Last updated: 22 September 2026
Quick Answer: Market research splits into four types by business objective: brand, customer, competitor and product. Each answers a different question, so the type you pick sets the method, the sample and the turnaround. Primary research buys new data you own; secondary research reuses published data, in an industry ESOMAR sized at $153 billion in 2024.
Which type you commission decides almost everything downstream. The method, the number of people you need to talk to and the weeks before anyone can act are fixed the moment you name the type, not when fieldwork starts.
That industry is large and still growing. ESOMAR put the insights sector at $153 billion in 2024, an 8% rise on the year before, split across market research, research software and business services (Global Market Research 2025). Most of that spend answers one of the four questions below.
The expensive mistake is rarely the vendor. It is commissioning a brand tracker when the live question was a pricing question, then finding out six weeks later that the data cannot answer it and the launch date has not moved.
What Counts as Market Research
Market research is the systematic collection and interpretation of information about people and organizations, done to give decision makers objective, fact-based evidence instead of opinion. The ICC/ESOMAR International Code, revised in 2025, applies that definition to qualitative and quantitative work alike, including studies run on self-service platforms.
The dividing line between research and data collection is disclosure. The American Association for Public Opinion Research sets out what has to be published alongside any finding: the data collection strategy, who sponsored and who conducted the work, the exact instrument wording, the population under study, and whether the sample came from a probability frame or an opt-in source (AAPOR disclosure standards). A number with none of that attached is a claim, not evidence.
Value comes from the decision the evidence changes, not from the volume of data collected. The case for treating it as a standing cost rather than a crisis purchase is made in why market research matters for sustainable growth, and delivery sits on the full-service market research page.
Primary and Secondary Research, and When Each Is Enough
Primary research is original data you collect or commission, so you own it exclusively and it is designed around your question. Secondary research uses data someone else has already collected, analyzed and published. The trade is specificity against cost, and the mistake is running the first when the second would have answered it.
Secondary comes first for a reason. The Small Business Administration points buyers at the Census Bureau and NAICS for industry structure, the Bureau of Labor Statistics for employment and income, the Federal Reserve for credit conditions, the Consumer Product Safety Commission for product incidents, and the Bureau of Economic Analysis for GDP and consumer spending (SBA market research guide). None of it costs anything.
It is also more current than most people assume. The American Community Survey randomly selects about 3.5 million addresses every year, which is a larger annual sample than any commercial panel will field for you. For market sizing and category baselines, published data is not the cheap option, it is the better one.
Primary research earns its cost when the question is specific to you: why your checkout abandons, what your buyers think your brand stands for, whether your concept beats your own current product. The full split, including where each one breaks down, is covered in primary versus secondary market research.
Qualitative and Quantitative Data Are No Longer Separate Worlds
The old wall between qualitative and quantitative research was a measurement problem, and measurement has largely caught up. Social comments that were once anecdotal are now tracked at volume and counted, and open-ended responses are coded automatically while fieldwork runs.
What has not changed is inference. Counting something precisely does not make the people you counted representative of anyone else. Pew Research Center benchmarked six online samples against government sources and found online opt-in samples averaged 11.2 points of error for 18 to 29 year olds and 10.8 points for Hispanic adults, against 3.6 points for both groups on probability-based panels (Comparing accuracy of 2 types of online survey samples). The arithmetic was fine in every case. The sample was not.
Underneath accuracy sits a cruder problem. In a separate Pew study of more than 60,000 interviews across six online sources, bogus respondents ran 4% to 7% of widely used opt-in sources against 1% in panels recruited by address, and they were not answering at random: they tended to approve of whatever was put in front of them, which biases an estimate rather than just adding noise.
Qualitative work still explains why: what frustrates you about the current solution, what made you pick this over the alternative. Quantitative work sizes and ranks: what you would pay, how often you use it.
The practical answer is to run both and to be honest about which one carries which claim. Numbers find the pattern, verbatims explain it, and neither survives a sample drawn from the wrong people, which is why where good survey respondents come from is worth settling before the discussion guide.
Four Types of Market Research by Business Objective
Four types cover almost every commissioned study: brand research, customer research, competitor research and product research. They are named after the decision they serve, not the technique they use, which is why two of them can share a method and still produce very different reports. The table maps each one to its objective, method, sample and turnaround.
