Localize in four layers, keyword research redone in the target language, copy rewritten by a native speaker, images adjusted for local units and norms, and pricing set per marketplace. Machine translation alone fails because shoppers in Germany or Spain search with different words than a translator would pick. Budget one full workday per listing per marketplace.
The short version
- Translation is not localization. German shoppers type search terms a translator would never choose.
- Redo keyword research per marketplace. Demand lives in the local language, not in your English keyword list.
- A native speaker rewrites the copy after the local keywords are locked, never before.
- Images carry half the message. Units, plugs, lifestyle context, and text overlays all need local versions.
- Price per marketplace. Fees, VAT, and freight differ, so a copied price is a wrong price.
Why a translated listing underperforms
Amazon's search engine in each marketplace matches queries against the text indexed in that marketplace's language. When you translate an English listing, you preserve your meaning but lose the query. A translator picks the grammatically correct word. A shopper types the colloquial one. If your indexed text contains the first and the search box receives the second, you are invisible for the exact demand you paid to reach.
That is the whole mechanism, and it is why the order of operations matters. Keyword research comes first, in the target language, using that marketplace's own search data. Copy comes second, written around those terms by someone who natively speaks the language. Translation of your English bullets is not a step in this process at all.
Flapen manages about 70 brands across all 23 Amazon marketplaces and produces listing content in English, German, Spanish, and French in-house, so this workflow runs weekly on our side rather than living in a slide deck.
Score your localization before you publish
A scorecard keeps this honest. Rate each criterion from 1 to 5, multiply by the weight, and add it up. Publish at 80 or above. Below that, fix the weakest row first.
| Criterion | Weight | What a 5 looks like |
|---|---|---|
| Local keyword research | 30 | Ranked terms pulled from the target marketplace's search data, not translated from English |
| Native-language copy | 25 | Written by a native speaker around the local terms, reads like it was born there |
| Image localization | 20 | Units converted, plug types correct, overlay text in the local language, local lifestyle context |
| Pricing and fee model | 15 | Price rebuilt from local fees, VAT, and freight, then checked against local competitors |
| Compliance and conventions | 10 | Local labeling rules met, sizes in local conventions, claims legal in that country |
Two rows deserve a note. Image localization gets skipped most often because it costs studio time, yet the main image is what earns the click. And pricing is not a currency conversion. A marketplace with higher fulfillment fees and VAT needs its own margin math, or you will scale a loss.
The capacity question behind localization
Here is the uncomfortable part. Everything above is manual, skilled work, multiplied by every marketplace you enter. It cannot be batched away with software, which means the real constraint is operator time.
If an agency handles your expansion, ask one question before signing anything, how many brands does each account manager carry. Our ratio is about 1.4 brands per operator, and that number is the reason redone keyword research and native copy are possible at all. An account manager carrying 15 brands does not have a workday per listing per marketplace to give you. They have a machine-translation button, whatever the proposal said.
The same logic applies if you are doing this yourself. One person can localize into two or three new marketplaces per quarter. Plans that assume ten are really plans to translate. Our launch playbook sequences expansion for exactly this reason, and the full method is laid out at Amazon FBA launch.
What most agencies will not tell you
Most global expansion packages are machine translation with a human skim, sold at localization prices. The tell is in the deliverable list. If it does not name per-marketplace keyword research as a separate line item with its own hours, the keywords are being translated, and translated keywords are the single most common reason an expansion underperforms its home market.
The second thing kept quiet is sequencing. Selling you eight marketplaces at once is a bigger invoice than selling you the two where your product's economics actually work. Fee structures, category demand, and return behavior differ by country, and some marketplaces will never clear your margin bar. A good partner shows you the per-marketplace math and tells you which ones to skip. Skipping is the advice that costs them revenue, which is why you rarely hear it.
Related answers
- Which products work well for global shipping
- Step by step Amazon seller roadmap global
- Global lead times and shipping costs for Amazon restock
- Global Amazon marketplaces product research help
- Amazon seller roadmaps and capital: the complete guide
If you would rather have operators who do this weekly run your expansion, start with Flapen.
