Deep interviews: guide, technique, rookie mistakes

A deep (semi-structured) interview is a generative method: you use it when you need to understand why people behave the way they do, before you have hypotheses worth testing. It sits in the qualitative-attitudinal quadrant of the methods map — it tells you what people say and think, and must be triangulated with behavioral data.

The guide is a map, not a script

Structure the guide in funnels: broad context first (“walk me through a typical week of…”), then narrower episodes (“tell me about the last time that failed”), then details. Follow the participant when they open an unexpected door — the guide exists so you can return, not so you never leave.

A workable one-hour structure: 5 minutes of framing and consent, 10 minutes of context, 30 minutes on 2–3 specific recent episodes, 10 minutes on artifacts (screens, receipts, workarounds — “can you show me?”), 5 minutes of wrap-up (“what should I have asked about?”).

Technique

• Ask about the last specific time, not about “usually”. “Usually” produces averaged fiction.
• One question at a time. Double-barreled questions get half-answers.
• Follow up on emotion words. “That was annoying” → “what made it annoying that day?”
• Count to five after their answer ends. The best material arrives after the pause you almost interrupted.

Rookie mistakes (all fixable)

• Leading questions. “Was that frustrating?” plants the answer. Ask “How did that feel?”
• Hypotheticals. “Would you use X?” produces polite fiction. Ask what they did.
• Pitching. The moment you defend your product, the interview is over.
• Interviewing only fans — recruit for recent relevant behavior, not for enthusiasm.

Try it: rewrite these into non-leading, past-tense form — “Do you like our onboarding?”, “Would a dark mode help you?”, “Isn’t the search hard to find?”