Your quarterly plan says the positioning decision was informed by research. Now find the research. Most marketing teams cannot, because the plan cited research and the only techniques the team actually ran were a competitor page and a gut check. Here are 12 techniques grouped by what they answer, what they cost and what they cannot tell you.
Your quarterly plan says the positioning decision was informed by research. Now find the research. If you had to hand someone the notes, the sample and the date the field work closed, could you do it in under a minute? Most marketing teams cannot, and it is not because they are careless. It is because the plan cited research, and the only market research techniques most teams actually ran were a competitor page and a gut check. Roughly 58% of companies still base at least half of their regular business decisions on gut feel rather than data, according to compiling statistics from 2026. Six in ten marketers say their budgets are under more scrutiny than ever, and only a third invest in the structured data and metadata that would make their own reporting trustworthy. So which of your last three decisions was actually based on something you learned about the market, rather than on what you already believed?
The uncomfortable part is not the amount of research you skipped. It is that a focused study now costs less than a single misplaced campaign flight, and the barrier is rarely money.
Primary vs Secondary Research: The Split That Decides Your Budget
Every technique below is either primary or secondary, and the types of market research split on that line before anything else. That split decides what your study costs before you have chosen a technique.
Primary research means you collect the data yourself. You recruit the people, you write the questions, you own the answers. It is the only way to find out something the market has not already written down, and it is the only way to test willingness to pay. It is also the slow and expensive half: a custom study bought from a traditional research firm runs $15,000 to $50,000 or more per project across four to twelve weeks, and a boutique firm or freelance researcher lands somewhere in the $5,000 to $25,000 range across two to six weeks (published 2026 cost estimates).
Secondary research means the data already exists. Competitor pages, review sites, search demand data, industry reports, your own sales calls and support tickets. It costs you the time to find it and the judgement to verify it, and in a marketing team that time is usually already spent somewhere else in the week.
Sorting studies into these two buckets also fixes the single most common sequencing error. Teams commission primary research for questions their own data already answers, then complain that research takes too long. Before you approve any paid study, run the free pass first. Your own review page, your sales-call recordings and your search demand data will settle a surprising share of buyer questions.
Two of the techniques below are technically secondary, and both belong before anything you pay for.
Search and demand analysis tells you whether the market is real, growing or shrinking, and how buyers phrase the problem. You can run it this week on your own site: Search Console queries, keyword volumes, trend direction, the exact language in the impression data. Cheapest way to avoid researching a market that does not exist.
Industry report triangulation is the other half of that free pass: market size, direction and stated priorities. Good context for a deck, useless as an answer to your pricing question.
The practical rule: exhaust secondary before you pay for primary, and be honest with yourself about what "exhaust" means. Two hours on competitor pricing pages and review mining counts. Copying a stat from a report you have not read does not.
Qualitative Techniques: Four Ways to Find Out Why
Qualitative research answers "why" and "how". It is directionally useful at sample sizes a statistician would reject as a number, and that is fine as long as you never quote a percentage from it. Four market research methods carry most of the load on this side.
In-depth interviews (IDIs) ask people to explain a decision they have already made: why they chose you, why they left a competitor, what made them hesitate at the checkout or at the signature. Twenty interviews is a serious study. Cost lands between $200 and $500 per interview on one published 2026 estimate and $500 and $1,500 per interview on another, a 2.5x disagreement. Assume the middle until a vendor quotes you, and budget a 20-interview study at four to eight weeks including recruitment. What interviews cannot tell you is prevalence. They will not tell you how many buyers feel that way, and interview samples are too small to extrapolate.
Focus groups put a concept in front of a group and let the discussion go where it goes. The useful output is not consensus, it is the moment someone disagrees. A single session runs $6,000 to $15,000, and a real study of four to six groups runs $24,000 to $90,000, over four to eight weeks (published 2026 estimates). One group is not a study. It is a listen, and the sample is too small to project onto anyone outside the room.
