5 tips to evaluate market research skills

To evaluate market research skills, give the candidate a real research brief and score how they handle it. Watch which question they ask first, which method they pick, how they defend the sample size, and what recommendation they actually land on. A resume shows you none of that. A 45-minute work sample shows you all of it.

Here is the gap that costs teams money. Plenty of candidates can define a focus group. Far fewer can tell you when a focus group is the wrong tool, or notice that a survey with 25 responses cannot carry a pricing decision. Vocabulary is easy to rehearse. Judgment is not, and judgment is the thing you are paying for.

So this guide does two jobs. First it walks the research process properly, because you cannot score a skill you cannot describe. Then it turns that process into a rubric, a work sample, and questions you can use this week.

TL;DR

• Market research skills are judgment skills, not vocabulary. Test them with a short brief and a messy dataset, never with a multiple-choice quiz about research terms.
• Run the process yourself first. Objective, method, sample, collection, analysis, recommendation. Those six steps become the six rows of your scorecard.
• Score five things: how they frame the problem, how they choose a method, how they handle data, how hard they push on the evidence, and how clearly they explain it to someone who will never read the appendix.
• The strongest signal is a candidate who tells you what their own research cannot prove. That instinct is rarer than technical skill and it protects you from expensive decisions.
• Ask for the recommendation, not the report. Research that does not end in a decision is a hobby.

How do you do proper market research?

Proper market research means starting from a decision you need to make, choosing a method that can actually answer it, collecting data from people who represent your real market, and ending with a recommendation someone can act on. Skip the decision at the start and you get interesting facts nobody uses.

The process breaks into six steps. Learn them in this order, because the order is what separates research from data collection.

1. Write the decision, not the topic. Not "research our customers" but "decide whether to launch the mid-tier plan at $39 or $49". A decision has a deadline and a cost of being wrong. A topic has neither.
2. Pick the method that fits the question. Qualitative work (interviews, focus groups, open-ended questions) tells you why people behave the way they do and what language they use. Quantitative work (surveys, transaction data, experiments) tells you how many and how much. Most real questions need a little of both, in that order: talk to 12 people to learn what to ask, then ask 400 people to find out how common it is.
3. Decide who counts as your market, then sample it honestly. This is where most research quietly breaks. If you survey your existing customers about why people buy, you learn nothing about the people who did not. Write down who you are excluding before you start.
4. Collect the data without leading the witness. "How much do you love the new dashboard?" is not a question, it is a prompt. Neutral wording, randomized answer order, and a pilot run with five people to catch questions that read differently than you intended.
5. Analyze for patterns, then look for the pattern that would prove you wrong. Anyone can find support for the thing they already believed. Good analysts go hunting for the counter-example, and say so when they find it.
6. Report a recommendation with its confidence attached. One page. What you learned, what you recommend, what would change your mind, and what the research cannot tell you.

Secondary research (published reports, government data, competitor material that is already public) comes before all six steps, not instead of them. It is cheap and it stops you paying to rediscover something that is already known. Primary research is what you run yourself when the answer you need does not exist yet.

How to write a research brief that gets used

Most research fails at the brief, not at the analysis. A brief that works fits on one page and answers five things: the decision being made, who makes it, when they make it, what they currently believe, and what evidence would change their mind. That last item is the one teams skip, and skipping it is how you end up with a study that confirms whatever the loudest person already thought.

Write the brief with the decision-maker in the room. Ten minutes of "what would actually change your mind here" saves three weeks of research nobody uses. If the honest answer is "nothing", stop. You do not have a research problem, you have a disagreement, and no amount of data will settle it.

One more test before you commission anything. Write the two headlines the research could produce, the one you expect and its opposite, and ask whether you would do something different in each case. If both headlines lead to the same action, the research is decoration.

How big should your sample be?

There is no universal number, and any candidate who offers one without asking about the decision has told you something useful about themselves. Sample size follows from three things: how precise the answer needs to be, how varied the population is, and what it costs to be wrong.

