International keyword research starts with native seed terms and per-country database pulls, then sizes demand against revenue. Nic's five-step method.
Size the market before you build a page for it. International keyword research is the sizing step, and done properly it hands you three numbers: how much demand exists in the local language, how much of it you can realistically take, and what that share is worth in revenue against the cost of entering.
The moment that made this real for me was small. A client, a sensible team who knew their home market well, had several thousand monthly searches for their service with “near me” attached and not a single page built for it. The local data had never been put in front of them, so the demand had never come up in a planning meeting. If that can sit unseen in a market you know, it will certainly sit unseen in one you have never sold in, which is why a country you haven’t entered yet deserves the maths before anyone commissions a translation.
This piece covers why translated lists fall short on their own, the five steps I run, where per-country volume data comes from and what each source is good for, how the workbook is scored, and what a finished sizing looks like.
The short version
- A translated keyword list is a seed, not a plan. Local buyers often use completely different words (the Dutch term for accounting software search is boekhoudprogramma, not the literal translation), so a native speaker has to rebuild the vocabulary before any tool is run.
- Five steps make up the process: native seed terms, per-country database pulls, intent clustering per market, competitor evidence from the local SERP, then converting demand into a revenue estimate against the cost of entering. The numbers say wait more often than they say go.
- This sizing work is a fixed piece of Clear Click's process for B2B and SaaS companies entering a new market, covered in detail below.
Can I just translate my UK keyword list?
No, though it is a natural first move: the UK list took months to build and feels like a head start. It still is one, as the seed for step one below. What it can’t do on its own is tell you what buyers in that country type into Google. A translated list shows you your terms in another language, and the two differ far more often than anyone expects.
Take accounting software. The direct Dutch translation is "boekhoudsoftware", and it does get searched. But the term Dutch small businesses lead with is "boekhoudprogramma", literally "bookkeeping programme", a word no UK translator would pick because it sounds dated in English. A fair chunk of that market also searches in English. So one UK head term becomes three Dutch demand pools of very different sizes, and the translated list sees one of them. Search demand is built from local habit rather than from dictionaries. In plain English, the market already has words for what you sell, and the job is to find them.
The same trap runs inside a language. Spanish for Spain and Spanish for Mexico split on everyday nouns. German buyers use English loanwords for some software categories and refuse them for others. Translate straight across and you inherit every one of those splits, then spend twelve months ranking for phrases the market doesn’t use. Hope is not a strategy, and a translated spreadsheet on its own is hope with column headings.
What are the five steps we actually run?
I will say this up front: the order matters. Skip a step and the number at the end is fiction.
1. Native-speaker seed terms
Start with 20 to 40 seed terms from someone who buys or sells the product in that market, not a translator. We ask how they'd describe the problem, what they'd type, and what local competitors call the category, then check autocomplete on the local Google domain and the navigation labels on local competitor sites. That gives you the vocabulary. The tool comes after.
2. Per-country database pulls
Run every seed through SEMrush or Ahrefs with the database set to the target country, and through Google Keyword Planner with location set to the country and language set to the local language. Both settings, every time. The slip we see most often in workbooks we inherit, and it is an easy one to make because the tool won’t warn you, is a UK database pull with the keywords in German, which returns numbers describing German speakers in Britain. Export everything, including the terms with no volume.
3. Cluster by intent, per market
Group the pulled terms into clusters around a buying intent, then tag each cluster BOFU, MOFU or TOFU using local phrasing rather than your home funnel. Price modifiers dominate commercial searches in some markets, provenance modifiers in others. A native reader labels intent in an hour; a tool can only infer it from the words.
4. Competitor evidence from who ranks locally
Pull the top ten for the primary term in each cluster from the local Google and classify every result: local specialist, global brand, marketplace, directory, government or media. Three or more local specialists on page one means real competition. Directories and translated corporate pages mean an open door. Difficulty scores come from a global backlink model, so they describe link strength in general; the local SERP shows who is actually winning. Read the SERP.
5. Demand-to-revenue sizing
Add up the BOFU and MOFU cluster volume, apply a share you could realistically hold in year one, apply your conversion rate and your average contract value, and put the result next to the cost of entering. That's the output. Everything above it is working.
Where does search volume by country actually come from?
Just so you know, for reference: there is one first-party source and everyone else models on top of it.
Google Keyword Planner is the first-party source. Set location to the country and language to the local language and the volumes are Google's own. Two things to know. Without active spend it shows ranges (1K to 10K) rather than figures, and it merges variants, plurals and misspellings into one bucket, so head terms look bigger than any single phrasing and you can’t see which spelling buyers use.
SEMrush has the widest coverage, over 140 country databases, which is why it's our default for sizing. Volumes are Keyword Planner data blended with clickstream and then modelled. The catch is depth: the German and French databases are deep, but smaller markets and long-tail terms are thin, and a zero in a small database often means "not sampled" rather than "not searched".
Ahrefs shows a lower number than Keyword Planner for most terms because it splits the variants Google merges, and its twelve-month average smooths seasonal spikes. The feature I use most is the by-country breakdown on a single keyword, which answers "where in the world is this searched" in one click. Good for the first pass, not the final sizing.
Google Search Console is the only source that describes your own demand, filtered by country. It covers only markets you already appear in, but if English pages are picking up impressions from Spain today, that is free evidence, and it counts things the tools show as zero. Visibility without volume still counts.
