The 7-Step Product Research Process (And Where Most Teams Skip a Step)
Product research is the process of gathering evidence about customers, competitors and demand before and after a product is built. To do product research, define the decision you need to make, review secondary data, specify and recruit a sample, choose qualitative or quantitative methods to match the question, run fieldwork, and deliver a verdict with its conditions.
What Product Research Is, in One Sentence
Product research is the structured collection of evidence about customer needs, competitive alternatives and willingness to pay, used to decide whether a product should be built, changed, priced differently, or abandoned.
Product research runs across the whole product lifecycle, not only before launch. It answers four questions:
Is there a real problem here, and who has it badly enough to pay? Does our solution actually solve it, judged by people outside the building? What will they pay, and what do they give up to buy it? What is happening now that we did not expect, after the product is live?
Product research combines primary research, which is evidence you collect directly from people, with secondary research, which is existing data you analyse. Both belong in a serious study.
Product Research vs Market Research
Product research and market research overlap, and they answer different questions. Market research describes a market. Product research interrogates a specific offer inside that market.
| Product research | Market research | |
|---|---|---|
| Central question | Should we build, change or price this product? | How big and attractive is this market, and who is in it? |
| Unit of analysis | A concept, prototype, feature set or live product | A category, segment or territory |
| Typical outputs | Concept test scores, feature priorities, price sensitivity curves, usability findings | Market size, growth rate, segment profiles, competitor share |
| Who commissions it | Product, innovation and go-to-market teams | Strategy, corporate development and investors |
| Cadence | Continuous through development and post-launch | Periodic, often annual or deal-driven |
A product research programme that ignores market context tests a good product into a market that cannot sustain it.
The 7-Step Product Research Process
The product research process runs in seven steps, from defining the decision through to delivering a verdict. Teams that skip steps 1 and 3 produce research that is interesting and unused.
| Step | Output you should be able to show | Usually owned by |
|---|---|---|
| 1. Define the decision | A one-page research brief | Engagement lead |
| 2. Secondary research | A gap list of what desk data cannot answer | Analyst |
| 3. Specify and source the sample | Screener plus confirmed participant list | Research operations |
| 4. Select methods | Method-to-question map | Research lead |
| 5. Fieldwork | Transcripts, recordings, response file | Moderator or field team |
| 6. Analysis | Findings mapped to the original decision | Research lead |
| 7. Delivery | Verdict, confidence level, kill condition | Engagement lead |
Step 1 — Write the Decision Before You Write the Questionnaire
Product research fails most often because nobody wrote down what the research was meant to settle. Before any method is chosen, write the 6-Line Research Brief:
The decision. The specific choice this research must inform. The date. When the decision gets made regardless of whether research is finished. The audience. Who has to be convinced, and what they currently believe. The respondents. Whose opinion legitimately counts, stated as a job title or behaviour, not a demographic. The falsifier. What finding would change the team’s mind. If nothing would, stop here. The cost of error. What being wrong costs in money, time or reputation.
Line 5 is the one that gets skipped and the one that matters. [EXPERIENCE: insert a Nexus engagement where an explicit falsifier changed the study design or killed a planned build.]
Step 2 — Exhaust Secondary Research Before You Spend on Primary
Secondary research is existing evidence: regulatory filings, industry reports, patent databases, pricing pages, app store reviews, support ticket exports, sales-call recordings, and public statistics. Work it first, for two reasons. It is cheap, and it stops you spending expensive primary interview time asking questions a public document already answers.
The output of step 2 is a gap list. Every remaining unknown becomes a primary research objective
Step 3 — Define the Sample, Then Check You Can Actually Reach It
Most research plans define the ideal respondent and assume availability. Run the Recruitment Reality Check on the sample before the timeline is agreed. Four factors decide whether a sample is easy or brutal:
Incidence rate. What share of the general or professional population qualifies. A sample that is 1 in 1,000 needs a different sourcing approach than 1 in 10. Seniority. Director-level and above respondents rarely sit in consumer panels and rarely respond to unsolicited survey invitations. Disclosure exposure. Employment agreements, NDAs and regulated-industry rules narrow who can legitimately speak and about what. Geography and language. A sample split across time zones roughly doubles scheduling overhead.
If three of the four factors are hard, plan for custom recruitment rather than panel sampling.
Step 4 — Match the Method to the Question
Choose methods against the question type, not against team preference or the tool already on contract. The method table in the next section is the working map. One rule covers most cases: if the question starts with “why” or “how,” the answer needs qualitative work; if it starts with “how many,” “how much” or “which of these,” it needs quantitative work.
