How to Do Market Research for a Startup, Step by Step

Market research at a big company means commissioning a sixty-page report. Market research at a startup means doing enough work to make better decisions about who to sell to, what to build, and what to charge — usually in two to three weeks, mostly by yourself. The two activities are not the same, and confusing them is how founders end up with binders full of data and no clearer view of their customer.

The goal is not to produce a deliverable. The goal is to change what you do. Every piece of research below should end with a sentence that starts 'because of this, we will…' If it does not, you collected information instead of insight.

Step one: write the question, not the topic

Most research goes nowhere because it begins with a topic ('the fitness app market') instead of a question ('do busy parents pay for one-to-one strength coaching online?'). A sharp question forces you to define the customer, the behavior, and the proof you would accept. Write three to five questions before you do anything else, and order them by how much each answer would change your plan.

Then decide what evidence would count. 'I would believe demand exists if I find ten parents already paying a personal trainer over a hundred dollars a month' is testable. 'I would believe demand exists if the market is large' is not.

Step two: secondary research to set the stage

Spend a day, not a week, on secondary research. The point is to learn the language of the industry, find rough size estimates, and identify the players — not to write a literature review. Use industry reports (Statista, IBISWorld, Gartner summaries), public filings of comparable companies, government statistics, and reputable news.

Cap your time. Diminishing returns hit fast, and secondary data ages quickly. Two reliable sources for any number is enough; three becomes vanity. Write down each source so you can defend the figure later.

• Industry size and growth rate, with the source named.
• Five to ten companies operating in the space, with rough revenue or funding if known.
• The most common pricing models and price points in the category.
• Major regulatory or technological shifts in the last three years.

Step three: customer interviews are the core

The single most useful research activity is talking to fifteen to twenty people who match your customer description. Recruit through LinkedIn, niche communities, customer referrals, or paid panels like Respondent if needed. Aim for thirty-minute conversations, not surveys.

Ask about the past, not the future. 'Tell me about the last time you ran into this problem. What did you do? What did you spend? What was annoying?' is far more useful than 'Would you use a product that…?' People are bad predictors of their own behavior and excellent reporters of what they have already done.

Step four: study competitors as a market signal

Competitors are the cheapest research subjects you have. Read their reviews on G2 or Trustpilot for unmet needs. Sign up for their products and time the onboarding. Subscribe to their emails to see how they sell. Pull their pricing pages into a single spreadsheet. Note where they cluster — and more importantly, where they do not.

Pay particular attention to one-star and three-star reviews. Five-star reviews flatter; three-star reviews reveal the workflows that broke, the missing integrations, and the segments that were promised but underserved. That is where positioning opportunities hide.

Step five: lightweight quantitative checks

Once interviews suggest a pattern, validate it with cheap quantitative tests. A fifty-dollar ad campaign to a tightly targeted audience will tell you whether the message resonates. A short survey to two hundred members of a relevant community will tell you whether the pain is broad or narrow. Google Trends, Keywords Everywhere, and Reddit search will reveal whether people search in your category at all.

Do not confuse a large survey sample with truth. Twenty people who match your exact customer beat two thousand random respondents every time. Match the sample to the question, not the spreadsheet to its size.

Step six: synthesize into decisions

Write a one-page synthesis. At the top, restate the original questions. Below each, write the answer and the evidence behind it. At the bottom, write three decisions you are now making because of what you learned: who you will target first, what you will build first, and what price you will test.

The synthesis is the only artifact that matters. Founders who skip it usually accumulate research without changing course; founders who write it discover their plan is now different from what they imagined three weeks ago. That difference is the entire return on the work.

When to stop researching and start selling

Research becomes procrastination once the marginal interview stops changing your mind. If interviews eleven through fifteen mostly confirm what you heard in interviews six through ten, you have enough to act. The next experiment should not be another interview — it should be an offer, a landing page, or a pilot.

Match the depth of research to the size of the bet. A weekend project needs almost none. A business that will absorb years of your life and significant capital deserves real evidence — but even then, that evidence is meant to make you start, not to delay starting forever.