# How to Do Useful Market Research: Listen Before You Measure

Learn how to do useful market research: explore needs with open questions, measure stable themes, and make evidence-backed decisions.

- Canonical URL: https://www.harmate.com/en/blog/how-to-do-useful-market-research-listen-before-you-measure
- Author: Harmate Team
- Published: 2026-08-23
- Updated: 2026-11-20T01:38:43.931890+00:00
- Language: en

## Content

# How to Do Useful Market Research: Listen Before You Measure

Useful market research does not begin with a list of boxes to tick. It begins with a decision to prepare, followed by enough open exploration to surface what the team did not anticipate. Measurement comes later, once needs, objections, behaviors, and customer language are stable enough to become categories.

This sequence prevents precise percentages from being built on answers that customers would never have volunteered themselves. It does not replace competitive research, industry data, or a suitable sampling plan. It connects them to a practical question: **what will we decide differently because of this research?**

This guide focuses on demand and offer-market fit: needs, current behavior, alternatives, objections, language, and differences between audiences. It complements rather than replaces the financial, legal, and regulatory analysis required in a full business plan. The linked guides cover question wording and interview technique in more depth; the subject here is the complete sequence from hypothesis to decision.

## What market research should help you decide

Market research brings together information about demand, supply, competitors, and the environment around a project. The [U.S. Small Business Administration](https://www.sba.gov/business-guide/plan-your-business/market-research-competitive-analysis) recommends combining market research and competitive analysis to understand customers and the competitive landscape. Gathering the information is only the first step. You still need to connect it to a decision.

Depending on your situation, that decision may be whether to:

- pursue or abandon a need hypothesis;
- prioritize one audience segment;
- rewrite a value proposition;
- change an offer or its pricing logic;
- choose a distribution channel;
- investigate one remaining uncertainty before committing resources.

Before collecting anything, complete this sentence: “At the end of this study, we need to decide whether…” Then state the hypothesis, the audiences involved, and the evidence that would lead you to confirm, qualify, abandon, or postpone the decision.

This framing reduces two risks. The first is encyclopedic research: rich in facts, poor in consequences. The second is validation research: a study designed mainly to reassure the team about a solution it has already chosen.

Our guide to [actionable questionnaires](https://www.harmate.com/en/blog/actionable-questionnaires-start-with-the-decision-not-the-questions) develops this decision-first principle.

## The expensive mistake: measuring categories too early

Imagine a team preparing a home meal service. It assumes that price, speed, and variety are the deciding factors. Its survey asks people to rate those three criteria. The answers will be easy to compare, but the survey cannot reveal what it excluded: the mental load of planning, trust in ingredients, the challenge of feeding several people, or the need to improvise.

A closed-ended question compresses reality into predefined options. That compression is useful when you already know what needs to be measured. It becomes risky when the team's assumptions quietly become the boundaries of the study.

The [American Association for Public Opinion Research](https://aapor.org/standards-and-ethics/best-practices/) describes the trade-off clearly: closed-ended questions place less burden on respondents and are easier to interpret, but the response options are more likely to influence the answer. Open-ended questions let people answer in their own words, while requiring coding before quantitative interpretation.

This is not a contest in which qualitative or quantitative research must win. For discovery, open exploration should come first. For measurement, comparison, and tracking, closed categories become useful after they have met the reality of the audience.

## A market research method that listens before it measures

### Frame the hypothesis and the decision

Start with four elements: the hypothesis, the audience, the object being studied, and the decision. A useful hypothesis must be capable of being contradicted. “Our solution is convenient” is not one. “Freelancers managing several clients lose time consolidating information and would consider switching tools if data migration were straightforward” gives you something to investigate.

Also state what the study will not establish. Exploratory conversations can reveal problems, language, and exceptions. They cannot, by themselves, estimate how much of an entire market thinks the same way.

### Use existing evidence before asking people

Desk research helps you understand the market before recruiting participants: official statistics, professional publications, regulation, trends, competitor offers, public pricing, reviews, and indirect alternatives.

Look at what competitors promise, what they charge, which audiences they appear to target, and the specific reasons for satisfaction or frustration in public reviews. These observations do not prove demand. They help you formulate sharper hypotheses to take into the field.

