Customer conversations fail in two predictable ways. Either you ask people to predict their future behaviour, which they are bad at, or you ask in a way that invites them to be nice to you, which they will be.
People are bad futurists, and kind
"Would you use this?" and "would you pay for that?" produce enthusiastic answers that predict almost nothing. The person answering is imagining a version of themselves with more time and discipline, and they are also trying not to hurt your feelings.
Ask about the past instead
The past is factual. What someone did last month is real data, and it is the best available predictor of what they will do next month.
- Not "would you pay for this?" but "what have you paid for to solve this?"
- Not "is this a problem?" but "when did it last happen, and what did you do?"
- Not "would you use it weekly?" but "walk me through the last time you needed it."
Compliments are the danger sign
If a conversation is full of "that's a great idea," you are being flattered rather than informed. A useful interview often feels slightly disappointing, because you learn the problem is smaller, weirder, or different from the one you built for.
The goal is not validation. It is to find out what is actually true while it is still cheap to change your mind.
Try it
Take the three questions you would naturally ask a prospect and rewrite each one to be about their past behaviour. Notice how much harder they are to answer with a polite yes.