Back
Market Research Methods: A Practical Guide
SmartInterview Team
The Short Answer
Market research methods divide along two axes. Primary versus secondary: you collect it, or somebody already did. Qualitative versus quantitative: you are understanding, or you are counting. The method you need follows from the question you are asking, and the most common failure in commercial research is running the method you are comfortable with instead of the one that fits.
Secondary first. Desk research is the cheapest hour you will spend, and it usually shortens the primary study that follows.
Qualitative to find the dimensions: depth interviews, focus groups, ethnography, open-ended survey questions.
Quantitative to size them: surveys, conjoint and trade-off exercises, brand tracking, panel research.
Sequence beats scale. A small qual phase before a survey saves more money than a bigger survey ever earns you.
Judge the output by the method, not the deck. Ask who was sampled, how they were recruited, and what the question actually said.
For the underlying distinction, see qualitative vs quantitative research. For the software layer, see market research tools.
The Methods at a Glance
Eight methods cover the large majority of commercial market research. Everything else is a variant of one of them.
Method | Type | Answers | Typical timeline | Where it breaks |
|---|---|---|---|---|
Desk research | Secondary | Market size, structure, competitor and regulatory context | Hours to days | Definitions and dates rarely match your question |
Surveys | Primary, quant | Incidence, preference, segment size, tracked attitudes | 1-4 weeks | Only reports what you thought to ask |
Depth interviews | Primary, qual | Reasoning, buying process, customer vocabulary | 2-4 weeks | Small n, not projectable, expensive to scale |
Focus groups | Primary, qual | Reaction to stimulus, range of opinion, shared language | 2-4 weeks | Group dynamics distort; poor for sensitive topics |
Ethnography and observation | Primary, qual | What people actually do in context | 3-8 weeks | Time-hungry; observation changes behavior |
Conjoint and trade-off | Primary, quant | What buyers will give up for what, and price sensitivity | 4-8 weeks | Design-heavy; garbage attribute list gives garbage utilities |
Brand tracking | Primary, quant | Awareness, consideration and perception over time | Continuous or per wave | Worthless unless the instrument stays identical |
Panel research | Access model | Reach to audiences you do not own | Days to weeks | Sample quality varies enormously by provider |
Panel research sits slightly apart: it is not a way of asking questions, it is a way of finding people to ask. It is listed here because in practice it determines the credibility of any quantitative study you cannot run on your own customer list. We unpack how that market works in what is panel research.
Primary vs Secondary Research
Secondary research: cheap, fast, and usually skipped too quickly
Secondary research means analyzing data that already exists: national statistics, industry and trade association reports, regulatory filings, published academic work, competitor disclosures, and your own internal records. That last category is the one companies forget. Support tickets, sales call notes, churn reasons and product analytics are all market research data you have already paid for.
Do this phase before commissioning anything, and end it with an explicit list of what you still do not know. That list is the brief for the primary work, and it is usually much shorter than the survey somebody had already started drafting.
The limitation is fit. Somebody else defined the terms, drew the sample and picked the timing. Before you rely on a secondary figure, check who funded it, when the fieldwork happened rather than when it was published, how the key terms are defined, and whether any method is disclosed at all. A number with no stated sample or method is marketing, not evidence.
Primary research: for the gaps
Primary research is data you collect for your question. You set the sample, the wording and the timing, and you can probe when an answer is interesting. That control is what you are paying for in cost and elapsed time. The general treatment of the collection side is in data collection methods.
Qualitative and Quantitative: Which Question Does Each Answer?
The choice is not about rigor. Both are rigorous when run properly. It is about the shape of what you do not know.
Use qualitative when you do not know the options yet
If you cannot write a credible list of answer choices, you are not ready to run a survey. Qualitative methods find the dimensions: what people are actually comparing, what words they use, what the decision process looks like, what alternatives you have never heard of. Use them for entering a new category, diagnosing an unexplained result, exploring a sensitive topic, or understanding a purchase with many steps and multiple people involved.
