How to Conduct Market Research for New Business Ideas That Win

You have an idea. You think it's good. But "you think" is worth exactly nothing until a stranger opens their wallet. The gap between a business idea that feels brilliant in the shower and one that survives contact with the market is almost always crossed by the same unglamorous tool: honest market research.

Here's what most guides get wrong. They describe market research as a single big project you do once before launch — a report, a spreadsheet, a slide deck. In practice, the useful version is a short sequence of cheap tests designed to kill your idea fast if it deserves to die. That's the whole game. Not validation, but disconfirmation.

This piece walks through how to conduct market research for new business ideas in a way you can actually execute in a few weeks, on a small budget, without a research background.

How to conduct market research for new business ideas without wasting months

Most failed research projects I've seen share one flaw: they started with "let's learn about the market" instead of "let's test this specific belief." A general market scan produces a document nobody reads. A targeted test produces a decision.

Start with a falsifiable hypothesis

Write your idea as a claim you could prove wrong. Not "small businesses want an easier invoicing tool" — that's a wish. Something like: "At least 15% of independent consultants I can reach will pay €19/month for an invoicing tool that auto-chases late payments."

Notice what the second version does. It names a population, a price, a threshold, and a behaviour. That's a hypothesis you can kill in a week.

The sequence that works

1. Write the hypothesis with a numeric threshold.
2. Pick the cheapest method that could disprove it.
3. Define your sample before you start recruiting.
4. Collect data and record it raw — no interpretation during collection.
5. Interpret against the pre-set threshold.
6. Make a go, pivot, or kill decision and write down why.

Sound mechanical? It is. That's the point. The moment you start judging results while collecting them, you begin rationalising.

What are the 5 P's of market research?

The 5 P's are a framing device for what to investigate: People (who your customer is), Product (what they need it to do), Price (what they'll pay and how they compare alternatives), Place (where they discover and buy), and Promotion (how they hear about you).

Treat the framework as a checklist, not a method. It tells you the dimensions; it says nothing about how to measure any of them. That's where most beginners stall — they can recite the five P's and still not know what question to ask a potential customer.

What each P actually means in practice

P	Question to answer	Cheap way to test it
People	Who has this problem badly enough to fix it?	5–10 interviews with people who match your description
Product	What must it do to be worth switching for?	Walk through a rough prototype; watch where people hesitate
Price	What will they actually pay?	Pre-sale page with a real checkout or a deposit
Place	Where do they look for solutions?	Ask directly: "Where did you find the last tool you bought?"
Promotion	Which channel reaches them profitably?	Small paid ad test — €100–200 across two channels

The table is not a plan. It's a way to notice which P you've ignored. In my experience, Place is the one people skip, and it's the one that quietly kills otherwise solid products — you build for a customer who never looks where you're selling.

How to do market research as a beginner

You do not need a research methodology course. You need discipline about three things: who you talk to, what you ask, and how you record it.

Primary versus secondary, and which to do first

Secondary research is what already exists — competitor pricing pages, review sites, industry forums, public filings, job postings, Reddit threads. It's free and fast and it should come first, because it tells you what questions are worth asking.

Primary research is what you generate yourself: interviews, ad tests, pre-sale pages, prototype sessions. It's the only thing that tells you about your customer rather than the market in general.

A rough rule I use: one afternoon of secondary research (competitors' pricing pages and the last six months of a relevant forum) before any primary work. It saves hours of asking questions with answers already published.

The beginner's interview mistake

Do not ask "would you buy this?" People are polite. They'll say yes to avoid hurting your feelings, and then not buy.

Ask about behaviour instead: what they did last time they faced the problem, what they currently use, what it costs them, and — this is the important one — what they tried and abandoned. Someone who's already cycled through three solutions has a real problem. Someone who's never looked is not a customer.

From my own experience: a founder I worked with ran eight interviews for a hiring tool. Six said the idea was interesting. One had actually built a spreadsheet workaround and spent four hours a week maintaining it. Guess which one became the design partner. The other six never responded to a follow-up email.