Google Trends gives relative interest, never absolute volume. Use it for seasonality and regional splits inside a country. Outside Google, Yandex Wordstat, Naver and Baidu's planner are the first-party equivalents.
Realistically, no single source is complete, and none of them claims to be. We pull two and treat any disagreement over 30% as a prompt to check with a native speaker.
How do you build a multi-country keyword workbook?
One workbook, one tab per market, identical columns on every tab, so markets compare side by side. Ours carries: market, language, cluster, keyword in the local language, literal English gloss, intent tag, country-database volume, second-source volume, KD, CPC, who ranks (local, global, marketplace, directory), our current position, mapped page (hub or spoke), status (new or refresh), and priority score.
The score is what makes the sheet useful. We calculate it per cluster: volume weighted by intent (BOFU counts triple, TOFU once), multiplied by a winnability rating from the SERP read (1 to 3), divided by the number of pages you'd need to build. Sort on it and your P1 clusters fall out. P1 gets the first quarter, P2 the second. P3 are the zero-volume terms for things you sell that nobody types yet, and they still get a page once P1 is live, because the visibility is worth having.
The English gloss column looks trivial. It's what lets a UK stakeholder sign off a cluster they can't read. Keep it.
What does a good market sizing look like?
Here is the shape of the output, with numbers I've made up to be plausible for a UK accounting software company weighing the Netherlands. They are illustrative only.
- Cluster A, "boekhoudprogramma" and variants: 9,900 searches a month, BOFU, four local specialists on page one. Winnability 1.
- Cluster B, "boekhoudprogramma zzp" (freelancer bookkeeping software): 2,400 a month, BOFU, a mixed SERP with two directories. Winnability 3.
- Cluster C, "online boekhouden" comparisons: 4,400 a month, MOFU, comparison sites and one bank. Winnability 2.
Total commercial demand: 16,700 searches a month. A year-one share weighted towards Cluster B, where the door is open, is 4%, so 670 visits a month. At a 2% visit-to-trial rate that's 13 trials; at 30% trial-to-paid, 4 new customers a month. At £1,200 annual contract value that's £4,800 of new ARR a month, roughly £57,000 over the first twelve months if nothing improves.
Now put the cost of entry beside it: native content for three clusters, localisation, hreflang, a year of building, call it £45,000 to £60,000. Year one is break-even at best. Year two, with retention and a P2 wave, is where it pays. That is a sensible market to enter on a two-year horizon. To be fair, it is also the answer the sizing gives more often than not: the numbers say wait more often than they say go. That is worth knowing before the budget is committed, and a wait is rarely a never: it usually means a smaller first wave or a longer horizon, either of which is cheaper than going in on a feeling.
How does this feed the wider international SEO plan?
The workbook decides whether and where. Structure decides how. Once the P1 clusters exist for a market they drive three downstream decisions: which pages get built first, which domain or folder setup carries them, and which language and region pairs the hreflang has to declare. The structure side is in our international SEO guide, the tagging in the hreflang implementation guide, and the multilingual SEO for Spain piece shows how one language splits into several demand pools inside one country.
This kind of sizing work looks different again for a SaaS business selling one product into several markets at once. Read about our B2B and SaaS SEO work.
Questions we get asked
How do I find search volume by country?
Set the country in the tool before you type the keyword: the database selector in SEMrush, the country in Ahrefs, location plus language in Keyword Planner. Then pull the same term from a second source. If you only want to know where in the world one term is searched, the Ahrefs by-country breakdown answers it.
Can I just translate my UK keyword list?
Use it as a seed list, never as the plan. Have a native speaker rewrite it in the words buyers use, then pull the country data on the rewritten list. Expect a third of your original terms to fall away and the real head term to be something you would not have guessed.
Which tool has the best international keyword data?
Keyword Planner for absolute volume in Google markets, if your account has spend and you can live with merged variants. SEMrush for coverage across the most countries. Ahrefs for variant splitting and the per-country view of one term. We run SEMrush and Keyword Planner side by side and use Ahrefs as the tiebreaker.
How many keywords do we need before entering a market?
Keywords are the wrong unit for that decision, though it is the question everyone starts with. You need three to five commercial clusters whose combined sized revenue clears the cost of entry inside a period your board accepts, usually 24 months. That typically sits on 200 to 400 keywords, but a market with 60 well-clustered BOFU terms and open SERPs beats one with 2,000 terms and four local specialists on every page one.
How do you research keywords in a language you don't speak?
Not alone. A native speaker supplies seeds and labels intent, the tools supply numbers, and the English gloss column keeps you in the loop. Budget a day of a native speaker's time per market, half for seeds and half to read the SERPs with you. It's the cheapest line in the project and the one most often cut.
Where we fit
We do this for B2B and SaaS companies entering their second, third or fifth market, usually from the UK, and the first deliverable is the sized workbook. Sometimes it says wait, and we'll say so. Our work with Hubject, a multi-market B2B business, produced 610% acquisition growth and saved around £40,000 in budget modelling; the case study has the detail.
610%: The acquisition growth Clear Click helped Hubject achieve, a multi-market B2B business, by sizing markets properly before entering them.
Source: Clear Click case study, Hubject.
If you have a market in mind and a translated list on your desk, the sizing is a fixed piece of work under our market opportunity analysis, and the build sits in international SEO. Get in touch with the market and the product and we'll tell you what we'd pull first.
