UX Design Fundamentals · 18 min · 120 XP

Interviews that reveal something

Ask about what people actually did, not what they say they would do, and know when you have heard enough.

Almost every bad research interview fails the same way: it asks about the future. “Would you use this?”, “How much would you pay?”, “Would a dark mode be useful?” People are generous, they want to be helpful, and they are genuinely bad at predicting their own behaviour. You will come back with a notebook full of encouragement and no information.

The fix is to ask about the past, because the past already happened and people can remember it. “Walk me through the last time you had to do that.” “What did you do right before?” “What did you try first?” “What did it cost you when it went wrong?” Every one of those has a real answer, and the answer is evidence rather than a prediction.

The same curiosity, asked two ways
Future intent (worthless)     Past behaviour (evidence)
──────────────────────────    ────────────────────────────────
"Would you use a shared      "Tell me about the last time two
 inbox?"                      of you answered the same email."

"Is invoicing painful?"      "Walk me through sending your
                              last invoice. What took longest?"

"Would you pay £20/month?"   "What do you spend on this today,
                              and what did you buy last?"

Compliments are not data. “That's a great idea” and “I'd definitely use that” feel like progress and mean nothing. Rob Fitzpatrick's The Mom Test is the short book on why. The useful currency is specifics people give up without being prompted: what they've already tried, what they've already paid for, who else they had to ask, what they do in the meantime.

Talk to people who have the problem, not people who are easy to reach. Your colleagues, your Twitter followers and your most enthusiastic customer are all convenient and all wrong in the same direction. A two-question screener (have you done this thing, and when was the last time?) is usually enough to keep the sample honest.

When to stop: when new interviews stop surprising you. That point is called saturation, and in one reasonably narrow group it usually arrives somewhere around the fifth to eighth conversation. It's a signal to watch for, not a quota to hit. A second, different group of users resets the clock.

Do it this week: find one person who has the problem you're working on and ask them to walk you through the last time it happened. Ask no question about the future for fifteen minutes. Count how many times you nearly did.

Loading your workspace…