Qualitative findings are not projectable. Fifteen interviews cannot tell you that 40% of a market thinks something, and any deck that implies it should be sent back. What they can tell you is that a reason exists, that it is coherent, and how customers describe it, which is often the more valuable finding.
Use quantitative when you need a number to allocate against
How many, how often, how much, which segment, is it moving. Quantitative methods are how you size an opportunity, compare markets, prioritize a roadmap or defend a budget. The constraint is that they only measure what you thought to ask, which is precisely why the qualitative phase belongs first.
The cost boundary has moved
The historical reason projects skimp on qualitative work is cost per participant: recruiting, scheduling, moderating and analyzing an interview is expensive, so studies carry twelve of them and the survey does the rest. AI-moderated interviews and automated coding of open responses change that arithmetic, making it feasible to run open, probed conversations at sample sizes that used to be closed-question territory. We go through what that changes, and what it does not, in qualitative research with AI.
Sequencing Methods in a Real Project
Methods work in order. Running them in the wrong order is the most expensive avoidable mistake in a research plan, because a survey built without a qualitative phase measures the wrong list of things at full cost.
The default sequence
Write the decision down. One sentence, with the person who will make it named, and the date they need it. If no plausible finding changes that decision, stop here and save the budget.
Desk research. Internal records first, then external sources. Output: what is already known, and what is genuinely open.
Qualitative exploration. Interviews or open probed surveys against the open questions. Output: the dimensions, the vocabulary, the candidate answer options.
Quantitative measurement. A survey built from what the qualitative phase found, sized against the smallest subgroup you intend to report on.
Qualitative follow-up. The step almost everyone skips. Go back to people to explain whatever came back surprising in the numbers.
Tracking, if the question repeats. Freeze the instrument and re-run it on a schedule.
When to break the sequence
If the decision is narrow and the options are already well defined, go straight to quant. Choosing between three finished concepts does not need a discovery phase. If the decision is broad and nobody can articulate what customers even want, do not run a survey at all yet: a survey written from internal guesses returns internally-shaped answers.
Budget-Tiered Approaches
The same question can be answered at very different price points. What changes is the confidence you can claim, and being honest about that is the professional part of the job.
Minimal budget
Desk research plus your own records, then a survey to a list you already own, plus a handful of customer conversations. Sample access is usually the largest cost line in research, and using your own list removes it entirely. The limitation is real and must be stated: you are learning about existing customers, which tells you little about the people who never bought, and those are often the ones you actually need.
Mid budget
Add purchased sample so you can reach non-customers and competitor customers, and add a properly recruited qualitative phase. This is the tier at which most genuine market questions become answerable, because you can finally compare people who chose you with people who did not.
Full budget
Add specialist methods where they pay for themselves: conjoint for pricing and feature trade-offs, a continuous brand tracker, ethnography for categories where stated behavior and real behavior diverge. Add an agency where you need fieldwork you cannot run yourself or independence you cannot claim internally.
Two spending rules hold at every tier. Spend on sample before you spend on sample size, because reaching the right people beats reaching more of the wrong ones. And budget analysis time at design stage: teams routinely book a week of fieldwork and an afternoon of analysis, then leave recordings untranscribed and open responses uncoded.
How to Judge Whether the Output Is Trustworthy
Whether the work came from an agency, a platform or your own team, these are the questions that separate a finding from a slide.
Who exactly was in the sample?
Not "n=500 consumers". Which country, which age range, which screening criteria, and how many were excluded to get to 500. A high screen-out rate is not a problem in itself, but combined with an incentive it creates pressure for respondents to claim eligibility they do not have.
How were they recruited?
Owned list, panel provider, river sampling from ads, or social. Each carries a different bias. Panel and river sources are non-probability samples, which means the margin of error people quote from them rests on assumptions rather than sampling theory. The honest phrasing is "among respondents", not "of the population".
Can you see the questionnaire?
Ask for the actual wording, in order. Leading stems, unbalanced scales and missing "none of these" options are invisible in a topline and decisive in the result. If nobody will show you the instrument, treat the numbers as unverified.
Are the bases reported per chart?
