"Would you use a tool that did X?" produces lies. Not because users are mean — because they want to be helpful and they're guessing about a future they haven't lived in. The Rob Fitzpatrick rule still holds in 2026: ask about the past, not the future. Ask about behavior, not opinions.
Three questions to never ask
- "Would you pay for X?"
- "Do you think this is a good idea?"
- "Would you use this if it existed?"
All three produce flattering answers and zero signal.
Three questions that work every time
- "Walk me through the last time you did [workflow]."
- "What did you try before? Why'd you stop?"
- "What did you do this morning between waking up and lunch?" (only if relevant)
All three are about real, recent behavior. Behavior doesn't lie.
The "money signal" question
"Have you ever paid for anything to solve this?" is the single most predictive question in user research. A "yes" with a dollar amount is gold. A "no, but I would" is noise.
The 90/10 rule
Talk 10% of the time, listen 90%. Most founders flip this and walk away with no signal, just an echo of their own pitch. If you catch yourself explaining the product, stop and ask another question.
Take notes verbatim
Record the call (with permission) or write down their exact words. Their phrasing IS your future landing page copy. Paraphrasing destroys this.
The graduation move
After 10 interviews, look for sentences that 5+ people said almost verbatim. Those are positioning gold. Then run a real demand test (see 48-hour validation experiments).