Step 5 — Run Fieldwork in Waves
Run the first three to five interviews, then stop and review. Early fieldwork nearly always exposes a flawed screener, a leading question or a wrong assumption about vocabulary. Fixing that after interview four costs one day. Finding it after interview twenty costs the study.
Step 6 — Analyse Against the Decision
Analysis means mapping evidence to the decision written in step 1, not summarising transcripts. For qualitative data, code responses against the research objectives and count how many respondents support each finding, so that “several participants said” becomes “9 of 14 heads of operations said.” For quantitative data, report the confidence interval alongside the headline figure.
Step 7 — Deliver a Verdict and Its Kill Condition
Deliver a direct recommendation, a stated confidence level, and the condition under which the recommendation would flip. A verdict with a kill condition survives scrutiny. A findings deck without one gets debated for a month.
Product Research Methods, and the Questions Each One Answers
Product research methods divide into qualitative methods, which explain behaviour, and quantitative methods, which measure it. The table below maps the common methods to the question each is genuinely good at answering. Sample figures are planning conventions used widely in commercial research, not fixed rules.
| Method | Best question for it | Type | Typical sample per segment |
|---|---|---|---|
| In-depth interviews (IDIs) | Why do buyers behave this way? | Qualitative | 12–20 |
| Focus groups | How do opinions form in a group, and what language do people use? | Qualitative | 2–4 groups of 6–8 |
| Usability testing | Where does the product break for real users? | Qualitative | 5–8 |
| Diary or longitudinal studies | How does usage change over weeks? | Qualitative | 8–15 |
| Surveys and questionnaires | How widespread is this need or preference? | Quantitative | 200–400 |
| Concept testing | Which concept wins, and by how much? | Quantitative | 200–400 |
| Conjoint analysis | Which feature and price combinations drive choice? | Quantitative | 300+ |
| Van Westendorp / Gabor-Granger | What price range is acceptable or optimal? | Quantitative | 200+ |
| Product and web analytics | What are users actually doing in the product? | Quantitative | Full population |
| A/B testing | Which version performs better on a live metric? | Quantitative | Power-calculation dependent |
Qualitative vs Quantitative, by Question Type
Qualitative research explains why customers behave as they do, using interviews, focus groups, usability sessions and open-ended responses with small samples. Quantitative research measures how many customers behave that way, using surveys, analytics and structured experiments with larger samples.
The sequencing that works: qualitative first to learn the language and the range of possible answers, quantitative second to size those answers, qualitative again to explain anything the numbers cannot. Running a survey before any interviews produces clean data about the wrong questions, because the answer options were written from internal assumptions.
Jobs-to-be-Done and Voice of the Customer
Jobs-to-be-done (JTBD) is a framework that defines a product by the progress a customer is trying to make, rather than by customer demographics. In product research it is used to write interview guides that ask what the customer was doing immediately before they went looking for a solution, and what they were using instead.
Voice of the customer (VoC) is the ongoing programme that collects customer feedback across channels and routes it into product decisions. JTBD shapes what you ask during development. VoC is what keeps the evidence flowing after launch.
User personas are the summary output, not the research. A persona built from three internal opinions and a stock photo is a decoration. A persona built from 15 coded interviews is a working tool.
How to Run Customer Interviews and IDIs That Produce Usable Evidence
In-depth interviews produce usable evidence when they ask about past behaviour rather than future intent. Customers are unreliable predictors of what they will buy and reliable reporters of what they already did. Plan 45 to 60 minutes per interview, and expect thematic saturation, the point at which new interviews stop producing new themes, at roughly 12 to 20 interviews per distinct segment. Five practices that separate useful IDIs from pleasant conversations:
Ask for the last time, not the usual time. “Walk me through the last time you had to solve this” beats “how do you usually solve this.” Ban the pitch. If the participant learns what you want to hear in the first five minutes, the remaining 40 are worthless. Chase the workaround. The spreadsheet, the WhatsApp group or the manual step a respondent built themselves is the clearest statement of unmet need in product research. Ask what they stopped using, and why. Churn stories contain the objections your sales team will meet. Record, transcribe and code. Evidence that exists only in a moderator’s memory cannot be defended to a client.
Interviews are only as good as the person on the other end. A sharp guide asked of the wrong respondent produces confident, worthless data.
