Pick your next 10 markets without guessing: the expansion blueprint

What you put in, and what you get out

You put in one thing: your product feed — the same Google Merchant Center file you already have. You get back one thing: a ranked table of countries, top to bottom, telling you where to expand next and why — backed by real search demand, real CPCs, real competitor prices and a simulated return on ad spend for each market. No gut feeling, no “Germany because it’s big”. A spreadsheet that overrules the meeting.

This article is the whole blueprint, and I’m not going to wave my hand at any of it. At every step I show you the real intermediate output — the actual table, the actual counts, the actual seed that came out the other side — from one overnight run on a real bargain-segment Czech e-shop. So you can see, step by step, exactly what each stage produces and decide “yes, that’s the thing I need”.

A word on why this is worth doing at all. The honest version of an expansion analysis — demand, prices, competition and unit economics across ten or fifteen markets — was always possible. It just meant hundreds of hours of pulling keyword volumes country by country, translating seeds, eyeballing competitors, and building a spreadsheet nobody fully trusted. So it mostly didn’t get done; teams picked their next country because someone spoke the language or a distributor called. That manual cost is what collapsed — not the difficulty of the idea, the difficulty of the execution. What used to be weeks is now one overnight run plus an afternoon.

The blueprint at a glance

Seven steps. For each one: what you do, what you’re looking at, and why — because every step exists to answer one specific question about a market.

Now the same seven steps, one at a time, each with the real thing it produced on this run.

Step 1 — The shortlist the logistics radius produced

You don’t analyze all of Europe; you analyze where you can deliver. Starting from the home market’s shipping and customs reach, the candidate set came out to 13 target countries across Central and Eastern Europe and the Balkans. Here is that literal shortlist, with the entry signals the scoring uses later — the home market (Czechia) sits in the table too, but only as the calibration baseline, not a target. And note that two countries get flagged out before a single keyword is pulled, for a reason no slide deck predicts:

That last column is the kind of constraint you only learn by running it: Google Ads doesn’t support Bosnian or Macedonian for keyword ideas, so those two markets fall out of the keyword stage no matter how attractive their growth looks. Thirteen planned, eleven the API can actually research.

Step 2 — The feed, collapsed to 400 seeds

The raw material is the product feed — roughly 252,000 catalogue items, with product_type paths glued together by &, a and slashes the way real feeds always are (“Fridges and freezers”, “Toys & games”). The parser splits those into atomic categories, drops generic top-level nodes (“home”, “electronics”, “garden”), and keeps only paths that clear a frequency threshold. A quarter-million items collapses to a tight, meaningful seed list:

And the seeds themselves are exactly what you’d query a market on — categories, not product names. The heaviest ones by catalogue weight:

To make this concrete: say the shop sells small home goods, toys and phone accessories. Its feed has a path like Electronics > Mobile > Phone cases. The parser throws away “Electronics” (too generic), keeps “phone cases” as a seed, and counts how many products sit under it — here, “cases” tags 14,411 items, so it’s clearly a real category worth researching, not a one-off. Product names mostly have zero search volume, so they’re ignored; the 40,000-plus catalogued brands get held back as a separate signal, and the category seeds carry the keyword research.

Step 3 — Every seed, in the market’s own language

This is the step people skip, and it quietly poisons everything downstream. You never query a market in English. Before a single API call, an LLM translates the full seed list into each country’s native language — the 400 core seeds, expanded with synonyms and spelling variants to 488 entries × 12 languages = 5,856 machine translations sitting ready. Here is one seed, “toys”, as it actually went out to each market:

Why this matters in plain terms: if you researched Poland with the English word “toys”, Google would hand you the volume and CPC for English-language searches in Poland — a tiny, weird slice of expats and tourists. Query zabawki instead and you get the actual Polish market. The numbers are different by an order of magnitude, and every downstream decision rides on them. Brand names are the one exception — those never get translated.

