A founder showed me a slide last year with eleven countries ranked by addressable market, each with a score out of 100 built from GDP growth, category penetration and regulatory friction. Germany came first. He had budget approved, roughly a year of runway for the move, and one question: how do I know Germany will work before I commit?

I told him the slide was not wrong, it was simply answering a question nobody needed answered. Germans buy his category. They bought it last year and the year before. He could have confirmed that in forty minutes with industry data and a competitor’s annual report. What he did not know, and what no ranking could tell him, was whether he could put his product in front of a German buyer, at his price, for less than his gross margin could absorb.

That is the whole game. Not the market. The channel.

The market is not a testable object

When people say they want to test a market, they usually mean they want reassurance that demand exists. But demand for an established category is a published fact, not a hypothesis. If you sell industrial sensors, logistics software, skincare or heat pumps, somebody in that country is already buying them in volume, and you can find out roughly how much. Spending six months and a quarter of a million euros to rediscover this is not validation, it is procrastination with a budget line.

The genuinely unknown variables are all downstream of demand. Who holds the buying decision in that country, and is it the same role as at home? What does it cost to reach that person? How long does the cycle run? Does your price survive the local intermediary structure, the import duty, the certification, the payment terms, the discount that distributors in that market treat as normal rather than exceptional? None of those are properties of “the market”. They are properties of a specific route to a specific buyer.

I learned this building consumer hardware in China. We started with one product and over several years grew the range to eight. Not one of those launches turned on whether people wanted the thing. They turned on which shelf, which platform, which partner, and what that route extracted from the unit economics before the money reached us. Two products with nearly identical demand profiles performed completely differently because one found a channel with a 20 point lower cost of reach. The market was the same market. The channel was not.

Later, leading digital transformation for a private healthcare chain expanding across mainland China, the same pattern repeated in a sector where you would least expect it. Demand for quality private care in a tier 2 city is not in doubt. What decided whether a new site filled its appointment book was the acquisition route: referral networks, insurance panels, employer contracts, digital. Those four channels had cost-per-patient figures that differed by multiples, and the ranking changed from one province to another. You cannot test “the city”. You test the referral network, and you get a number.

Why the friendly first customer ruins your data

Here is the failure mode I see most often, and it is seductive because it feels like progress.

A founder enters a new country through a contact. A former colleague, an investor’s introduction, a distributor met at a trade fair who was genuinely enthusiastic. Within four months there are three signed deals and a case study. The board is pleased. The founder commits to a local entity, two hires and a year of fixed cost.

Eighteen months later the pipeline behind those three deals is empty, and nobody can explain why.

The explanation is that the three deals measured the contact, not the country. A motivated intermediary with personal credibility can sell almost anything once. What those deals told you is that your product is not actively rejected, which is useful but nearly worthless as a basis for capital allocation. They told you nothing about repeatable cost of acquisition, because the acquisition cost was somebody’s goodwill, and goodwill does not scale and does not appear on a rate card.

There is a second, quieter version of the same mistake: the discounted pilot. You cut 30% to get a reference logo, and you learn that a buyer will take your product at 30% off. You still have no idea whether they will take it at list. Any test run below your real price answers a question you will never ask again commercially.

I apply a simple filter now. If a sale happened because of a relationship that cannot be reproduced by a process, or at a price you do not intend to charge, it is not evidence. It is encouragement. The two feel identical in a board meeting and they are opposite in value.

One channel, one price, one buyer type

The alternative is deliberately narrow, and it frustrates founders because it looks small relative to the ambition.

Pick one channel. Not a mix. If it is outbound to a defined buyer role, then it is outbound only, and the paid search, the trade fair and the partner conversation wait their turn. Mixed-channel pilots produce mixed attribution, and mixed attribution produces a number you cannot act on. You will end the quarter knowing you got twelve meetings and having three plausible stories about where they came from.

Pick one price: the price you intend to charge at scale, in local currency, with local payment terms and local delivery cost already inside it. If that price does not work, you want to discover it in week six on a test budget rather than in year two on a payroll.

Pick one buyer type, defined by role and company profile rather than by sector. “Operations directors at manufacturers with 200 to 800 employees” is testable. “Industrial companies” is not, because you will end up with a scatter of conversations across five different buying processes and no comparable data points.

Then run it long enough to see a full cycle. If your home-market sales cycle is 90 days, a 60-day test in a new country tells you almost nothing except that nobody signed yet. Plan for one and a half cycles, and resource it accordingly.

The output you are after is not revenue. It is three numbers: cost to generate one qualified conversation, conversion rate from conversation to order, and the cycle length. With those, you can calculate acquisition cost per customer and compare it against your gross margin. Everything else in the pilot is colour.

Write the kill number before you write the plan

This is the part almost nobody does, and it is the part that makes a test a test.

Before the first outreach, define the number at which you stop. Something like: if the fully loaded cost of acquiring one customer exceeds 40% of first-year gross profit, we do not enter this market through this channel. Write it down, send it to whoever holds the budget, and date it.

The reason to do this in advance is psychological, not analytical. By month five you will have a local contact you like, a half-built relationship with a distributor, two meetings that went well, and a personal investment in being right. Every one of those creates pressure to reinterpret the data generously. A number agreed when you were still neutral is the only defence against your own future optimism.

Allowable acquisition cost falls straight out of your margin structure, so it is not an arbitrary line. If your gross margin is 45% and your average customer is worth 20,000 euros in year one, you have around 9,000 euros of gross profit to work with, and a CAC above 3,500 to 4,000 euros usually means you are buying revenue rather than building a business. Run those arithmetic before the pilot and the kill number writes itself.

Also define the inverse: what result would make you double the budget immediately. Tests that can only fail tend to be designed to fail slowly.

Treat the first local relationship as the experiment

The last shift is about how you hold your first partner, distributor or country hire.

Founders tend to treat the first local relationship as the prize: the thing that proves the market is open. I would reverse it. That relationship is the single most informative experiment you will run, and you should design it like one. Give it a defined scope, a defined period, a defined volume expectation and an explicit review date. Not because you expect it to fail, but because a partner who cannot agree to a measurable scope is telling you something important about how they intend to work.

In the climate-tech business I run now, this is the discipline I insist on most. A partner who commits to specific numbers over two quarters is a data source. A partner who commits to enthusiasm is a cost centre with good manners.

So the practical change is this. Stop asking whether the country will work. Ask which single route to which single buyer you can test at full price in the next ninety days, what the number is that stops you, and who signs off on that number before you start. You will spend less, you will learn faster, and when the answer is no, you will still have the capital to go and test the next channel.

I write from twenty years of building businesses between Europe and Asia. If your company is facing this, start a conversation.