Rows run in the order a growth question usually arrives, from who knows you to what you sell, so the order carries no ranking. Sample and turnaround are typical commissioned ranges for one wave, not guarantees, and recruitment rather than analysis is usually what sets the clock.
| Type | Business objective it settles | Typical method | Typical sample per wave | Typical turnaround |
|---|---|---|---|---|
| Brand research | Whether the market knows you, and what it thinks it knows | Awareness and perception surveys, positioning interviews | 200 to 400 for the survey, 40 to 80 for the qualitative layer | 2 to 4 weeks |
| Customer research | Why your existing buyers behave the way they do | In-depth interviews, segmentation surveys, journey mapping | 10 to 20 interviews per segment, 400 or more for projectable scores | 1 to 3 weeks |
| Competitor research | Where you can differentiate and where you cannot | Desk analysis, switcher interviews, message and feature benchmarking | 15 to 30 switcher interviews alongside desk work | 1 to 2 weeks |
| Product research | Whether the thing you are building is worth building | Concept tests, usability sessions, pricing research, in-market tests | 8 to 15 users per persona for usability, 200 to 400 per cell for concept tests | 1 to 3 weeks |
The qualitative column starts from published saturation work. Across three coded interview datasets, the median number of interviews before new information fell to 5% or less of what the first interviews produced was six in homogeneous samples and eight to nine in a more varied one, at a run length of two; waiting for no new themes at all pushed those medians to 11 and 12 (Guest, Namey and Chen, PLOS ONE). The quantitative column comes from what the claim needs: a cell of 400 carries a margin of sampling error of roughly 5 points at 95% confidence for an estimate near 50%.
Brand Research
Brand research tells you how the market sees you and whether that matches what you intended. It covers awareness, who knows you and how they found you; perception, what they think and feel; positioning, where you sit against the alternatives; loyalty, whether they come back; and value, what premium the name can carry.
Methods are unglamorous and well established: recall and recognition surveys, interviews that surface brand associations, and concept sessions that test positioning language before a campaign. Which measures actually move, and how often to take them, is set out in brand perception metrics and how often to measure them.
The output is a gap list. Where perception and intent diverge, brand research tells you which of the two is wrong. A recurring qualitative layer beside an existing tracker, as in deeper brand tracks, exists because a score that moved without an explanation is not actionable.
Customer Research
Customer research puts your existing buyers under the microscope: who they are, what drives their decisions and where you are failing them. The recurring themes are satisfaction, loyalty, segmentation, purchase triggers and barriers, and the journey from first awareness through to repeat purchase.
Techniques span net promoter score surveys for loyalty, satisfaction interviews for depth, journey mapping for touchpoints, and purchase-record analysis for the behavior people do not report accurately. Segments only earn their cost when teams can act on them, which is the distinction drawn in segmentation research that produces usable segments.
Start the work with a spending anchor rather than a persona. The federal Consumer Expenditure Surveys put average total annual US household spending at $78,535 in 2024, with housing at $26,266 and transportation at $13,318. That is the budget your price competes inside, and it is published before you commission anything.
The payoff is that you stop guessing what customers want and start knowing. Personas are only as good as their inputs, which is why personas built from real interviews behave differently in a roadmap meeting than personas assembled from analytics.
Competitor Research
Competitor research maps where you stand and where the open ground is. It goes past a list of names into strategy, strengths, weaknesses and how customers perceive those competitors when you are not in the room.
The activities are a SWOT for you and each serious rival, a positioning map, an honest read of what they do well, and share-of-market movement. The interviews that matter most are with switchers, people who chose a competitor over you, because they can name the reason in a way your own customers cannot. Category movement worth tracking alongside it sits in market research trends.
Public reviews are a legitimate input here and a cleaner one than they used to be. The Federal Trade Commission announced its final Consumer Reviews and Testimonials Rule on 14 August 2024, banning the buying, selling and suppressing of reviews with civil penalties for knowing violations, effective 60 days after publication in the Federal Register (FTC final rule). Cleaner is not the same as representative: a review set is still whoever chose to write.
Product Research
Product research runs from concept validation through post-launch optimization, and its job is to remove risk at each stage. The questions are product-market fit, feature priority, design comprehension, message resonance and performance of what is already shipped.
Early on, concept tests and exploratory interviews establish whether the problem is real and whether your solution reads as a solution. The mechanics of running one properly, including comprehension checks before a rating, sit on the concept testing page. Between versions, usability sessions and trade-off exercises settle which features earn their build cost.
