL re-syncs to the targets automatically — typically within hours, sometimes within minutes.

The trap is in step 4. BIL does not respect manual edits made on the target marketplace listings. If you carefully wrote a German bullet that uses the local high-volume keyword "Gymnastikmatte" instead of the literal translation of the English source, BIL will overwrite that bullet with its own machine-translated version the next time the English bullet changes — and the change can be as small as a typo correction. The result: weeks of translation work silently destroyed on a Tuesday afternoon when a teammate fixed a typo in the source listing.

The defence: either commit to BIL fully (accept that machine translation is the canonical version on target marketplaces and stop manually editing them), or disable BIL entirely and translate manually. Half-on is the failure mode. Half-on is also where most multi-marketplace catalogues end up by accident because BIL is enabled once during a launch and forgotten about for years.

Workflow for a clean multi-marketplace launch

1. Confirm regulatory readiness. Certifications, hazmat declarations, country-specific labelling, VAT registration in countries where you'll hold inventory.
2. Pull the reverse feed for the source-marketplace SKUs (Episode 12).
3. Do local keyword research per target marketplace with native-speaker review. This is the input to the translation brief, not an output of it.
4. Translate manually or via TM tool. Brief the translator with the source listing, the local keyword priorities, the Style Guide rules, the forbidden-word list, and any brand-voice constraints.
5. QA the translated listing against the destination-marketplace Style Guide before upload. Catch byte-limit overruns, capitalisation issues, and forbidden-word violations at this step rather than at suppression.
6. Upload as PartialUpdate through the destination marketplace's flat-file pipeline (Episode 10).
7. Verify the live listing against the reverse feed in the destination marketplace — confirm that the stored values match what you uploaded.
8. Decide on BIL. Either enable it pointing at the now-translated target listing (which becomes the source for forward syncs in the other direction, a less common but valid pattern), or leave it off.

Common multi-marketplace failure modes

• BIL silently overwriting manual edits. Covered above.
• Direct translation that ignores local keyword behaviour. Conversion underperforms forever; the brand thinks it's a market-fit issue when it's actually a keyword issue.
• Unit conversion missing in imagery. Returns spike on "wrong size received" complaints that are actually unit-confusion issues.
• Certifications attribute populated for one marketplace, missing for another. Listing approved in DE, suppressed in FR.
• A+ Content not localised. A+ documents are per-marketplace. Publishing the source A+ document on the target marketplace shows English copy to French shoppers; conversion drops measurably.
• Brand Story not localised. Same issue as A+ — the carousel is per-marketplace and needs translation per locale.
• Variation-theme name mismatch across marketplaces. A "color" variation theme in EN can become "Couleur" in FR, but if the underlying attribute value drifts (Red vs Rouge), the variation family integrity breaks.
• Translation drift over time. Source listing edits don't propagate; target marketplaces gradually lose parity with source. Quarterly cross-marketplace audit using the reverse feed catches this.

Seller vs Vendor — translation paths differ

Sellers self-service translation through the workflows above, with BIL available as the auto-sync option. Vendors do not use BIL — Vendor-to-Vendor cross-marketplace replication runs through internal Amazon catalogue-team workflows that the Vendor cannot directly configure. Vendor brands launching in a new marketplace typically work with their Vendor Manager to set up the localisation, or run a parallel Seller account in the target marketplace specifically to control the translation themselves.

What to take into the next episode

Translation handled, the next episode wraps Module 4 with the one-page end-to-end content-creation checklist that pulls all sixteen episodes into a single repeatable workflow.