Ethnographic and observational research is what people actually do, rather than what they say they do. Watch a buyer use your product, sit with a customer through their first week, watch how a prospect builds a shortlist. It is slow, it is expensive, and it does not scale, which is exactly why it finds things a survey cannot reach. Most teams run it once, learn something structural, and never repeat it.
Open-ended survey questions are the hybrid instrument. They cost almost nothing to add to a survey you are already fielding, and they return the language buyers use, verbatim. Do not treat them as statistically meaningful. Treat them as a free dictionary for next quarter's message testing.
One trap runs through all four. Leading questions ruin qualitative research faster than small samples do. "How frustrating was the pricing page?" guarantees a complaint. "Walk me through what happened when you looked at the pricing page" leaves room for the answer you did not expect, which is the only kind worth paying for.
Hanover Research's State of Market Research found that only 40% of companies run in-depth interviews and just over half ran a focus group in the last year. For the technique most responsible for explaining a decision, that number is low.
Quantitative Techniques: Four Ways to Count What You Found
Quantitative research answers "how many", "how much" and "how often". It is where a finding becomes defensible in a room with finance, and it is where small samples get quietly over-claimed.
Structured surveys are the workhorse. The response rate depends far more on the list than on the questionnaire, and the questionnaire has a hard ceiling: roughly 15 to 20 questions before completion falls off, and engagement drops past about ten minutes of completion time. Pricing sits anywhere from $2 to $50 per response, with general-population samples at $15 to $50 and niche B2B audiences, such as mid-market SaaS CTOs, at $30 to $80. A 1,000-response study typically fields over two to four weeks (published 2026 estimates). Split those responses into groups and you have an audience model, which is where segmentation, targeting and positioning starts.
Conjoint analysis shows people realistic product combinations at different prices and works out which features actually trade off against each other. It is the closest thing to a pricing instrument in this list, and it needs a real panel plus someone who can read the output. Not a DIY instrument.
MaxDiff forces a ranked preference across a long attribute list, which is the cure for the "everything is important" answer that flattens most feature surveys. Like conjoint it needs a real panel, though modern platforms build the analysis in for you.
A/B and pricing tests are the fastest quantitative method and the only one that requires no recruitment at all. You already have the traffic. What they cannot do is explain the result, which is why the winning variant tells you what worked and the interviews tell you why.
Inside the survey bucket sits one more option, and it needs a straight answer rather than a debate. Synthetic respondents, meaning AI-generated survey participants, are usable for a directional read on quantitative trends and should not be trusted for much beyond that. In a Strat7 study reported by Research Live in 2026, synthetic respondents came within 2 to 3 percentage points of real participants on simple headline results, but overshot willingness to pay by 16% and failed to build coherent respondents on sub-groups, drivers, segments and change over time. Independent 2026 practitioner roundups put the match at 85 to 95% on calibrated quantitative trends and behavioural patterns, falling to 37 to 60% on multi-factor studies and lower still on qualitative depth. Use them to pilot your instrument. Do not use them to replace the answer, and treat any study that hides which kind of respondents it used as a study you cannot trust. If the tooling layer is what you are actually shopping for, our AI market research guide covers that workflow end to end; this article stops at the method.
The discipline underneath all of this is one sentence: a number is only worth quoting if you can name the sample and the question wording that produced it. A 100-response survey supports "worth investigating". A 400-response survey supports a decision.
Desk Research Techniques: Four Methods That Need No Participants
This is the half of the work that market research tools and platforms now do in the background, and it is the group a marketing team can run this week with no recruitment, no panel and no budget approval. It is also where most teams should start, regardless of what else they can afford.
Competitor teardowns answer what rivals ship, charge and position against. Their pricing pages, changelogs and release notes are public, and reading them monthly is the fastest available signal on where a competitor is heading. A teardown built on a template you reuse beats a one-off deep dive you never repeat, which is the argument for keeping a competitor research template in the same folder as your positioning work.