Rough working rules that hold up in practice. For qualitative work, new themes usually stop appearing somewhere between 8 and 15 interviews within a single, well-defined audience, so 12 is a sensible default and more is often waste. For quantitative work, precision improves with the square root of the sample, which is why moving from 100 to 400 responses halves your margin of error while going on from 400 to 700 buys you far less. That square-root relationship is worth knowing because it explains why the jump from a tiny sample to a modest one is the valuable one, and why paying for a very large sample rarely pays back.

Sub-groups are where sample plans quietly fail. A survey of 400 that you then want to split four ways is really four surveys of 100, and nobody notices until the segment charts start moving on two or three responses. Decide your cuts before you field, not after.

Competitor and market analysis without a big budget

Market sizing does not require a paid report. Build it from the bottom up: how many organizations fit your buyer description, what share realistically could buy in a year, and what they would pay. Show the arithmetic so anyone can challenge an input. A bottom-up estimate you can defend beats a top-down number from a report nobody on your team can interrogate.

For competitor work, the public surface tells you more than most people use. Pricing pages, job postings (which reveal what a company is building before it ships), documentation, release notes, and review sites all update themselves for free. Track a handful of specific things on a schedule rather than doing one enormous sweep a year, and record the date next to every observation, because a competitor fact without a date is a liability the moment someone quotes it six months later.

Keep it grounded in your own decision. Competitor analysis that is not tied to a choice you are about to make turns into a newsletter.

Why continuous research beats the annual study

A single large annual study gives you a precise read on a market that has already moved by the time the deck is finished. Small continuous studies give you a rougher read that is always current, and they let you spot a change while you can still respond to it.

The practical version for a small team: one recurring question set you run every quarter with the same wording, plus short ad-hoc studies when a specific decision comes up. Keeping the wording identical is what makes the comparison real. Change the question and you have restarted the clock, which is a mistake experienced researchers protect against and inexperienced ones make without noticing.

Which market research skills actually predict performance?

Five capabilities predict whether someone will do useful research: problem framing, method selection, data handling, critical evaluation of evidence, and communication. Everything else, including tool familiarity, can be taught in a couple of weeks. These five cannot, and they are the ones a resume hides best.

Analytical thinking is the skill employers name most often across every industry. Seven in 10 companies call it essential, according to the World Economic Forum's Future of Jobs Report 2025. For a research hire it is not one skill among many. It is the job.

Skills needed for market research roles

The demand is steady rather than explosive. The U.S. Bureau of Labor Statistics projects employment of market research analysts to grow 7% from 2025 to 2035, with about 82,000 openings a year over the decade, and reports a median annual wage of $78,760 as of May 2025. A bachelor's degree is the typical entry point, which means the credential filters almost nobody out. You need a different filter.

Here is what each capability looks like in practice, and how to get evidence of it rather than a claim about it.

Marketing research skills vs market research skills

People use the two phrases interchangeably in job ads, and then interview for the wrong thing. Market research studies a market: customers, demand, competitors, pricing, size. Marketing research is broader and covers the whole marketing operation, so it also takes in campaign testing, message performance, channel mix and brand tracking.

The practical difference is what you put in the work sample. A market research hire should be able to size an opportunity and describe a buyer. A marketing research hire should be able to read an A/B test and tell you whether the lift is real. Decide which one the role needs before you write the brief, because a candidate strong in one can look weak in the other for no good reason.

Market research capability assessment rubric

A market research capability assessment works best when the same five rows are scored by more than one person, on the same evidence, without anyone seeing the others' scores first. Use a simple 1 to 5 band per capability and write the anchor for a 3 before you start, so "average" means the same thing to everyone on the panel.