Focus Groups vs In-Depth Interviews: Which Fits Which Question
Focus groups and in-depth interviews are not interchangeable. Focus groups surface shared language and social dynamics. In-depth interviews surface individual reasoning and sensitive detail.
| Focus groups | In-depth interviews (IDIs) | |
|---|---|---|
| Best for | Reactions to naming, packaging, messaging, positioning | Buying processes, workflows, objections, switching decisions |
| Group size | 6–8 participants per group | 1 participant |
| Session length | 90–120 minutes | 45–60 minutes |
| Main risk | Dominant participants create false consensus | Sample too small to generalise |
| B2B suitability | Low. Competitors will not talk candidly in front of each other | High |
| Sensitive topics | Poor | Good |
| Scheduling difficulty | High. One diary clash can collapse a session | Moderate |
For B2B product research, IDIs are the default and focus groups are the exception. Senior professionals do not disclose procurement frustrations, vendor pricing or internal failures in front of peers from adjacent companies.
How to Validate a Product Idea Before You Build It
To validate a product idea, confirm four things in order: that the problem exists, that people currently pay something to solve it, that your specific concept is preferred over the alternatives, and that the preferred concept is acceptable at a viable price. A product idea that clears the first three and fails the fourth is a hobby.
Problem validation. 12 to 20 IDIs with people who have the problem now, looking for existing workarounds and current spend. Solution validation. Show a prototype or a clickable concept and watch what they do, not what they say. Concept testing at scale. Quantify preference against competing concepts and a “none of these” option. Pricing research. Establish the acceptable range before the roadmap is locked.
Concept Testing
Concept testing measures how a defined target audience responds to a product concept before development is committed, usually through a survey that presents one or more concepts and measures purchase intent, uniqueness, relevance and believability. Always include a “none of these” option. Forced-choice concept tests reliably overstate demand, because in the real market “I’ll keep doing what I do now” is always on the shelf.
Pricing Research
Three named methods cover most product pricing research. Van Westendorp price sensitivity asks four open price questions to establish an acceptable range. Gabor-Granger tests purchase intent at specific price points to model a demand curve. Conjoint analysis presents bundles of features at different prices and derives which attributes actually drive choice. Conjoint is the most informative and the most expensive, and it needs the largest sample.
Product-Market Fit Signals
Product-market fit validation is a matter of behavioural evidence, not survey sentiment. The signals that hold up in front of an investor or a board:
Retention that flattens rather than decaying to zero Customers using the product in ways nobody instructed Unprompted referrals and inbound demand Willingness to pay before a feature is finished Rising usage in the absence of marketing spend
Competitive Analysis and Market Sizing for a New Product
Competitive analysis in product research means documenting what a customer would realistically buy instead of your product, including the option of doing nothing. Capture, for each alternative: target segment, headline benefit, pricing band, distribution channel and the specific complaints customers make about it in public reviews. Those complaints are the fastest route to a defensible positioning gap.
Market sizing establishes the commercial ceiling: TAM (total addressable market) is everyone who could theoretically buy, SAM (serviceable addressable market) is the portion your model can actually serve, and SOM (serviceable obtainable market) is the share you can realistically win in a defined period. Size the market before running expensive fieldwork, because a study that proves demand inside a market too small to fund the build has answered the wrong question.
How to Recruit Research Participants (Including the Ones Who Don’t Answer Panels)
Recruiting research participants means defining a qualifying respondent, screening candidates against that definition without telegraphing the right answers, and confirming their availability inside your fieldwork window. Recruitment is where product research timelines usually break, and it gets harder as respondents get more senior, more niche or more regulated.
Panels, Own Customers and Expert Networks Compared
| Source | Best for | Speed | Main limitation |
|---|---|---|---|
| Consumer panels | High-incidence consumer samples, large quantitative studies | Fast | Professional respondents; almost no senior B2B coverage |
| Your client’s own customers | Usability testing, retention and churn research | Fast | Survivor bias; excludes people who never bought |
| Lost deals and churned accounts | Objection and pricing research | Moderate | Access depends on the client’s CRM hygiene |
| Social and community sourcing | Niche enthusiast and practitioner samples | Variable | Self-selection bias; heavy screening load |
| Expert networks with custom recruitment | Low-incidence, senior or specialist samples | Fast when recruited to spec | Higher cost per completed interview |
Screening Without Coaching the Answer
A screener that reveals what you are looking for will find you people who say it. Rules that keep a screener honest:
Ask about behaviour and responsibility, never about interest in the category Bury the qualifying criterion among plausible alternatives in multiple-choice options Verify role with a task question (“who signs off on this purchase in your organisation?”) rather than a title question Screen out anyone working for a competitor, a research agency or the client Re-verify at the start of the session, because screeners get gamed
When Your Sample Is Hard to Reach
Some product research samples do not exist in any panel. Air traffic controllers, hyperscaler practice directors, and heads of crypto trading are not going to be found by widening a panel filter. For those samples, recruitment has to be done from scratch for the engagement: identify the specific individuals who meet the specification, approach them directly, confirm they can speak within their compliance obligations, and schedule inside the client’s window.
