Rob Fitzpatrick's "The Mom Test" (2013) is still the best book on customer interviews. The main rule: if questions are such that even your mom would answer honestly (not just nicely), you're on the right track.
Most interviews fail because they ask about the future (hypothetical) instead of the past (factual). "Would you pay $20 for this?" → people lie. "What did you pay last time for something like this?" → fact.
Three Mom Test rules
- Talk about their lives, not your idea. The idea comes last, if ever.
- Ask about the past, not the future. Hypothetical answers are worthless.
- Less talking, more listening. You need to learn from them, not sell to them.
Good vs. bad questions
Bad: "Would this be useful to you?" Good: "When did you last try to solve this problem? How did it go?"
Bad: "Would you pay $20/month for this?" Good: "What similar product are you currently paying for? How much?"
Bad: "Do you like my idea?" Good: "When someone last pitched an idea like this to you, what did you think?"
AI's role in interview guides
1. Generation
Describe to AI:
- The problem you suspect the customer has
- The target segment
- What you want to confirm or refute
Ask: "Give me 10 Mom Test-compliant interview questions. Each must be about the past, not the future. Include 2 wilder questions that might reveal an unexpected angle."
2. Validation
You've written your own questions. Check:
"Here are my interview questions. Which are leading questions? Which ask about the future hypothetically? Which are too general to get a useful answer?"
AI quickly shows you the holes.
3. Scenario practice
"Pretend you're a 35-year-old consultant considering starting your own business. I'll ask you questions, you respond realistically."
You use AI to practice before real interviews. You don't learn about the customer, but you learn about your questions — do they sound right, do they leave unclear gaps?
4. Transcript analysis
You've done 5 interviews — 2 hours of audio. AI transcribes (Whisper) and you ask:
"What themes recurred across interviews? What surprised? What didn't I ask but came up spontaneously?"
A practical interview script
A good starting point for a 30-minute call/zoom:
- Context (5 min): "Describe your day-to-day and current role."
- Problem existence (5 min): "When did you last try to do [problem-area task]? Walk me through it."
- Current solution (5 min): "What tools/methods do you use now? What works, what doesn't?"
- Trigger (3 min): "What triggered the need last time?"
- Price context (3 min): "Are you paying for other tools in this area? How much?"
- Open (5 min): "What would be most useful to talk about that we haven't covered?"
- Referrals (2 min): "Who would be the next 1–2 people in your circle this kind of conversation would be valuable with?"
Pitfalls
1. Selling instead of listening. The biggest mistake. You need to be confident your solution is good, while open to hearing it isn't. A good interviewer doesn't pitch.
2. No follow-ups. "Why?" and "What do you mean?" and "Can you give an example?" — these three open 80% of an interview. Don't move to the next question too fast.
3. Ending with "Would you buy?". This destroys the interview. You get false "yes" answers, not facts.
Innovaidor and Interview Guide
Innovaidor's Interview Guide is a craft skill the AI can activate automatically when the conversation moves to interview prep. E.g., you're sparring a concept on BMC, and AI suggests: "Validating this would now need 5 interviews. Switch to Interview Guide to build the questions?"
One click → AI generates Mom Test-compliant questions tailored to your target segment.
Closing
Interviews aren't hard when the questions are right. AI accelerates generation and validation — but the conversation itself is still your work. A good interviewer listens more than talks, asks about past not future, and welcomes criticism as much as praise.
Start: write ten questions for your next interview, check them with AI against Mom Test principles, and do one real interview this week.