Why keyword lists do not translate
The keyword clustering guide states the rule: clusters do not translate, because phrasings, intents and the engine's treatment of them all differ by market. The translation and transcreation guide explains why copy built on translated terms misses demand. This tutorial is the research that replaces the translation.
| What changes across markets | Example of the difference | What it means for research |
|---|---|---|
| The word for the category | A borrowed English term, a brand used generically, or a native word | Seeds come from how the market talks, not from a dictionary |
| How specific people are | One market searches by activity, another by product feature | The clusters, and so the pages, differ |
| The stage of awareness | A category that is mature at home may be new elsewhere | More informational queries and fewer product-level ones |
| What the engine shows | Marketplaces, local retailers, forums or guides | The page type that can compete differs |
| Regional variation | The same language spoken in several countries with different terms | Research per country, not only per language |
Google's documentation on multi-regional and multilingual sites recommends that each page use a single language for its content and navigation and that each language version has its own URL, and its page on localized versions explains how to tell Google which pages are alternates of each other. Neither tells you which words to use on those pages. That is the job of this research.
Before you start
- The home market's clusters as a list of needs: what each cluster is for, its intent and the page that owns it. The prerequisite is the keyword clustering guide.
- The edition's structure decided, as the international hub describes: which country or language, and where the URLs will live.
- A native-speaking researcher who knows or can learn the category, with time agreed and a budget.
- Keyword tools set to the market's country and language, and a way to see that market's results pages.
- Search Console access for the existing site, to see what the market already finds.
- The localisation brief template, which will receive the output: the localisation brief.
The steps
Brief a native-speaking researcher
Give the researcher the needs, not the words: what each home cluster is for, who searches it and what the page must do, with the home market's terms only as background. Ask them to work in their own language from the start, to note every term they would use and every term they would never use, and to flag regional variants, borrowed words and terms with unwanted meanings. Agree the output format, the sheet described in the last section, so the results can be compared and merged.
Build seed terms from the market, not the dictionary
Seeds come from how the market already talks about the category: the navigation and product names of local retailers and competitors, local marketplaces, forums and community groups, reviews in the language, local press and the questions customers from that market have already sent to support. Where the business has customers there, interview a few. Translations of the home terms go into the list only as candidates to test, marked as such.
Expand them with the market's own data
Run the seeds through keyword tools set to the market's country and language, and through autocomplete and related searches as the market sees them. Filter the existing site's Search Console performance report by country: queries from that market, even in the home language, show what it already looks for. Before any market is chosen, the tutorial on reading the demand you already get from other countries applies the same filter with brand queries set apart, so a country's searches can count as evidence in the market scoring. Treat every Search VolumeThe estimated number of times a query is searched in a period, usually per month. figure as an estimate and expect thin data in smaller markets, where many long-tail queriesA specific, usually longer query searched rarely on its own but often as a group. show nothing at all; the low-volume research guide covers how to judge them.
Read that market's results pages
For each candidate head term, look at the results page (SERPThe page of results a search engine shows.) as a searcher in that market sees it, using the search engine's region and language settings or the researcher's own searches. Record which kinds of site rank (local retailers, marketplaces, international brands, publishers, forums), which formats appear and whether the results answer the need well. Two terms that return the same results belong together; a translated term whose results are about something else is a false friend and comes off the list.
Compare intent with the home market
Label each term's Search IntentThe goal behind a search query: to learn, to find a site, to compare or to act. from its results page, then set the labels beside the home market's for the same need. Differences are findings: a need that is commercial at home may be informational in a market where the category is newer, and a product-level query at home may be an activity-level one abroad. Note each difference, because it changes the page that should own the need.
Cluster in the market's language
Group the market's terms by results-page overlap and meaning, as the clustering guide sets out, working only with the market's own terms and results. Do not force the market's terms into the home clusters: where two home clusters collapse into one in the market, or one splits into two, follow the market. Choose each cluster's head term from the market's data and record the dominant format of its results page.
Decide which pages the edition needs
Map each market cluster to a page. Core pages that every edition carries get the market's head terms; home pages with no demand in the market can wait; clusters with no home counterpart get new market-specific pages. Keep each page in one language, as Google's multi-regional documentation recommends. Where the market shares a language with an existing edition, the pages are alternates and the hreflang guide covers declaring them (hreflangAnnotation for language/region versions.); Google's page on localized versions notes that versions whose main content stays untranslated are treated as duplicates, so translate the content, not only the template. The new market launch checklist lists the other checks the edition must pass.
Hand the findings to the localisation brief
Put the approved terms for each page into the brief's keyword field: the head term, the main variants, the intent and the terms to avoid, with the reason. Add them to the termbase so writers and translators use the searched-for words in titles, headings and copy rather than their own synonyms; the tutorial on building a termbase and translation memory shows where the searched-for term sits beside the reviewer's approved term in each entry. After launch, filter Search Console by the edition's country and pages and compare the queries the pages earn with the sheet; the gaps are the next round of research.
A market keyword sheet (illustrative)
A running-gear brand finds its translated categories are not what runners search
Tumblewhin Running Gear, an invented running-gear brand, prepared to enter a market whose language its team does not speak. Its first plan used translations of its home category names. A native-speaking researcher, briefed with the needs rather than the words, built the sheet below. Terms are shown as English glosses; volumes are monthly tool estimates for the market.
| Home category | Literal translation (gloss) | Estimate | What local runners search (gloss) | Estimate | Decision |
|---|---|---|---|---|---|
| Trail running shoes | off-road running footwear | 50 | mountain running shoes | 2,900 | Market term owns the category page |
| Running vests | running waistcoat | 0 | the borrowed English word "vest" with "running" | 720 | Borrowed term in title; translation dropped |
| Compression socks | compression stockings | 1,300 | the same, but results show medical retailers | 1,300 | Product page uses "running" plus the term to separate it from medical use |
| Hydration packs | drinking backpack | 210 | drinking backpack | 210 | Translation confirmed; kept |
| (none at home) | not applicable | not applicable | winter running spikes | 1,600 | New market-specific category page |
Of the four translated category names in the original plan, one was kept unchanged, one needed a qualifier and two were replaced by the market's own terms. The edition launched with five category pages: the four existing categories under the market's terms and one category the home site does not have.
The volumes on the market keyword sheet feed the SEO traffic estimator for international markets, which pairs them with a position read from the market's own results pages to compare markets.
Common mistakes
- Translating the keyword list. The literal translation is often a term the market does not use.
- Research by a non-native speaker with a dictionary. False friends, unnatural phrasing and missed regional variants.
- Forcing the market into the home clusters. The page set should follow the market's results pages.
- One list for every country that shares a language. Terms and intents vary by country as well as language.
- Mixing languages on a page. Google's documentation recommends one language for content and navigation.
- Translating the template but not the content. Versions with untranslated main content are treated as duplicates.
- Stopping at launch. The queries the new pages earn in Search Console are the best research for the next round.