Nexus Expert Research recruits from scratch for every engagement for exactly this reason, and delivers experts in that seniority band in under 24 hours.
How Long Product Research Takes and What Drives the Cost
Product research timelines are driven by recruitment difficulty far more than by method complexity. The ranges below are planning guidance for a single, well-specified study. [STAT NEEDED: Nexus internal medians for each row, which would replace this generic guidance with proprietary, citable data]
Study type Typical end-to-end timeline Main cost driver Desk and secondary research 3–5 working days Analyst time, paid database access 12–20 IDIs, high-incidence sample 2–3 weeks Recruitment and moderator time 12–20 IDIs, senior or low-incidence sample 2–4 weeks Custom recruitment and honoraria Focus groups (2–4 sessions) 3–4 weeks Scheduling, facility or platform, incentives Quantitative concept test (n=300) 2–3 weeks Sample cost, which rises steeply as incidence falls Conjoint study 4–6 weeks Design complexity and sample size Usability testing round 1–2 weeks Participant recruitment and session analysis
Four factors move the cost of a product research project more than anything else: incidence rate (rarer respondents cost more per completed interview), seniority (honoraria scale with the respondent’s time value), number of distinct segments (each segment needs its own sample), and speed (compressed fieldwork windows cost more to staff).
When Product Research Is the Wrong Investment
Product research is not always the right spend, and saying so protects the studies that do matter. Do not commission it when:
The decision is already made. Research commissioned to justify a decision produces evidence that gets quoted selectively and trusted by nobody. The cost of being wrong is lower than the cost of the study. If a feature takes two days to build and one day to remove, ship it and measure. The concept cannot be described in one screen. Respondents cannot react to something they do not understand, and their confusion will be recorded as rejection. You need an answer faster than fieldwork allows and no shortcut exists. A rushed study with a compromised sample is worse than an openly stated assumption, because it carries false authority. The question is legal, regulatory or safety-critical. Those need qualified professional advice, not customer opinion.
Product research also has limits worth stating to a client up front. Stated intent overstates real purchase behaviour. Small qualitative samples describe a range of views and do not measure their prevalence. And every study is a snapshot of a market that keeps moving.
Product Research FAQ
How many people do I need for product research?
It depends on the method. Qualitative work commonly reaches saturation at 12 to 20 in-depth interviews per segment, usability testing typically uses 5 to 8 participants per round, and quantitative concept testing usually needs 200 to 400 responses per segment for subgroup analysis.
What is the difference between primary and secondary research?
Primary research is evidence you collect directly from people, such as interviews, surveys and usability tests. Secondary research is existing data you analyse, such as industry reports, filings, public statistics and customer support records. Most product research projects use both, secondary first.
Can AI or synthetic respondents replace real research participants?
No, not for decisions that carry real cost. Synthetic respondents can pressure-test a questionnaire, generate hypotheses and screen early concepts cheaply. They cannot tell you what a specific head of trading did last quarter or why a procurement process stalled, because they have no access to that reality. Use them to sharpen the study, then run it with real people.
How often should product research be repeated after launch?
Continuously for lightweight feedback, and on a fixed cadence for structured studies. Most teams run in-product feedback and satisfaction tracking constantly, and repeat structured concept, pricing or positioning studies annually or whenever the competitive set changes materially.
What is product discovery research?
Product discovery research is early-stage exploratory work that identifies which customer problems are worth solving, before any solution is specified. It is almost entirely qualitative, uses interviews and observation, and is the stage where jobs-to-be-done framing is most useful.
Who should do product research, an in-house team or an agency?
Use in-house teams for continuous work on your own users, where product access and context matter most. Use an external research partner when you need an unbiased moderator, a sample you cannot reach, specialist methodology such as conjoint, or capacity inside a fixed deadline.
Your Next Step
Before the next research plan leaves your office, price the recruitment, not the methodology. Check the four factors in the Recruitment Reality Check against your respondent specification: incidence, seniority, disclosure exposure, geography. Three hard factors means panel sampling will miss your window.
That is the point to bring in custom recruitment rather than the point to extend the fieldwork dates. Nexus Expert Research sources named specialists worldwide, built to each engagement’s specification.