The SBA distinguishes between existing sources, which can answer broad and quantifiable questions, and direct consumer research, which is more specific but also more demanding. The distinction travels well beyond the United States: do not ask participants for information that a reliable public source can answer better.

Keep four columns in your research notes:

- what an identifiable source supports;
- what remains a hypothesis;
- what appears contradictory;
- what must be explored directly with the audience.

### Explore real experiences with open-ended questions

Exploration should focus on real situations rather than abstract opinions about an idea. Ask people to describe the last time the problem occurred, what they tried, what got in the way, the consequences, and the compromises they accepted.

Useful prompts include:

- “Tell me about the last time this happened.”
- “What did you do next?”
- “What required the most effort?”
- “What have you already tried, and why was it not enough?”
- “How would you describe this problem to a colleague?”
- “What would make you decide not to change anything?”

Avoid presenting your solution before you understand the situation. “Would you use a tool that…?” invites a judgment about an imaginary proposal. “How do you handle this today?” reveals behavior, alternatives, constraints, and the language people already use.

Open-ended does not automatically mean neutral. Wording, order, context, and follow-up prompts can frame the answer. [Krosnick and Presser's review of questionnaire design](https://web.stanford.edu/dept/communication/faculty/krosnick/docs/2010/2010%20Handbook%20of%20Survey%20Research.pdf) discusses these effects in detail. Prepare and test an interview guide, then use simple probes that ask for an example or clarification without suggesting a preferred answer.

For a detailed collection process, see our [semi-structured interview guide](https://www.harmate.com/en/blog/semi-structured-interview-guide-method-steps-and-15-examples).

### Structure responses without erasing disagreement

After collection, resist the urge to find “the” conclusion immediately. Connect every finding to source excerpts. Group similar formulations, but preserve the words people used, the conditions around them, exceptions, and disagreement.

A practical evidence table can contain:

- the observed theme;
- the situations in which it appears;
- source wording;
- the people or segments concerned;
- counterexamples;
- the level of uncertainty;
- the possible implication for the decision.

Frequency is not the only signal. A rare objection may matter if it blocks a purchase. A common preference may be secondary if it does not affect behavior. One unusually clear phrase may improve a message without representing the whole market.

Keep a path back from every theme to the [open responses that support it](https://www.harmate.com/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions). This makes it harder for a smooth summary to hide inconvenient evidence.

### Turn stable themes into measures

Measurement begins when multiple sources or conversations produce categories clear enough to test. You can then build closed response options, rating scales, or priority choices using language the audience actually used.

Suppose exploration repeatedly surfaces “avoiding duplicate data entry,” “finding information quickly,” and “knowing who changed what.” A quantitative questionnaire can measure the frequency, importance, or consequences of those situations. It can also include an appropriate “other” option when the categories may not be exhaustive and an open follow-up when the reason matters.

Before distributing the survey widely:

- make sure each option expresses one idea;
- make options mutually exclusive when the format requires it;
- include “do not know” or “not applicable” when those are legitimate states;
- test how people close to the target audience interpret the questions;
- place general questions before specific questions that could prime them.

The open phase is not decoration added to the final survey. It helps create better categories. The closed phase does not replace verbatim evidence. It shows how the categories are distributed within a defined scope.

### Compare audiences, then make a human decision

An overall average can hide useful differences. Compare segments that were part of the framing: new and long-term customers, users and buyers, small and large organizations, adopters and people who abandoned an alternative.

Do not manufacture segments after the fact only because they produce an interesting story. Document the definition of each group, its coverage, and the limitations of the comparison. A small base can generate clues or cases to explore; it should not be described as a representative result.

The verdict belongs to the team. Evidence can support four reasonable outcomes: confirm the hypothesis, qualify it, abandon it, or conclude that the study is still insufficient to decide. The final option is sometimes the most rigorous one.

## Example: when listening changes the hypothesis

Consider a company planning an expense-tracking service for freelancers. Its initial hypothesis is that price is the main weakness of existing tools. The team is ready to send a closed survey about acceptable prices.

In a set of open interviews, people tell a different story. They first mention scattered receipts, fear of making a mistake, duplicate entry, and uncertainty about what is missing before a deadline.

