How to Conduct Market Research Before Launching Your Startup

Last year, a founder I know burned through €18,000 and nine months building a scheduling tool for freelance physiotherapists. He had a landing page, a logo, a waitlist of 340 emails. What he did not have was a single recorded conversation with an actual physiotherapist. When he finally called ten of them, seven said the same thing: they already solved that problem with a WhatsApp group and a shared Google Calendar. He shut the product down three weeks later.

That story is not unusual. It is the default outcome. And it is almost always preventable with three weeks of unglamorous work that costs less than a used laptop. So here is the actual method for conducting market research before launching a startup — the sample sizes, the scripts, the tools, the costs, and the go/no-go thresholds I wish someone had handed me the first time I tried this.

How to do market research before starting a business

The sequence matters more than the tools. Most founders start with market size reports because they're easy to download and feel like progress. That's backwards. You want to start narrow — with humans — and widen only once the human signal is clear.

Four phases, in order

Here is the loop I now run on every project, and the one I recommend to anyone asking how to do market research for a product or a service:

1. Desk scan (2–3 days). Understand the terrain. Who already sells to this audience? What do they charge? Where do these people gather online?
2. Problem interviews (2 weeks). 20–30 conversations about the problem, never about your solution.
3. Demand tests (1–2 weeks). A landing page, a pre-order button, a paid ad, a manual concierge version of the service.
4. Competitive and pricing teardown (3–4 days). Now, and only now, look at competitors in detail — because you finally know which customer you're comparing them for.

Notice what's missing: nobody in this sequence validates the idea. You validate the problem and the willingness to pay. The solution can change ten times. Mine did.

Primary vs secondary research, and what each costs

Secondary research is everything already published: industry reports, company filings, public registries, forum threads, competitor pricing pages. It's cheap and fast and it will not tell you whether anyone wants your thing. Primary research is you talking to, observing, and testing on real people. It's slow and uncomfortable and it's the only kind that produces decisions.

Research type	Typical cost (solo founder)	Time	What it actually answers
Public registries & filings	€0	Half a day	How many businesses exist, how they're structured
Competitor pricing teardown	€0	1 day	What the market already pays, and for what
Problem interviews	€0–€150 (coffee, gift cards)	2 weeks	Whether the pain is real, frequent, and expensive
Survey to 150+ respondents	€50–€300	5–10 days	Rough proportions across a wider audience
Landing page + paid traffic	€150–€800	1 week	Whether strangers click, sign up, or pay
Concierge / manual pilot	Your time	2–4 weeks	Whether the outcome is valuable enough to pay for

The €2,000 ceiling I mentioned isn't arbitrary. It's roughly the point where a pre-seed founder starts spending money to avoid talking to people. Past that threshold, the marginal euro buys you very little additional certainty.

How do startups do market research?

Differently from corporations, and mostly for the better. A big company runs a study to reduce the risk of a decision that's already half-made. A startup runs research to find out whether there's a decision worth making at all.

The interview script that works

Here's the structure I use. It runs 25–40 minutes and it works for almost any B2B or B2C problem:

• Walk me through the last time you dealt with [problem area]. What happened, step by step?
• What did you do about it? What tools, workarounds, or people were involved?
• How much time or money did that cost you, roughly?
• Have you ever looked for a better way? What did you find, and why didn't you switch?
• Who else is affected when this goes wrong?

And here's what you never say: "Would you use a product that does X?" That question measures politeness, not demand. Everyone says yes. Nobody pays.

Where to find 30 people who will actually talk to you

Cold outreach works better than most founders expect, provided you're specific. In my experience the reply rate on a short, personalised message that names the person's actual job and asks for 20 minutes of their expertise lands somewhere around 10–20% — which means 200 messages gets you 20 to 40 conversations. That's a week of evenings.

Reddit and niche forums are a second channel: search for the complaint, then message the people who wrote it. Not to pitch. To ask how they handle it today.