Subgroup findings are where credibility goes to die. A chart comparing four markets on a 500-person study is reporting on bases around 125, and a chart comparing four markets by three age bands is reporting on bases too small to mean anything. Every chart should carry its own n.
Does the qualitative work stay qualitative?
Percentages from twelve interviews are a category error. Well-run qualitative output reads as "three of the twelve described this pattern", and it earns its place through explanation and quotation, not arithmetic.
Can you get back to the raw material?
You should be able to click a theme and read the verbatims behind it, and open the response-level data behind a crosstab. Coding you cannot audit is a black box, whether a person or a model produced it. This applies to automated coding specifically: the requirement is not that the model is perfect, it is that you can check it against the original words.
Did the instrument change between waves?
For any tracker, ask what changed. A reworded question, a moved question, or a switch from phone to online can move a number with no underlying change in the market. If a metric jumps and the method also changed, you have learned nothing about the market.
Common Mistakes
Researching a decision that has already been made. If the answer is fixed, the study is theater and everyone involved knows it.
Asking about future intent. "Would you buy this at 40?" is weakly related to buying it at 40. Anchor to what people actually did, or use trade-off methods that force a choice.
Confusing sample size with sample quality. Five thousand responses from a skewed source just make a biased estimate look precise.
Only researching your own customers. The most important people in most market questions are the ones who did not choose you.
Leaving the open responses uncoded. The richest data in the study, routinely dumped in an appendix nobody reads.
Where to Go Next
Pick the toolchain: market research tools
Understand sample sourcing: what is panel research
Go deeper on the qual side: qualitative research with AI
Choose the collection technique: data collection methods
Run the qualitative phase you keep cutting
In most research plans, the qualitative phase is the one that gets dropped when the timeline tightens. The survey survives because it is cheap to field, and it goes out built on internal assumptions about what customers care about.
SmartInterview makes that phase affordable enough to keep. It runs AI-moderated interviews and surveys by voice or text, probes each answer the way a moderator would, works across languages, and codes the open responses into themes you can count and trace back to the exact words a respondent used.
Put it in front of your next quantitative wave and see how much the questionnaire changes. Start free or book a demo.
Frequently Asked Questions
What are the main market research methods?
Desk research, surveys, depth interviews, focus groups, ethnography and observation, conjoint and trade-off exercises, and brand tracking. Panel research sits alongside these as the access model that lets you reach people outside your own customer list. Most projects combine two or three of them.
What is the difference between primary and secondary market research?
Primary research is data you collect yourself for your specific question, so you control the sample, wording and timing. Secondary research reuses data collected by others, including published statistics, industry reports and your own internal records. Secondary is faster and cheaper, so it belongs first, with primary work reserved for what it leaves unanswered.
Should I use qualitative or quantitative methods?
Use qualitative methods when you cannot yet write a credible list of answer options, when you need customers' own language, or when a number has come back inexplicable. Use quantitative methods when you need to size, rank, compare or track. Most substantial projects need both, with qualitative work first to define what the survey should measure.
How much does market research cost?
It varies enormously with how hard your audience is to reach, and sample access is usually the largest line. Surveying a customer list you already own removes that cost almost entirely; reaching low-incidence B2B audiences through a panel can cost an order of magnitude more per participant than general consumer sample. Budget analysis time separately, since it is the line most often forgotten.
In what order should I run market research methods?
Write down the decision, do desk research including your own internal data, run a qualitative phase to find the dimensions, then a quantitative phase to size them, then go back qualitatively to explain anything surprising. Skip the discovery phase only when the options are already well defined, such as choosing between three finished concepts.
How do I know if a research finding is trustworthy?
Ask who was sampled and how they were recruited, read the actual questionnaire wording, check that each chart reports its own base size, and confirm you can get from a summary back to the raw responses. For trackers, ask whether anything about the instrument or the mode changed between waves.
Can I do market research without a big budget?
Yes. Start with your own records and free published sources, survey a list you already own, and hold a small number of real customer conversations. The honest limitation is that you will be learning almost entirely about existing customers, which says little about the people who considered you and chose something else.