Pricing is the question most teams under-research and the one where guessing costs the most, since a price set too low is margin gone for the life of the product. Methods that produce a defensible number, rather than a stated willingness to pay, are compared in pricing research and what customers will actually pay.
How to Choose the Right Market Research Approach
Start from the decision you have to make, then work backward to the method. A question about whether to launch needs a different study from a question about why checkout abandons, and the wrong pairing wastes the budget rather than the schedule.
Three constraints then narrow the field: the claim the sample has to carry, whether you can reach those people at all, and what budget, timeline and in-house skill allow.
Match the Sample to the Claim You Want to Make
The claim sets the sample, not the other way round. If the finding will be quoted as a national number, it needs a probability frame and a sample large enough to carry it, and a specialist survey house with an address-recruited panel is the better buy than any interview platform, however good the conversation.
If the finding will be quoted as a reason, twenty well-run interviews per segment will beat two thousand survey responses, because the reason lives in the follow-up question. Most disagreements about research design are really disagreements about which of those two claims the business is about to make.
Can You Actually Reach the People You Need
Reach is the constraint that quietly decides the method. A browser-link study reaches people with a stable connection, a smartphone and the patience for a landing page, which is a narrower slice than most sample plans admit, particularly outside major metros and among older and lower-literacy respondents.
This is where channel choice does real work. Alchemic runs AI-moderated interviews as text natively inside WhatsApp, with voice notes supported and no link or app, as voice and video interviews on the browser, and by outbound AI phone interview to any working number, feature phones included. It publishes 57+ languages including Hindi, Tamil and Telugu, with managed fieldwork or bring your own recruitment across 14 markets including the USA and the UK.
None of that helps when the answer has to be projectable to a national population. That remains a probability-sample job, and an address-recruited panel run by a survey specialist is the right purchase even though it is slower and costs more.
Budget, Timeline and In-House Skill
Cheap research is usually worthless research, but expensive research aimed at the wrong question is worse, because it also buys confidence. Spend on the single decision that most needs evidence rather than spreading the budget across four.
Timeline is a constraint, not a preference. If the decision is Friday, a multi-month design does not help, and the honest answer is secondary data plus a handful of interviews. Buying access to a platform also does not buy the judgment to design the instrument, and the tooling options are compared in the best market research software platforms.
What Market Research Cannot Settle
Market research cannot tell you what people will actually do, only what they say and what they have done. It cannot make a sample representative by measuring it more precisely, and it cannot settle a decision that is really about appetite for risk. Three limits are worth stating before any study starts.
Saturation is a stopping rule, not a claim about the population. Knowing that new information fell to 5% or less of the early total after eight or nine interviews tells you those interviews stopped being informative. It says nothing about how common those themes are among everyone you did not interview, and reporting a qualitative theme as a percentage is where most research misuse begins.
Published data lags unevenly, and the cadence matters more than the source. The Economic Census is the official five-year measure of US businesses and the benchmark for GDP and the Producer Price Index, which makes it authoritative and, in a fast-moving category, two or three years behind the question you are asking. Check the collection year before quoting any published number.
Cheaper designs are sometimes genuinely sufficient. In one county-level seroprevalence study covering a population of roughly 800,000, a convenience sample and a probability sample produced similar estimates once both were weighted to known sex, race and geography (American Journal of Public Health commentary). The lesson is not that convenience samples are fine. It is that the premium for a probability design is worth paying when the estimate is the deliverable, and harder to justify when the direction is.
Putting Market Research Into Action
Research earns its cost at the point where a finding changes a decision, not when the report lands. Three moves do most of that work: share findings beyond the team that commissioned them, join them to operational data, and use them as the evidence base for budget requests.
Sharing is the cheapest of the three and the most often skipped. One customer research project can inform a product roadmap, a messaging rewrite, a sales training deck and a support priority list at once, and none of that happens if the deck stays with whoever paid for it.
Integration compounds the value. Survey responses read differently next to purchase history and support tickets, and the combination catches the gap between what people said and what they did. Speed helps here: a 200-interview qualitative study from Alchemic takes about three days to a live dashboard, and five to seven for complex designs, short enough to sit inside a planning cycle rather than after it.
Business cases are where research converts into budget. When you can show that a named unmet need exists and size the opportunity, funding the work becomes an argument about numbers rather than an argument about taste.
The organizations that get the most from this invest before a crisis forces them to and act on findings that contradict what they already believed. Combining the types is what makes that possible: secondary sets the context, primary fills the gaps, and the four objective-led types answer the four questions a growth strategy actually asks.