Step 4 — 1.4 million keyword ideas, per market

Now the heavy lifting, and the first thing you actually download: each translated seed, in each country, through the Google Ads GenerateKeywordIdeas endpoint, with a rate limit and the link from every keyword back to its source product_type preserved. Across eleven researchable countries the run produced 1,402,486 keyword ideas — and the per-market split is itself a finding, because raw demand does not track the markets you’d expect to win:

The run is bookkept seed-by-seed, country-by-country, so you can see exactly how complete it is: 4,220 seed-country jobs attempted, 3,619 returned data, 601 errored — quota hits and the unsupported-language aborts from step 1. That’s the honest texture of a real run; it isn’t one clean sweep, and you want the log that proves which cells actually have data behind them.

Step 5 — AI cleans the scope (at the keyword level)

A broad seed drags in noise, and the temptation is to kill the noisy seed. Don’t. “Football” in Germany pulls millions of Bundesliga searches that have nothing to do with your shop — but the same seed also pulls “football boots”, which you very much want. So you filter at the keyword level, not the seed level: throw away the seed and you throw away its buying-intent children.

Here is what the cleanup actually removed on this run, and why — out-of-segment categories first, then pure noise patterns:

A concrete example of the keyword-vs-seed rule: the seed “auto-moto” is great for this shop — it sells car accessories. But in German it also pulls “gebrauchtwagen” (used cars) and “auto kaufen” (buy a car), which are worthless to a small-goods e-shop. You don’t delete the seed; you delete those specific keywords and keep “Handyhalter Auto” (phone holder for cars). That’s 2,100-plus junk keywords gone, the good ones untouched.

And the flip side of cleanup isn’t only subtraction — it’s tagging the signal you do want. For a bargain / second-hand catalogue, the buying intent lives in the deal words, so 49,770 keywords got tagged as bargain-intent in eleven languages:

This is the step that separates a usable deliverable from a noisy one. On the first pass the spreadsheet was full of Bundesliga, 2026 calendars and used-car listings; it took a two-pass review across ten languages to get to a core a client would actually trust.

Step 6 — Pull the competitive page out of a SERP

Volume tells you demand; it says nothing about whether you’ll get crushed on price. So for the top 20–30 % of best-selling products per category, you localize the title and scrape the live results page. One POST to DataForSEO returns the entire page as JSON — the part you can’t reconstruct from your own account:

import requests

# One POST → the whole results page as JSON:
# paid ads (competitor copy), shopping blocks with prices, organic, PAA.
resp = requests.post(
    "https://api.dataforseo.com/v3/serp/google/organic/live/advanced",
    auth=(LOGIN, PASSWORD),
    json=[{
        "keyword": "lodówka do zabudowy",   # localized to the market's language
        "location_code": 2616,               # Poland
        "language_code": "pl",
        "device": "desktop",
    }],
)
# ~$0.0035 per query → 3,000 top products per market ≈ $10.50

The same response carries three things for one price: the paid block (competitor ad copy — raw material for your messaging research), popular_products (shopping blocks with live prices you can’t get from your own Merchant Center), and the organic results (the market’s real players, including the ones who don’t advertise). You pay for a price check and walk away with a copy library and a competitor map. And the whole-program cost is the punchline:

A full competitive SERP pass across twelve countries runs $22–126 depending on mode — against a Semrush subscription at ~$140+/month flat just for the UI.

Step 7 — Simulate the economics, then score

Now you fold everything together. From real CPC, a defensible conversion rate and a per-market AOV you simulate ROAS for every country, then weight demand, prices, logistics, growth and entry friction into a single score. The output is one ranked table — the deliverable you came for — and its punchline is reliable: the countries that come out on top are almost never the ones the team expected.

Look at the top three. Slovakia, Poland and Serbia — not Germany, not Austria, the two markets everyone defaults to. Germany has the most demand in the set (348k keyword ideas) and lands fourth, dragged down by a simulated ROAS of barely 1.0. The reason is the hard penalty that keeps the table honest: when simulated ROAS drops below the viable threshold, the score is multiplied by 0.4 — because on a fresh market paid is your only channel, you have no SEO yet, and a market where paid doesn’t pay can’t be your entry point.

The scoring model: nine weighted factors

The score isn’t a vibe. It’s nine components, each weighted by how much it actually moves the decision. ROAS dominates, because a huge market with brutal CPCs loses to a small market with cheap clicks every single time.