Search and demand analysis, the free pass from earlier, answers whether demand is real, whether it is growing, and how buyers phrase the problem. Volumes, trend direction and the actual query language. If you need the layer that ranks and optimises for this, our guide to keyword research covers it. The strategic version is simpler: if nobody searches for the problem in your words, your message is probably not their message.
Social listening answers what gets said about the category when nobody asked. Unprompted language and objection patterns are the payoff, and the honest limit is volume. Below a certain number of mentions a month you are reading anecdotes, not patterns. If you are choosing where to run this, the social listening tools comparison covers the range, and a wider competitive intelligence programme is what this becomes once it repeats monthly.
Review mining is the highest-signal free method in the list. G2, Capterra, Trustpilot and app-store reviews are structured complaints written by people with no reason to be polite. Read your competitors' two-star reviews and you have a message-testing brief in an afternoon.
What desk research cannot do is answer "will this specific group pay for this specific thing". That question requires asking, and asking costs money. Every method above is a way to spend less on the methods below, not a way to avoid them.
Which Technique for Which Question
Knowing how to conduct market research comes down to one habit: name the question on the table and the confidence the decision actually needs, then pick the method. Almost nobody sequences it that way, which is why this matrix is the part worth keeping.
The question on the table | Technique to run | Time | Published 2026 cost | Confidence you can honestly claim |
|---|---|---|---|---|
Will they pay this much? | Conjoint analysis, pricing tests | 3–8 weeks (tests: 2–6) | Panel cost from $2–$80 per response | Decision-grade with a 400+ sample |
Why did they choose or leave us? | In-depth interviews (IDIs) | 4–8 weeks | $200–$1,500 per interview | Directional only, never a percentage |
How does the category react to a new concept? | Focus groups | 4–8 weeks | $6,000–$15,000 per session | Directional, small sample |
Is demand real and growing? | Search and demand analysis | Same week | Effort only | High on direction, none on willingness to pay |
Which words land with buyers? | Message tests, review mining | 1–2 weeks | Effort only | Directional, strong on language |
Who is actually in this audience, and how do they split? | Structured survey, segmentation work | 2–6 weeks | $15–$80 per response | Decision-grade at 400+, 600+ for sub-groups |
What will our competitor do next quarter? | Competitor teardown, release notes | Same week | Effort only | Directional, high on facts, low on intent |
What do buyers complain about in rivals' products? | Review mining | 2–3 hours | None | High signal, no statistical weight |
Everything feels important and we cannot rank it | MaxDiff | 1–3 weeks | Panel cost | Decision-grade with a real panel |
We build this in-house and it costs us nothing | In-house analyst time | Ongoing | $150,000–$175,000 per year fully loaded | Not a method, a budget line |
Two numbers in that table make one point. A 500-response survey and a single in-person focus group session land in roughly the same cost range: published 2026 estimates put them at $7,500 to $25,000 for the survey against $7,000 to $12,000 for one session. They answer completely different questions, so their prices are not comparable until you have named the question. The second number is the one teams skip. An in-house analyst is not a free option, at roughly $150,000 to $175,000 a year fully loaded (published 2026 estimate), and that budget line belongs in the comparison.
Freelance marketplaces will sell you survey-based research for an average of about $73 for a fixed-price project. That buys a template and a fast turnaround, and nothing about it supports a decision worth more than the fee. Use it for a first look, never for the spend you are about to commit.
Running Research Without a Research Budget
A lean market research process is a sequencing problem, not a spending problem. Four decisions carry it.
Pick two methods, not twelve. A qualitative pass to find out which questions are worth counting, then a quantitative pass to count them. Going quantitative first means measuring options you have not yet discovered, which is how teams end up with clean data about the wrong problem.