For AI-assisted desk research, the honest verdict is this: language models are genuinely useful for summarising competitor positioning, generating interview questions, and clustering 40 pages of interview notes into themes. They are useless for telling you whether a market exists. And they will confidently invent pricing figures if you let them. Use them as a sorting tool, never as a source.

What is the 80/20 rule for startups?

In market research terms, the 80/20 rule means roughly 20% of your research effort produces 80% of the insight — and that 20% is almost entirely the customer conversations. Not the market sizing spreadsheet. Not the competitor matrix with 14 columns. The conversations.

Concretely, if you have 15 hours for research this month, spend 11 of them talking to people, 2 reading public data, and 2 building a rough test. Founders who invert that ratio — 11 hours in spreadsheets, 2 talking — consistently launch into the void.

The same principle applies to the customer base. A small number of your interviewees will describe the problem with unusual intensity: they've built a spreadsheet workaround, they've hired someone part-time to handle it, they've complained publicly. Those people are your early adopters. The other 80% will nod politely and never buy. Learn to tell them apart in the first five minutes of a call — the tell is whether they describe a workaround they already built, not a frustration they merely feel.

Why does 90% of startups fail?

First, a correction: the "90%" figure is widely repeated and far less solid than it sounds. Depending on how you define a startup, how long you track it, and whether you count shutdowns, acquisitions, and zombie companies, the failure rate ranges across a much wider band. The number gets quoted because it's dramatic, not because it's precise.

The reason behind the failures is better documented than the rate. When founders are asked post-mortem why the company died, the single most common answer is no market need — they built something people didn't actually want. That beats running out of cash, which is usually a symptom rather than a cause. It beats being outcompeted. It beats pricing.

Which is precisely what good pre-launch research is designed to catch. A cash crunch is a lagging indicator; a lack of demand is a leading one, and it's visible months earlier if you look.

The survivorship bias trap

There's a second layer here. The startup stories you read are written by the survivors, and survivors tend to retroactively describe their process as deliberate. The founder who skipped research and won will tell you research is overrated. The founder who skipped research and lost isn't writing blog posts.

I'll admit I fell for this early on. I read three "we launched in a weekend" stories and concluded that speed was the whole game. It took one dead product and about €9,000 of my own money to notice that all three of those founders had years of prior domain experience doing the work research would have done.

The go/no-go criteria you should set before you start

Research without pre-committed thresholds is theatre. Before your first interview, write down what result would make you stop. Then honour it. Here are the numbers I use:

• Problem frequency: at least 6 of your 25 interviewees should describe the problem as occurring monthly or more often.
• Existing spend: at least 3 should be paying for something today, even badly, even if it's a freelancer or a spreadsheet subscription.
• Willingness to pay: on a landing page with real pricing, a conversion to paid pre-order or deposit of roughly 1–3% of qualified visitors is a workable signal. Under 0.5% is a stop sign.
• Concierge test: can you deliver the outcome manually for 5 customers without it consuming your entire week?

If two or more of those fail, the honest move is to change the customer segment or the problem — not to run more research hoping for a different answer. I've watched founders run four rounds of interviews looking for permission to proceed. The data doesn't change. The interpretation does.

When research genuinely is a waste of time

Two cases. First, when you're building something with no existing category and no comparable behaviour — there's nothing to measure because the behaviour doesn't exist yet. Second, when your product is a tool for yourself and your professional peers, and you are the customer. In that case, ship fast and iterate on your own usage. But be honest: "I am the customer" is the most abused sentence in startup land, and it's only true if you'd genuinely pay for the thing out of your own pocket.

For everyone else — for the 95% building something for people who aren't them — the conversations are the cheapest insurance you'll ever buy.

So the question isn't whether you can afford three weeks of interviews before you launch. It's whether you can afford to find out from your bank statement instead. My friend with the scheduling tool got his answer for €18,000 and nine months. Ten phone calls would have cost him an afternoon.