The team should not conclude that price does not matter. Instead, it revises the categories, preserves the source language, and separately measures the frequency of these situations, their perceived severity, current alternatives, and the role of price. It also compares freelancers working alone with those who share the process with an accountant.

The decision becomes more precise: do not sell only a cost saving; test a promise about visibility and fewer omissions. A strong verbatim can reveal an angle or vocabulary. It does not validate market size, willingness to pay, or price elasticity without dedicated measurement. This example illustrates the method; it is not evidence about all freelancers.

## How to choose whom to research

Research quality does not depend on response count alone. It depends on whether the people included match the decision.

For exploration, seek a useful range of situations: people who experience the problem intensely and occasionally, current and former users of alternatives, buyers and end users when those roles differ. The goal is to encounter contrasting experiences, not to imitate an entire population with a small qualitative sample.

For measurement, define the target population, recruitment method, nonresponse, and likely sources of bias. A survey distributed to your community primarily describes your community. A large number of responses does not automatically make it representative of the market.

Keep three ideas separate:

- **richness** describes the depth and context of responses;
- **comparability** requires sufficiently stable questions and conditions;
- **representativeness** depends on sampling and population coverage.

A study can be rich without being representative, or highly comparable while missing an unexpected cause.

## Investigate non-buyers and former customers separately

“Did not buy” and “stopped using” describe different histories. Before recruiting, define the event and time window you mean, then keep the groups separate in your notes and synthesis.

| Group | Include people who… | Question the account can inform | Do not conclude from it |
|---|---|---|---|
| Considered prospects who did not buy | Evaluated or discussed this offer, then chose another option, postponed, or made no decision | What alternatives, objections, constraints, or missing information they report at that decision point | Why every prospect declines, or what uncontacted visitors would have done |
| Former customers | Previously used or bought the offer, then cancelled or did not renew | What they remember about the decision to stop, what changed, and what they use now | That a stated reason alone caused the departure or predicts who will leave next |
| Current customers who considered leaving | Faced a concrete renewal or continuation decision and stayed | Which conditions they say mattered in that episode | That their reasons represent former customers or all customers |

Do not classify an abandoned enquiry as a lost sale unless the person actually considered the offer. Do not treat inactivity as a confirmed cancellation. Record the evidence source and stage, since a sales debrief, a customer’s later recollection, and an observed transaction are different materials. People may give several reasons, reinterpret a past choice, decline to respond, or be unreachable; keep those limits visible.

After a confirmed cancellation, the [churn survey](https://www.harmate.com/en/blog/churn-survey-questions-after-cancellation) offers optional feedback distinct from interviewing a prospect who never bought.

### A short interview guide for a real decision

Ask about a specific recent event before showing a concept or suggesting a reason. Adapt the prompts to the group, and leave room for an answer outside your list.

1. What was happening when you first looked for a solution?
2. What did you compare or try, and what happened next?
3. What mattered most when you decided to continue, choose something else, postpone, or stop?
4. Was there information, approval, timing, or a practical constraint that affected the decision?
5. What do you use or do now, if anything?
6. Is there a part of the story that does not fit the reasons we have discussed?

For a former customer, ask what changed over time and what they remember from the cancellation or renewal point. For a non-buyer, distinguish “not now,” “chose another option,” “no decision,” and “the offer was not relevant”; do not force a single category when the person describes more than one. If the interview concerns an offer still in development, use the [concept-testing protocol](/en/blog/concept-testing-an-exploratory-protocol-for-an-offer) to explore how people understand and react to the proposed offer. A concept reaction is not evidence of a past purchase decision.

For a B2B purchase decision, the [win-loss interview guide](https://www.harmate.com/en/blog/win-loss-interview-questions-b2b-guide-example) compares the buyer’s account with the CRM reason, including cases where no purchase occurred.