This is the Testlify Competency-to-Evidence Matrix applied to one role. The matrix maps every role to the competencies that matter, then connects each competency to measurable evidence through assessments, simulations, interviews, references and structured feedback. You start with the role, not with a test. For a research hire that means the five rows above become five evidence sources, and no row gets scored on a hunch.

Pro tip: weight the rows before anyone sees a candidate. If data handling matters twice as much as communication for this role, set it that way up front. Deciding weights after you have met a candidate you like is how a rubric turns into a justification.

What market research tips help you hire better?

Five market research tips, in the order they will save you the most time.

1. Replace the portfolio review with a fresh brief. Past work is real but unverifiable: you cannot tell what the candidate did and what their manager did. A short new brief on your own data is the only sample you can attribute with confidence.
2. Use your own messy data, not a clean demo file. Duplicate rows, a date column stored as text, 40 people who skipped question four. The mess is the test. Anyone can chart a tidy file.
3. Score the questions they ask, not just the answers they give. Send the brief with one detail deliberately missing. A strong candidate writes back asking for it. A weak one guesses and carries on.
4. Cap the task at 45 minutes and say so. Unpaid multi-day take-homes cost you good candidates who already have jobs, and they reward free time rather than skill. If you need more than 45 minutes to see the skill, your brief is doing too much.
5. Ask for the decision, then the confidence. "What should we do, and how sure are you?" The second half of that question is where the real ones separate themselves.

Which market assessment techniques work best?

The market assessment techniques that predict performance all share one property: the candidate has to produce something, not describe something. Ranked by how much signal they give per minute of everyone's time:

• The work sample. A one-page brief, a messy spreadsheet, 45 minutes, one recommendation. Highest signal by a wide margin, because it is the job.
• A spreadsheet task. Testlify's question types include live Microsoft Excel and Google Sheets work, so the candidate does the pivot in the real application rather than answering trivia about it.
• Situational judgement. Three methods, one budget, one deadline, pick and justify. Cheap to run at volume and it exposes method selection fast.
• A structured interview on the sample they just produced. Not "tell me about a time". Instead: "why this sample size", "who is missing from this data", "what would change your recommendation".
• Reference checks, last and narrow. One question only: did their research change a decision, or did it get filed?

Here is how that fits together for a hypothetical 40-person marketing agency hiring its first dedicated researcher. Sixty applicants arrive. All sixty take a 45-minute assessment combining a role-specific market research test with a spreadsheet task. Twelve clear the scoring threshold. Those twelve get a 30-minute structured interview built on their own submission, scored by two reviewers against the same five rows. Three finalists reach the founder, who now reads three comparable scorecards instead of sixty resumes. The point is not speed. The point is that every stage after the first is spent on people you already have evidence about.

If the role leans analytical rather than commercial, a business analyst assessment covers more of the modelling side, and a broader test of research and analysis ability suits generalist roles where research is one duty among several. For the interview stage itself, these questions for research assistants give you a structured starting set, and the market research analyst hiring guide covers the scorecard and the offer stage.

What to put in the candidate's work sample

Build the brief from a decision you genuinely made in the last year, with the outcome removed. Give them a short context paragraph, a spreadsheet with 200 to 400 rows of real (anonymized) data, one specific question, and a 45-minute cap. Leave one detail out on purpose, and tell the scorers which detail it is, so "asked for the missing input" becomes a row on the scorecard rather than a nice thing someone happened to notice.

Anonymize properly before it leaves the building. Strip names, emails, account identifiers and anything that could re-identify a customer, and check free-text columns, which is where personal data hides. If a file cannot be safely shared, build a synthetic one with the same shape and the same mess.

Score it blind where you can. Reviewers who do not see the candidate's name or CV score the work rather than the person, and the difference shows up most on candidates from non-obvious backgrounds, which is exactly the group a skills-first process is supposed to surface.

What are the limitations of market research?

Market research tells you what a sample of people said or did at one moment. It cannot tell you what everyone will do next. The honest limitations are sampling, self-selection, timing, and the gap between stated and actual behaviour, and a candidate who names them unprompted is worth more than one who does not.