Two hard penalties keep the table honest: if simulated ROAS drops below a viable threshold the score is multiplied by 0.4 (paid acquisition doesn’t work and on a fresh market paid is your only channel); and if a country returns no CPC data at all, its score is cut to a third — you don’t bet on a market you can’t price. That second penalty is exactly why the two markets dropped at step 1 (no Google Ads support for their language) never reach the ranking above: no keyword data, no price to model, no defensible bet.

The CPC spread, measured

That ranking is built on one factor above all others — the cost of a click. Here is the real thing: the median planner CPC per market from this overnight run, across 745,712 cleaned, priced keyword ideas pulled via the Google Ads API (GenerateKeywordIdeas, run 28–29 April 2026). Every figure is in CZK, the account’s own currency, so the markets compare cleanly.

The spread is the whole argument against defaulting to the obvious market: an Austrian click costs 5.1× a Serbian one and a German click 4.5×, before a single unit of basket size enters the math. The cheapest clicks sit where the meeting never points — Serbia first, then Poland and the Balkans — while the familiar Western markets carry an auction premium most teams underestimate. (Albania is left out of the table: with only 31 priced ideas its median isn’t worth trusting.)

One honest caveat — the same one the callout makes. These are planner estimates, and planner CPC runs high: the real Czech home account behind this run paid 2.54 CZK per click across 22.6 million search terms, below even the cheapest planner figure here. So read the column as relative market pressure, not the price you’ll actually pay.

Bonus output: the seasonality you’d plan budget around

The same keyword pull carries twelve months of search volume per keyword — 16.7 million monthly data points across 1.39 million keywords on this run. Aggregate it and the demand curve is unmistakable: a single peak month carrying 40 % more search demand than the quietest one.

December tops the curve and July bottoms it — exactly what you’d expect for a general-merchandise catalogue. That single swing is your launch-timing and budget signal: you don’t open a new market into its dead month, and you load spend before the December run-up, not during it.

Verified, not theorized

Every number above came out of one real run. Here is the headline set, with two more cross-border analyses for scale:

Two markets had to be dropped mid-run because Google Ads doesn’t support their language for keyword ideas — the kind of constraint you only learn by actually running the thing. The ~10 hours is machine time: the research grinds overnight across all countries in parallel while you sleep. The human work — shaping the feed, calibrating the economics, cleaning the noise, building the deliverable — is a few focused hours on top. What used to be weeks of manual labor is now one overnight run plus an afternoon.

(The −86 % cut is from a different, far more niche catalogue — a B2B construction-profiles shop where most broad-seed keywords were genuinely off-topic. The bargain shop above only shed a single-digit percentage to scope and noise. How aggressive the cleanup is depends entirely on how niche you are.)

Now the fun part: what you do with the ranking

1. You enter the market the spreadsheet picked, not the one you assumed

The whole point is to be surprised. On this run, Slovakia and Serbia — a small market and a non-EU one nobody had pitched — outscored Germany and Austria outright, purely because clicks were a fraction of the price and the economics actually closed. That’s the deliverable doing its job: overruling the gut with the math.

2. The competitive scrape becomes your launch playbook

Because step 6 already pulled every competitor’s ad copy and price point per market, you don’t walk in blind. You know who’s there, what they charge, which benefits they push in their headlines, and where there’s a price gap to undercut or a positioning gap to own — before you spend a single euro on traffic.

3. Demand per category tells you what to ship first

You kept the link from each keyword back to its product_type. So you don’t just know Poland is good — you know which categories Poland searches for, in what volume. The catalogue you launch with is the one the data says the market wants, not a copy-paste of your home assortment.

You are a keyword-research assistant. Input is one raw product_type path
from a Google Merchant Center feed, e.g. "Home & Garden > Fridges and freezers".
Return atomic, searchable category seeds in {LANGUAGE}, one per line:
- split glued categories ("fridges and freezers" → "fridge", "freezer")
- drop generic top-level nodes ("home", "garden", "electronics")
- expand each to 2–3 synonyms people actually search for
- never translate brand names

curl -s "https://api.dataforseo.com/v3/serp/google/organic/live/advanced" \
  -u "$LOGIN:$PASSWORD" -H "Content-Type: application/json" \
  -d '[{"keyword":"lodówka do zabudowy","location_code":2616,"language_code":"pl"}]'

FAQ

CTA: Wondering which markets actually pay back? Let’s run the blueprint on your feed.