### Keep a source-linked reason log

| Case and stage | Person’s words or excerpt | Situation and alternative | Reason as stated | Other or competing explanation | Follow-up / limit |
|---|---|---|---|---|---|
| Fictional former customer, renewal | “We stopped using it after the team changed.” | Handover; a spreadsheet was used for a while | Team change, in the person’s words | The replacement process may also have mattered; not established | Ask what happened during handover; one account, not a general pattern |
| Fictional considered prospect, no decision | “We could not decide who would maintain it.” | Evaluated the offer; kept the existing process | Ownership was unclear | Budget or timing may also have mattered but was not mentioned | Ask who needed to agree; absence from this interview is not proof it was irrelevant |

These examples are fictional and demonstrate how to preserve context, uncertainty, and competing explanations. During analysis, group similar stated reasons only after reviewing their source excerpts. Keep “not reported,” “not asked,” “no response,” and “unclear” distinct from a confirmed reason. Count or compare categories only when the recruitment, question, and denominator support that calculation; exploratory interviews alone do not establish prevalence, causation, or future churn risk.

A published survey study of service switching found that the mix of stated reasons varied by service and by the kind of switching considered. That is a reason to preserve context when comparing your own accounts, not a template of reasons to impose on respondents ([Keaveney and Parthasarathy, “Reasons for switching service providers”](https://doi.org/10.1016/j.ausmj.2011.12.001)). The international [ICC/ESOMAR Code](https://esomar.org/representation/icc-esomar-code-of-conduct) provides a professional framework for responsible research and respondent treatment.

## The right deliverable is a decision brief

A useful deliverable makes the reasoning reviewable. It can be short if it preserves:

- the hypothesis and the decision being prepared;
- the desk research sources;
- the audiences, methods, and limitations;
- themes, language, objections, and counterexamples;
- relevant measures and comparisons;
- verbatim evidence supporting the findings;
- evidence that contradicts the hypothesis;
- the chosen verdict and next actions.

This structure prevents conclusions from floating free of evidence. It also allows the reasoning to be revisited when the market, offer, or underlying data changes.

If the uncertainty concerns an offer still being developed, the [concept testing protocol](/en/blog/concept-testing-an-exploratory-protocol-for-an-offer) provides a stimulus, questions and a review framework for exploring comprehension, perceived usefulness and objections. This exploration does not predict sales.

## Where Harmate can help without deciding for you

Harmate helps marketing teams [collect questionnaire responses in Real Mode](/en/product/real-mode) or [import a structured file](/en/product/response-import). Imports support CSV, JSON, or XLSX after column review. The team can then analyze responses while keeping useful verbatims visible. It does not replace market analysis, participant recruitment, or human judgment; the team remains responsible for interpreting what the material supports.

View the public [Market Research – needs and fit template](/en/product/template-library/3e33cb96-9ad7-40b8-82c1-48ab1999eda8) to explore an example for this use case.

You can explore that workflow on the [Harmate for marketing teams](https://www.harmate.com/en/profiles/marketing-teams) page.

## Market research FAQ

### Should market research begin with a survey?

Not necessarily. Start with evidence that already exists and direct exploration when your categories are still hypothetical. A closed survey becomes more useful once you know what you need to measure and which population the result should describe.

### How many people should I interview?

There is no universal number. In exploration, the range of relevant situations and the arrival of genuinely new themes guide collection. For quantitative estimation, the required sample depends on the population, sampling method, desired precision, and comparisons planned.

### Are open-ended questions always better?

They are richer for discovering causes, language, and exceptions that predefined categories cannot contain. They also require more effort from respondents and analysts, are less directly comparable, and guarantee neither neutrality nor representativeness. Their value depends on the research objective.

### Can market research be done entirely online?

Yes for some objectives, provided that online sources and reachable participants correspond to the population being studied. If a channel excludes an important group or hides the context of use, combine it with other methods and make the limitation explicit.

### How do you know when the study is finished?

The study is sufficiently developed when the team can connect a verdict to evidence, explain the limits, and name the next action. If new responses no longer change the decision and only add variants of themes already understood, more collection may add little value. If a decisive disagreement remains unresolved, address it directly rather than hiding it in an average.

## The takeaway

Good market research does not begin by producing numbers. It begins by reducing an uncertainty that blocks a decision. Desk research establishes the context, open questions surface causes and audience language, analysis preserves disagreement, and closed questions measure categories once they are solid enough.

Listening before measuring does not mean giving up quantitative rigor. It gives the numbers better things to measure.