Market Research for Starting a Business: A 7-Step Guide

Market research for starting a business means finding out, before you build, who has the problem, how much it hurts, what they use today and whether they would pay for something better. You do not need a big budget or a research firm. You need a week, a clear idea and a habit of writing down what you learn.

The part most guides skip is what to do afterward. A pile of notes is not a result. This guide walks through seven steps, and each one ends in something you can use in your plan: a customer segment, a gap table, a positioning statement and a first-version MVP scope.

If you do the steps in order, you will have a decision at the end of the week: go, change direction, or stop. All three are good outcomes. The bad outcome is spending six months building the wrong thing.

What market research for a new business should actually answer

Market research reduces risk. It helps you find real demand and avoid building something nobody asked for. It does that by answering four questions:

• Who has the problem? A specific group, not “everyone.”
• How painful is it? Painful enough that they already spend time or money on it.
• What do they use today? Rival products, spreadsheets, an assistant, or nothing at all.
• Will they pay? Evidence from behavior, not compliments.

Research exists to make decisions, not to produce a report. If a finding does not change who you target, what you charge, what you build or whether you continue, it was interesting but not useful.

Primary vs secondary, qualitative vs quantitative

A good first pass uses all four: desk research (secondary), community reading and interviews (qualitative primary), then a quick numbers check (quantitative primary).

Step 1: Write down your idea and your riskiest assumptions

Start with one sentence: “[Product] helps [specific customer] do [job] instead of [current alternative].” For example: “A booking app helps small physiotherapy clinics fill cancelled slots instead of calling a waitlist by hand.”

Then list everything that must be true for the idea to work. Rank each assumption by two things: how much it would hurt if wrong, and how little evidence you have. The top three become your research questions.

• Clinic owners lose real money on cancellations (pain).
• They would try a new tool instead of their current calendar (switching).
• Patients will accept a text offering a freed-up slot (behavior).

Artifact: a one-page sheet with the idea sentence, the target customer, the problem, the assumed alternative and three ranked assumptions. This is also the first stage of how you validate your startup idea before you build anything.

Step 2: Do desk research first (secondary research)

Desk research is cheap and fast, so do it before you ask anyone for their time. Look at:

• Government and public data, such as U.S. Census Bureau data, for business counts and demographics.
• Google Trends to see whether interest in a topic is rising, flat or seasonal.
• App store listings and product reviews for existing tools.
• Industry association pages and public reports.
• The U.S. Small Business Administration’s market research and competitive analysis guide for a basic framework.

Rough market sizing, bottom-up

You will see TAM, SAM and SOM in many guides. In plain terms: TAM is everyone who could theoretically buy, SAM is the part you could reach with your product and model, and SOM is the part you can realistically win early.

For a first pass, skip top-down claims like “the market is worth billions” and build from the bottom: number of target businesses you can reach × realistic share × price per year. For the clinic example: clinics in your first region × the share that fit your profile × a monthly price × 12. The number will be rough. What matters is whether it is clearly too small to support the business you want.

Using AI for desk research

AI is useful for summarizing long documents, listing questions to ask and drafting a first outline of competitors. It is not a source. It can produce numbers and citations that do not exist.

• Ask it to summarize material you paste in, rather than to recall facts.
• Open every source it names and confirm the number is really there.
• If you cannot find the original, do not use the figure.
• Treat its output as a starting list to check, not a finding.

For tools worth trying, see AI tools for startup founders.

Artifact: a rough market-size estimate with its inputs listed, and a short list of trend notes with sources.

Step 3: Mine free online communities for real language

People complain in public all the time. Reddit, Quora, Facebook groups, niche forums and product reviews show the exact words customers use, the workarounds they invented and what annoys them about current tools.

Search for phrases like “how do you handle,” “anyone else struggling with,” “alternative to” and “is there a tool for.” Read threads, and also read the one-star and three-star reviews of competing products.

Log what you find in a simple table so patterns appear:

Artifact: a log of 20 to 30 entries. Use the exact phrases later in your positioning and landing page copy. Remember that online voices skew vocal, so treat this as a source of hypotheses, not proof.

Step 4: Talk to 8 to 15 potential customers

Interviews are the most valuable step. Aim for 8 to 15 people in your target segment. Fewer than 8 and one loud opinion can mislead you. Past about 15, you usually hear the same things again, which is a sign you can stop.

Where to find people without a network

• Message people who posted in the communities from Step 3.
• Reach out to owners of small local businesses in your target group.
• Ask for introductions in professional groups, and offer 20 minutes, not an hour.
• Ask each interviewee who else you should speak to.

Avoid friends and family. They are kind, and kindness ruins data.

A question script about past behavior

Ask about what people did, not what they might do. The approach comes from Rob Fitzpatrick’s book The Mom Test, which is worth reading before your first call.

