A structured conversation that uncovers what users actually do and feel — without pitching your product or fishing for the answer you want.
▸ Try the interactive toolA user interview is a structured one-on-one conversation designed to uncover real behaviour, motivation, and frustration — without pitching your product or hunting for confirmation. It's the foundational qualitative discovery tool: almost every other discovery artifact is built from interview data.
The core skill is restraint. A bad interview is the PM talking, leading, and hearing what they hoped to hear. A good one is the user doing 80% of the talking about specific past behaviour, while the interviewer resists the urge to pitch, defend, or suggest. What people did is data; what they say they'd do is speculation.
Ask about the last time something happened, not what someone “usually” does or “would” do. Past behaviour is evidence; hypothetical and habitual answers are flattering fiction.
The information that changes a roadmap rarely comes from analytics — it comes from understanding why behind the numbers. But interviews are deceptively easy to do badly.
| The shortcut | What it costs | What it gives you instead |
|---|---|---|
| Building on assumptions | Without talking to users, the team designs for an imagined person. | Interviews replace the imagined user with a real one, in their own words. |
| Leading the witness | Asking “wouldn't it be great if…?” gets a yes that means nothing. | Behaviour-focused questions remove the PM's thumb from the scale. |
| Confusing opinions with needs | “I'd pay for that” is the least reliable sentence in product. | Past-behaviour questions surface what people actually do, not what flatters the interviewer. |
| One-and-done research | A single chat with a friendly user feels like validation. | A series of interviews reveals patterns one conversation can't. |
Recruit people who match the segment you're designing for — not whoever is easy to reach. Prepare open questions grouped by theme, but treat the guide as a map, not a script.
Start easy, then anchor on real episodes: “Walk me through the last time you…” Specific stories beat general opinions every time.
Follow interesting answers with “why?” and then wait. Silence is the most productive interview technique — people fill it with the real answer.
The moment you describe your solution, the interview is over — you'll get politeness, not truth. If they criticise something, dig in; don't defend.
One interview is an anecdote. Run enough (commonly 5–10 per segment) to hear the same thing repeatedly, then synthesise the patterns — that's the signal.
A PM convinced their onboarding was fine ran five interviews and asked each user to “walk me through the first day you signed up.” Four of five described the same dead end — a moment they didn't know what to do next — and quietly gave up.
None of them had complained; they'd just churned silently. Analytics showed the drop-off but not the cause. The interviews named it in the users' own words within an hour.
Two artifacts: a discussion guide (used loosely) going in, and a synthesis of patterns coming out.
“Walk me through the last time you…” · “What were you trying to accomplish when…” · “What did you do right before / after…” · “What was frustrating about that?” · “What would you have done if this didn't exist?” — then stay quiet.
| Avoid | Why it fails | Ask instead |
|---|---|---|
| “Would you use a feature that…?” | Hypothetical — everyone says yes | “Tell me about the last time you needed to…” |
| “Don't you find X frustrating?” | Leading — plants the answer | “What parts of this are frustrating, if any?” |
| “How often do you usually…?” | Habitual — invites a guess | “When did you last do this? Walk me through it.” |
What people did is data. What they say they'd do is speculation dressed as data — and acting on it is how teams build features that test well and ship to silence.
The interviewer's discipline matters more than the question list. The instinct to pitch, lead, and defend is strong, and every time you give in you trade truth for politeness. The best interviewers feel almost passive in the moment — and walk away with the insight that reshapes the roadmap.
The instant you describe your solution, you get feedback on your pitch, not insight into their life. Hold the solution back entirely.
“Wouldn't it be helpful if…?” guarantees a useless yes. Ask neutral, behaviour-focused questions.
If you're talking more than 20% of the time, you're interviewing yourself. Embrace silence.
A single friendly conversation isn't validation. Patterns only emerge across a series.
Interview transcripts are the raw material for JTBD, the VPC customer profile, problem statements, and affinity maps.
Patterns from interviews become survey questions that measure how widely the pattern holds.
Interviews are the core activity of the Discovery track — the evidence that clears the build gate.
Advanced interview technique, contextual inquiry, and synthesis methods are covered in depth in the discovery module.
AI helps before and after the interview — but it cannot replace the human conversation in the middle.
The judgment that stays yours: The value of an interview is the unscripted human moment — the surprise, the contradiction, the thing you didn't think to ask. AI can sharpen your prep and speed your synthesis, but it can't be in the room. And never let it invent quotes.
Find someone and ask them to walk you through a recent purchase or tool choice — something specific and real. Your only job: ask “why?” and then stay silent. No pitching, no leading.
Count how often you're tempted to jump in, suggest, or relate it to your own experience. Resist every time.
Notice how much more you learn in the silences than in your questions. That restraint is the entire skill.
Ready to go deeper · Advanced User Interviews →