QUALITATIVE RESEARCH FOR PRODUCT DISCOVERY
Quick Answer: Qualitative research for product discovery aims to surface the right problem before the team commits to a solution. The core methods — in-depth interviews (IDIs), ethnography, diary studies, and qualitative concept testing — each answer a specific question. The most common mistake product teams make is running too little discovery too late; the second most common is running large generic interview batches instead of purposive ones. This piece covers methods, sampling, synthesis, and governance.
WHY QUALITATIVE IS NON-NEGOTIABLE FOR DISCOVERY
Quantitative data tells you what is happening. Qualitative tells you why. Most product bets fail not because execution was weak but because the problem was misframed. You cannot un-misframe a problem with survey data.
The outcome of discovery is not a list of features. It is a clear articulation of the job being done, the friction, the stakes, and the alternatives — in the customer's words.
THE FOUR CORE METHODS
In-depth interviews (IDIs). 45–60 minutes, semi-structured, one-on-one. The workhorse method. Use for: understanding jobs, contexts, and decision processes; exploring emotional and social dynamics; generating hypotheses.
Ethnography / contextual inquiry. Observation in real context. Use for: complex workflows; environments where self-report misses critical detail; B2B processes with many stakeholders.
Diary studies. Participant-logged entries over days or weeks. Use for: over-time behaviour; emotional arcs; behaviours the participant would not recall accurately in an interview.
Qualitative concept testing. Mid-fidelity stimuli reviewed with target users. Use for: validating direction before quantitative preference testing or investment.
THE SAMPLING QUESTION
Qualitative sampling is purposive, not random. The goal is coverage of segments, contexts, and extremes — not representativeness in a statistical sense.
Defaults we use:
- IDIs: 8–12 per primary segment; add a few extreme users (power, novice, lapsed)
- Ethnography: 4–6 sites; stratified by context
- Diary studies: 15–25 participants; typically 7–14 days
- Qualitative concept testing: 15–25 per concept
Larger samples add marginal insight; smaller samples miss segment variation. Saturation — when new interviews stop producing new themes — is the decision rule, not a fixed count.
INTERVIEW PROTOCOL — WHAT WORKS
Open with the story, not the concept. "Walk me through the last time you…" produces richer data than "What do you think of…"
Probe for specifics. Numbers, dates, names, workflows. Abstract answers hide the real behaviour.
Silence is a tool. Pauses draw out the next layer. The best moderators talk less.
Never pitch. The moment the moderator advocates for a concept, the data is compromised.
Record everything with consent. Audio and video where feasible. Transcripts for synthesis.
SYNTHESIS — FROM INTERVIEWS TO INSIGHTS
The single highest-leverage practice is synthesis, and it is the step most often cut short. Options:
Affinity mapping. Team-based thematic clustering. Best for complex qualitative datasets; slower but deepest.
Coding framework. Pre-defined or grounded-theory-derived codes applied to transcripts. Best for multi-researcher consistency.
AI-assisted synthesis. LLM-based theme extraction and sentiment analysis on transcripts. Accelerates first-pass synthesis; human review required.
Journey mapping. Synthesis artefact for discovery research; living document across touchpoints, emotions, and systems.
Most projects we run combine AI-assisted first-pass synthesis with human-led affinity mapping for the final interpretation.
ETHNOGRAPHY — WHEN AND HOW
Ethnography is over-prescribed and under-resourced. It is the right method when:
- Workflow is complex and multi-stakeholder (hospitals, factories, trading floors)
- Self-report data is known to be unreliable
- Decision stakes are high enough to justify the cost
- The product will live in the context being studied
It is the wrong method when quicker methods would answer the question. Honest ethnography takes 4–6 weeks minimum and is expensive.
DIARY STUDIES — UNDERUSED AND POWERFUL
Diary studies capture behaviour over time that interviews cannot access: medication routines, parenting decisions, trading patterns, travel behaviour. Modern platforms (Dscout, Sprig, custom mobile apps) make diary studies easier than they used to be.
Common use cases: habit formation, emotional arcs during onboarding, multi-day purchase journeys, health behaviours.
GOVERNANCE
Consent and compensation. Every participant informed, consented in writing, compensated fairly. Fair compensation is typically $100–$250 per hour of qualitative participation for consumer research and $250–$500+ for professional/B2B.
Privacy. Video and audio stored securely; retention per policy. PII redacted from synthesis artefacts before wider distribution.
Method transparency. Every study ships with a methodology note: sample, method, limitations, ethical review.
Inclusion. Recruitment actively includes participants across age, ability, socioeconomic background, and ethnicity as relevant to the research question.
COMMON MISTAKES
Interviewing only happy users. Recruiters over-index on easy-to-reach customers. Purposively include lapsed, competitive, and prospective users.
Too-long interview guides. A 40-question guide produces a survey, not an interview. Guides should list 8–12 topic areas with probe examples.
Skipping synthesis. Transcripts are not insights. Plan 1–2 days of synthesis per 5 interviews minimum.
Paraphrasing in reports. Verbatim quotes with context beat paraphrase. Video clips beat quotes.
One-and-done research. Discovery is a program, not a project. Organizations that run discovery continuously outperform those that run it once per roadmap cycle.
AI IN QUALITATIVE RESEARCH — WHERE IT HELPS AND WHERE IT HURTS
Where AI helps:
- Transcription (replaces hours of manual work)
- First-pass theme extraction across large transcript sets
- Synthesis acceleration (grouping, tagging, highlighting)
- Generating interview-guide drafts and recruitment screeners
- Summarizing verbatims for executive audiences
Where AI hurts:
- Replacing live moderation (loses rapport, loses probing)
- Generating synthetic participants (never a substitute)
- Single-pass synthesis without human review (misses low-frequency, high-severity themes)
- PII handling without governance (compliance risk)
FAQ
Q: How many interviews are enough?
A: Plan for saturation, not a fixed count. Most projects reach saturation between 8 and 15 IDIs per segment. If themes keep emerging at 15, add a few more; if nothing new emerged at 8, stop.
Q: What's the difference between IDIs and focus groups?
A: IDIs are one-on-one; focus groups are 6–10 people. IDIs give depth and reduce groupthink; focus groups generate cross-participant dynamics and can cover more ground faster. Use IDIs for discovery; use focus groups when you need to observe group reaction or surface cultural norms.
Q: Can we run our own discovery interviews without a research partner?
A: Yes, with training. Many strong product teams run their own discovery. A research partner adds value for specialized methods (ethnography, diary), hard-to-reach samples, and synthesis rigour.
Q: How much should discovery research cost?
A: A 10-IDI discovery study typically runs $25K–$60K end-to-end. Ethnography $80K–$250K depending on site count. Diary studies $40K–$120K. Strategic programs combining all three $100K–$300K.
Q: When should discovery research happen?
A: Before a product bet is committed. If the team is already building, evaluative research is the right mode. If the team is choosing between problem areas, discovery is the right mode.
Q: How do you pick participants?
A: Screener-based recruitment with quotas matching the research question. For B2B, panel databases, LinkedIn outreach, and snowballing through networks. For consumer, panel recruiters or programmatic panels.
Q: Do participants lie in qualitative research?
A: They rarely lie outright; they rationalize, mis-remember, and self-present. Triangulate self-report with observation where stakes are high.
Q: Does NUUN run qualitative research for clients?
A: Yes. We staff senior moderators (not juniors running behind one-way mirrors) and ship synthesis that is actionable, not prose-heavy. Published methodology notes on every study.
RELATED READING
- UX research protocol
- Voice of Customer program playbook
- Best market research firms Canada
- Qualitative research (glossary)