The most misunderstood limitation is response rate. Everyone knows response rates have collapsed. Pew Research Center reported that typical telephone survey response rates fell to 6% in 2018, down from around 9% in the years before. The obvious conclusion is that low response rates ruin the data.

That conclusion is wrong, and this is the part worth testing candidates on. Pew's own methodological work found that response rate is an unreliable indicator of bias: most political and social measures barely moved between low-response and high-effort surveys, while civic engagement measures were overstated by between 9 and 38 percentage points. The bias was large on exactly the topics where the kind of person who answers surveys differs from the kind who does not.

So the right question is never "what was the response rate". It is "who is more likely to answer this particular question, and does that skew this particular number". A candidate who gets that distinction will save you from a bad decision. Ask it directly in the interview and listen for whether they reach for the headline number or the mechanism.

Two more limits worth naming. People are unreliable narrators of their own future behaviour, so intent-to-purchase numbers run high and should never be taken at face value. Ask 100 people whether they would pay $20 a month for a new feature and a cheerful share will say yes; ask them to enter a card and the number collapses. The fix is not a better survey question. It is to treat stated intent as a ceiling, not an estimate, and to test the real behaviour with a small experiment wherever one is possible.

Research also decays. A market read in January can be wrong by June, which is why continuous small studies beat one annual monolith, and why a dated finding quoted without its date is one of the more expensive habits a team can pick up.

And some questions research cannot answer at all. It will not tell you whether a genuinely new product will work, because nobody can report on an experience they have never had, and it will not settle a question that is really about strategy or appetite for risk. Knowing which questions to refuse is part of the skill. A candidate who says "research would not help here, and here is why" in a work sample has just shown you something better than a clean chart.

Hire researchers on evidence, not a rehearsed interview

Assessment first, conversation second. When the candidate has already produced work on your data, the interview stops being a performance and becomes a discussion about something real, and your scorecards become comparable across every applicant rather than a memory of who interviewed well.

You can build the whole thing on Testlify: role-specific and situational-judgement tests, live Excel and Google Sheets tasks, custom questions of your own, weighted scoring so the capabilities that matter count for more, percentile benchmarking to see how a candidate compares, and multiple reviewers scoring the same evidence independently. Want to see it against your own brief? Book a 30-minute walkthrough and bring the role you are hiring for.

Key takeaways

• Test judgment, not vocabulary. Definitions are rehearsable and every candidate has rehearsed them, so a terminology quiz ranks people by preparation time rather than ability. Give a brief with a real decision attached and the ranking changes completely.
• Describe the process before you score it. The six steps (decision, method, sample, collection, analysis, recommendation) are not filler for the reader. They are the rows of your scorecard, which is why a hiring team that cannot articulate the process ends up scoring on confidence and polish instead.
• The credential filters nobody. A bachelor's degree is the typical entry point for the role, so requiring one removes almost no candidates and tells you almost nothing. Replace the qualification screen with an evidence screen and you widen the pool and sharpen it at the same time.
• Use messy data on purpose. A clean file tests whether someone can operate a chart tool. Duplicates, text-formatted dates and missing answers test whether they notice, which is the difference between a report that is pretty and a report that is right.
• Response rate is not bias. Pew's research shows a low response rate can leave most measures nearly untouched while badly skewing the ones where responders differ from non-responders. A candidate who reaches for the mechanism rather than the headline number will stop you acting on a broken finding.
• Ask for a recommendation with a confidence level. Research that ends in a summary rather than a decision creates work instead of removing it. Candidates who volunteer what their own analysis cannot prove are rarer than strong technical candidates, and they cost far less in bad calls later.
• Score independently, then compare. Two reviewers marking the same five weighted rows without seeing each other's scores gives you a defensible decision and a record of why, which matters more the moment a hire is questioned or a rejected candidate asks.

FAQs