Decide the sample before you decide the tool. 100 respondents carries a margin of error of about ±9.8% and supports early exploration. 300 gets you ±5.7%, which is enough for most business decisions. 400 carries ±4.9% and is the published industry standard for a reliable consumer read. Going from 100 to 400 respondents cuts the error roughly in half, while going from 1,000 to 2,500 buys about one percentage point. Above a population of roughly 100,000, market size barely changes the arithmetic.
Stop chasing response rate. Published 2026 benchmarks range from 5 to 15% for cold B2B outreach to 20 to 30% for existing account holders, with platform data reporting figures as high as 91.61% on paid panels, where you are simply paying for completion. Those numbers come from different channels and incompatible studies, so do not read them as one trend line. What matters is the line underneath them: non-response bias matters more than response rate. A 50% response from a biased list is worse than a 15% response from a representative one, and chasing the number is how teams get a confident answer to the wrong question.
Write down what would change your mind before you run anything. One line is enough. If the answer comes back and nothing you would do changes, the study was decoration. If it comes back and you can name the decision it moves, you have your next study scoped already.
Five questions make a weak study worthless, and all five are avoidable at the design stage. Who is in the sample, and who is missing from it. What the exact wording was, because a leading question produces the leading answer. How many responses, and what that count honestly supports. When the field work ran, since a 2026 decision built on a five-year-old read is a decision built on a different market. And what the study did not ask, because the gap tells you what to research next.
Report it so it gets used: three lines at the top covering what you asked, what you found and what you are doing differently, with the sample statement on the same page. Resist presenting a single chart as the finding. The margin of error that says you cannot read that number matters more than the number itself.
Frequently Asked Questions
What are market research techniques?
Named methods for answering a specific question about a market, buyer or competitor. They split two ways: by the kind of answer they produce, where qualitative tells you why and quantitative tells you how many, and by who supplies the data, where primary means you collect it and secondary means it already exists.
What is the difference between primary and secondary research?
Primary means you collect the data yourself, which costs money and weeks. Secondary means it already exists publicly or commercially, which costs you the time to find and verify it. The practical rule is to exhaust secondary before you pay for primary, because a large share of your buyer questions are already answered for free by your own reviews, sales-call notes and search data.
How many people do I need for a survey?
400 respondents carries a margin of error of about ±4.9%, and that is the published industry standard for a decision-grade read. 100 respondents carries ±9.8% and supports exploration only. Above roughly 100,000 population size barely changes the math, so do not recruit for the size of your market.
Do I need a market research analyst to run these?
Not for the desk or qualitative techniques, and that is most of the list. Conjoint and MaxDiff are the two that realistically need a specialist, either a freelance researcher or a platform with the analysis built in. What the in-house route is not is free: a fully loaded analyst costs roughly $150,000 to $175,000 a year, which is why sequencing matters more than headcount.
Qualitative vs quantitative: which should I start with?
Qualitative, in almost every case. You run the qualitative pass to find out which questions are worth counting, then the quantitative pass to count them. Starting quantitative means measuring options you have not discovered yet.
How do I do market research with no budget?
Run the desk techniques: competitor teardowns, search and demand analysis, social listening and review mining. None of them need participants. The honest limit is that they cannot answer whether someone will pay, because that question requires asking, and asking costs money.
Can AI replace survey respondents?
For directional quantitative trends, synthetic respondents match real participants at 85 to 95% on calibrated trends and behavioural patterns. On a willingness-to-pay exercise they ran 16% above what real people said, and on sub-groups, segments and change over time they failed to produce coherent respondents. Use them to test your instrument and read the limitations section of anything built on them.
The techniques have not changed much in a decade. What changed is which ones you can run without writing a purchase order. Half of market research is finding and reading what the market already published, and that pass now runs in a conversation instead of a project: competitor teardowns, demand data and audience research in one workspace, with your human hours left for the interviews that need a person in the room. That is what our market research feature is for. Start free at studio.allable.ai.