Can ChatGPT do market research?

It can help with parts of it. It cannot do the research. The distinction matters because a surprising number of founders in 2026 are skipping real customer conversations and asking a language model instead.

Here's what a model does well: drafting interview scripts, summarising transcripts you've collected, structuring your hypothesis, generating a list of competitor categories to investigate, and stress-testing your assumptions by arguing against them.

Here's what it does badly: it does not know your market's actual willingness to pay, it invents plausible-sounding numbers if asked for them, and it has no access to conversations that haven't been written down. Ask it "how big is the UK market for X?" and you'll get a confident figure with no traceable source. That figure is a liability in a business plan.

A safe way to use AI in your research

• Use it to draft interview questions, not to answer them.
• Feed it real transcripts and ask it to find contradictions — that's genuinely useful and low-risk.
• Never carry an unsourced number from a model into a document a bank or investor will read.
• Ask it to argue against your hypothesis. It's better at destruction than confirmation.

The honest summary: AI compresses the tedious parts, but the load-bearing evidence has to come from people who exist.

How to write market research in a business plan

Investors and lenders skim. They're looking for three things: is the market real, is your evidence credible, and do you know what you don't know.

Structure the section in four short parts. First, market size with your reasoning — top-down and bottom-up, showing your workings. Second, your customer profile, drawn from interviews, with the number of conversations stated. Third, competitive overview including indirect alternatives (a spreadsheet is a competitor). Fourth, your evidence: what you tested, what the results were, and what you concluded. Including a failed test is not weakness — it signals that your numbers are real.

What to leave out: unsourced percentages, phrases like "the market is growing rapidly," and any figure you can't explain the origin of. If a lender asks where a number came from and you can't trace it, everything else in the document becomes suspect.

Should you include the research method and sample size?

Yes, briefly. One sentence like "based on 11 interviews with independent consultants sourced from two professional communities, plus a two-week pre-sale test" does more for your credibility than three paragraphs of adjectives.

Cheap tests that beat surveys

Behaviour is the only vote that counts. Here are four tests, roughly in order of how much they tell you per euro spent.

Pre-sale page and deposit test

Build a one-page site describing the product, with a real price and a real "Reserve your spot" button that takes a refundable deposit. Drive traffic with a small ad spend. The percentage who pay — not who click — is your strongest signal. A conversion rate under 1% of visitors is usually a sign your offer isn't compelling; over 5% and you have something worth building.

Interviews with the right people

Five to ten interviews with people who match your customer profile will surface the objections you can't anticipate. The value isn't in the sample size — it's in the specificity. You're not measuring; you're listening for the sentence that makes you say "I hadn't thought of that."

Ad and message tests

Run two different value propositions to the same audience with a small budget. Whichever gets the higher click-through tells you which language resonates, not which product is better. Cheap, fast, and it stops you building around a slogan nobody cares about.

The test that usually fails: the broad survey

A survey of strangers answering hypothetical questions produces noise. People overstate intent, understate price sensitivity, and answer in a hurry. I've seen founders spend two weeks distributing a survey, get 200 responses, and learn less than one afternoon of five phone calls. Use surveys to confirm a hypothesis you've already formed from behaviour — not to form it.

When to stop researching and start building

There's a point where more research stops reducing risk and starts being a way to avoid the discomfort of committing. You've reached it when three things are true: your hypothesis survived at least one behavioural test, you've spoken to enough people that you can predict their objections before they voice them, and you've written down your kill criteria and they weren't triggered.

At that point, the honest move is to build the smallest version and let the market answer the questions you couldn't. No amount of research will tell you whether people renew in month three. Only month three will.

The founders who get this right aren't the ones who researched longest. They're the ones who designed tests that could embarrass them, ran them anyway, and read the results without flinching. Which is a discipline, not a method — and it's the part no framework will hand you.