• Tell me about the last time [the problem] happened. What happened?
• What did you do about it?
• How long did that take, and what did it cost you?
• What tools or workarounds do you use now?
• What have you tried that did not work?
• Why is this still a problem?
• Who else is involved when it happens?
• If you could not solve it, what would that cost you?

Questions to avoid

• “Would you use this?” People say yes to be polite.
• “Would you pay for it?” Opinions about future spending are unreliable.
• “Do you think this is a good idea?” This invites compliments.
• Anything that describes your solution before they describe their problem.

Signals of real demand vs polite interest

Decision rule: if fewer than about a third of your interviewees describe a recent, costly instance of the problem, the segment or the problem is probably wrong. Change one of them before continuing. That is a rule of thumb, not a law, so adjust it, but decide the threshold before you start so you cannot argue yourself out of it later.

Artifact: a customer segment definition based on who actually had the problem, plus a ranked list of the top problems.

Step 5: Map competitors and alternatives into a gap table

Three kinds of competition matter:

• Direct: products that solve the same problem for the same customer.
• Indirect: products that solve it a different way.
• Do nothing / manual: spreadsheets, paper, email, an assistant, or accepting the loss.

The third kind is often the real competitor. In the clinic example, the biggest rival is not another app, it is the receptionist with a phone.

Build a table, not a list:

Read the gaps across the table. Where is every option weak? Which weakness can you credibly own given your size and skills? That gap, matched to the segment from Step 4, is your differentiation.

Artifact: the gap table and one sentence naming the gap you will own.

Step 6: Run a quick quantitative check

Interviews tell you why. A small numbers test tells you whether the pattern holds beyond the people you spoke to. Pick one:

• Short survey: 5 to 8 questions sent to your target group, mostly about past behavior and current tools.
• Landing page test: a page describing the problem and promised outcome, with a sign-up or waitlist button, sent to your segment through a community, outreach or a small ad budget.

Be realistic about sample size. A solo founder will get dozens of responses, not thousands, so the results are directional. Look for strong signals, not small differences.

Set thresholds before you run it, so the result decides for you:

Pick the target numbers yourself, based on how many people you reached. The point is that you write them down first.

Artifact: a go, change or stop call, with the numbers that support it.

Step 7: Turn findings into a plan, positioning and MVP scope

This is the step most guides leave out. Take each finding and attach it to a decision:

A simple positioning statement: “For [segment] who [problem], [product] is a [category] that [key benefit]. Unlike [alternative], it [gap you own].”

Then cut. If a feature has no research evidence behind it, it does not go in the first version. This is how you decide what goes into version one without arguing from taste, and how you turn a raw idea into a product plan that is based on evidence. Scope that grows without evidence is how projects drift; see scope creep for how to spot it.

Artifact: a segment, a positioning statement, a pricing hypothesis and a must-have feature list.

Worked example: from research notes to a scoped MVP

Idea: a pet-walking marketplace for busy professionals in one city.

Rank the problems: trust, reliability, same-day cover. Then map user journeys before you write code for the two key people, the owner booking a walk and the walker accepting one. The journeys show the screens and steps version one needs: walker profiles with verification, booking, a replacement flow, and payment. Reviews, chat and tracking wait.

To turn notes like these into a structured plan, you can paste them into an AI business plan generator and then edit the result against your evidence. Inside TwoSprint the flow runs in the same order: research notes, then features, then journeys, then the build plan.

Common mistakes and how to know you have done enough

• Asking friends and family. They protect your feelings, not your budget.
• Leading questions. “Wouldn’t it be great if…” gets you agreement, not facts.
• Confirmation bias. Write down what would prove you wrong before you start, and look for it.
• Trusting unverified numbers. Check AI-produced figures against the original source.
• Researching forever. More interviews rarely change the answer after the patterns repeat.

You have done enough when you can say yes to all of these:

• Patterns repeat across interviews and communities.
• You can name one clear segment.
• You can point to a gap in the competitor table you can own.
• You have a go, change or stop call from a numbers check.
• You have a first-version scope tied to evidence.

Printable checklist

• [ ] Idea sentence and top three assumptions written
• [ ] Rough bottom-up market size with sources checked
• [ ] 20 to 30 community findings logged
• [ ] 8 to 15 interviews done using a past-behavior script
• [ ] Competitor and alternatives gap table complete
• [ ] Survey or landing page result against pre-set thresholds
• [ ] Segment, positioning, pricing hypothesis and must-haves written
• [ ] Features with no evidence moved to the later list

Doing this with TwoSprint

Once your research is done, the next job is keeping it connected to the plan. TwoSprint lets you bring your notes into one workspace and move from the problem to features, journeys and a scoped first version, so each feature points back to something you learned.

If you still feel unsure about demand, the natural next step is to validate your startup idea before you build anything with a small, real test.