Most bad interviews are lost before anyone presses record. The user interview questions you walk in with determine what you can walk out with, and questions written in a hurry produce the classic haul: polite opinions, confident predictions, and nothing you would bet a roadmap on.
A discussion guide is where those questions live, but the point of a guide is widely misunderstood. It is not a script to be read aloud in order. It is a safety net: preparation thorough enough that you can put it down, follow the human in front of you, and still cover what the study needs. Writing user interview questions well is mostly about building that net.
Start with research questions, then write interview questions
These are two different things, and conflating them ruins guides. A research question is what the team needs to learn: "What stops experienced users adopting the new workflow?" An interview question is what you actually say to a person, and it is never the research question read aloud. Ask someone "What stops you adopting the new workflow?" and you will get a theory, politely invented on the spot.
Write the research questions first, three to five for a typical study. Then, for each one, write interview questions that gather the raw material to answer it: stories, walk-throughs, specifics. The analysis is your job later, not the participant's job live. Keep the research questions pinned at the top of the guide, where a mid-session glance can remind you what all of this is for.
Structure the guide as a funnel
The reliable shape moves from broad to narrow:
- Warm-up. Easy, factual questions about the person and their context. You are building comfort, and their answers calibrate everything that follows.
- The grand tour. A wide-open invitation, borrowed from the ethnographer James Spradley: "Walk me through how this works for you." Let people map their own world before you steer.
- Focused topics. The heart of the session, organised around your research questions.
- Specific or sensitive material. Money, frustrations, comparisons with rivals: these land better once trust is built.
- Wrap. "What should I have asked about?" often outperforms everything above it. Then thanks, and what happens next.
Time-box each section and be honest about the arithmetic: a 45 to 60 minute session holds roughly eight to twelve core questions once you leave room for follow-ups. If your guide has thirty questions, you have written a survey with a person attached.
It also helps to mark every question as must-ask or nice-to-ask. Sessions run over, the good ones especially, and you want to make the cut deliberately rather than discover at minute fifty that your most important question is still waiting.
User interview questions that open people up
A few principles do most of the work:
- Past behaviour beats future intention. "Tell me about the last time you..." beats "Would you use...?" every time. People are honest historians and hopeless forecasters.
- Open beats closed, mostly. Closed questions ("Do you use the app daily?") are fine for facts. Stories need open ones ("How did that unfold?").
- Watch for smuggled assumptions. "How much easier was the new version?" assumes it was easier. Neutral framing asks: "How did the new version compare?"
- Go easy on "why". Fired repeatedly, it can feel like cross-examination, and it invites rationalisation. "What led you to...?" and "How did you decide...?" reach the same place more gently.
- One question at a time. Double-barrelled questions get half-answers to whichever half was easier.
Probes: where the insight actually lives
Scripted questions open doors; probes walk through them. The moves are simple. "Tell me more about that." "What happened next?" Echo their own words back: "It was fiddly. Fiddly how?" And the most underused probe of all is silence: count to five before rescuing an unfinished thought, because the second half of an answer is usually the honest half.
Write two or three probes under each key question in the guide itself. Under pressure you will not invent good ones live, and pre-written probes are what let a nervous interviewer sound calm. One more habit worth stealing: note the participant's own vocabulary early and reuse it. People open up faster when interviewed in their language rather than yours.
Pilot it before it counts
Your first session is a pilot whether you plan it or not, so plan it. Run the guide once with a colleague or a low-stakes participant and check three things: timing (guides always run long), comprehension (which questions earned blank looks), and leading (which questions came back to you in your own words). Trim, reorder, soften, and keep adjusting through the early real sessions; in qualitative work, refining the instrument as you learn is a feature, not a flaw.
The guide also has to fit its setting. Interviewing someone at their desk mid-task, as in contextual inquiry, needs looser, observation-led questions than a scheduled remote call. And if you are wondering how many of these sessions to run, that is a sample size question with a comforting answer: fewer than you fear, once the guide is doing its job.
A great discussion guide disappears in the room. You wrote it, piloted it, and pinned probes to every key question, and precisely because of that you can spend the session listening instead of reading. That is the paradox worth embracing: prepare the script so thoroughly that you earn the right to abandon it.
People of Research is a community and newsletter for UX researchers, market researchers, and anyone who learns for a living by asking good questions. Join us and get the good stuff in your inbox.