How to Run Customer Discovery Interviews

Most customer interviews are useless. Founders ask leading questions, hear what they want to hear, and walk away convinced their idea is brilliant. Then they build something nobody buys.

Good interviews are different. They reveal truth, even uncomfortable truth. They expose assumptions you didn't know you had. They tell you what customers actually do, not what they say they'll do.

This guide covers how to find the right people to talk to, what to ask, and how to turn conversations into actionable insights.

The fundamental problem: people are polite. When you describe your idea with enthusiasm, most people won't tell you it's bad. They'll nod, say "that sounds interesting," and never think about it again.

Common mistakes that produce garbage data:

Leading questions: "Don't you think it would be great if there was a tool that did X?" Of course they agree. You just told them what to think.

Pitching instead of learning: If you spend more than 20% of the interview explaining your solution, you're doing it wrong.

Asking about the future: "Would you use this?" and "Would you pay for this?" are nearly worthless. People are terrible at predicting their own behavior.

Talking to the wrong people: If your customer is restaurant owners and you're interviewing your friends who are software engineers, the data doesn't transfer.

Confirmation bias: Interpreting ambiguous responses as validation. "They said 'maybe'" becomes "They're interested!" in your head.

The solution is a framework called the Mom Test, from Rob Fitzpatrick's book of the same name. The idea: ask questions so good that even your mom couldn't lie to you.

The core principle: talk about their life, not your idea. Ask about problems they've experienced, solutions they've tried, and money they've spent. Don't ask if your solution is good.

Rules of the Mom Test

1. Talk about their life, not your idea.

Bad: "I'm building an app that helps restaurants manage inventory. What do you think?"

Good: "Walk me through how you handle inventory at the end of each week. What's that process like?"

2. Ask about specifics in the past, not hypotheticals about the future.

Bad: "Would you use a tool that sends automatic reorder alerts?"

Good: "Tell me about the last time you ran out of something critical. What happened?"

3. Talk less and listen more.

Your job is to extract information, not to convince. If you're talking more than 30% of the time, you're doing it wrong.

4. Seek out the negatives.

Actively look for reasons your idea might fail. "What would make this not work for you?" "Why haven't you solved this already?"

5. Get commitments, not compliments.

"This sounds great" means nothing. "I'd pay $50/month for that" means slightly more. "Here's my credit card" is real validation.

You need to talk to people who represent your actual target customer. Not friends. Not family (usually). Not random people willing to chat.

Where to Find Interview Subjects

Your network (carefully): Friends of friends who fit your customer profile. Ask for introductions. "Do you know any restaurant owners who might spend 20 minutes talking about their operations?"

LinkedIn: Search by job title and industry. Send a polite message: "I'm researching [problem area] and would love to learn from your experience. Would you have 15-20 minutes for a quick call?"

Industry communities: Reddit, Discord, Slack groups, Facebook groups. Participate first, then ask for conversations.

Twitter/X: People often tweet about their problems. Reach out to those complaining about what you're solving.

Events and meetups: Industry conferences, local business meetups, co-working spaces. Face-to-face requests convert better.

Cold email: Works for B2B. Find email addresses through Hunter.io or Apollo. Keep the ask short and specific.

How Many Interviews Do You Need?

Minimum: 15-20 conversations before drawing conclusions. Fewer than that and you're probably seeing noise, not patterns.

You'll know you've done enough when:

• The same answers keep coming up
• New interviews stop revealing new information
• You can predict what someone will say about a topic

Opening (2-3 minutes)

Set context without biasing the conversation:

"Thanks for making time. I'm trying to understand how [people like you] handle [general problem area]. I'm not trying to sell you anything. I just want to learn from your experience. There are no wrong answers."

Don't reveal your solution yet. The more they know about your idea, the more biased their answers become.

Main Questions (15-20 minutes)

About their world:

• "Walk me through a typical day/week when you're dealing with [problem area]."
• "What's the hardest part about [relevant task]?"
• "How do you currently handle [specific challenge]?"

About past behavior:

• "Tell me about the last time [problem] happened."
• "What did you do about it?"
• "How did that work out?"

About existing solutions:

• "What tools or processes do you use for this now?"
• "What do you like about them?"
• "What's frustrating about them?"

About importance:

• "How much time does this take you each week?"
• "What does this problem cost you? In money, stress, or lost opportunity?"
• "Where does this rank among your challenges?"

About motivation:

• "Have you looked for a solution to this?"
• "Why or why not?"
• "What would make you switch from what you're doing now?"

Closing (2-3 minutes)

• "Is there anything else I should know about this that I didn't ask?"
• "Who else should I talk to about this?"
• "Would it be okay if I follow up when we have something to show?"

Get referrals. Every good interview should lead to at least one more.

Strong Positive Signals

• They describe the problem with emotion (frustration, anger, desperation)
• They've already tried to solve it (searched for solutions, tried workarounds, spent money)
• They can quantify the impact (hours wasted, dollars lost, customers churned)
• They ask when they can use your solution
• They offer to pay before you offer to sell

Weak or Negative Signals

• "That's interesting" or "That sounds cool" (polite dismissal)
• They can't remember a specific time this happened
• Their workaround seems "good enough"
• The problem is annoying but not worth paying to fix
• They change the subject quickly

Red Flags About Your Idea

• Nobody has the problem you're solving
• People have the problem but it's low priority
• Good solutions already exist and people are happy with them
• The problem is acute but rare (not worth building a business around)
• People want a solution but won't pay for it

Listen for these even when you don't want to hear them.

Interviews tell you what to build, not just whether to build.

After your first batch of interviews:

Validate or invalidate your core hypothesis. Is the problem real? Is it painful enough to pay for? Does your general approach make sense?

Identify the sharpest version of the problem. Maybe the broad problem exists, but one specific sub-problem is much more painful. Focus there.

Find your early adopters. Which segment of interviewees was most enthusiastic? Those are your first customers.

Extract positioning language. The words customers use to describe their problem become your marketing copy.

Prioritize features. What's the minimum that would make someone switch from their current solution?

Don't wait until you have perfect certainty. You never will. Decide when you have enough confidence to build something small and test it with real users.

• Most interviews fail because founders ask leading questions and hear what they want to hear.
• The Mom Test: ask about their life and past behavior, not about your idea and hypotheticals.
• Talk to 15-20 people minimum before drawing conclusions. Look for patterns, not individual opinions.
• Strong signals: emotion about the problem, money already spent on solutions, quantified impact.
• Weak signals: "sounds interesting," can't remember specifics, problem is low priority.
• Record insights systematically. Quotes and patterns become product decisions and marketing copy.

How long should interviews be?

20-30 minutes is ideal. Enough time to go deep, short enough that people will agree to it. You can go longer if the conversation is productive, but don't plan for more than 30.

Should I pay people to interview?

For consumer products, sometimes a small incentive ($20-$50 gift card) helps with recruitment. For B2B, payment can actually reduce response rates because it feels transactional. Offer value instead: share insights, provide early access, etc.

What if I can't find people to interview?

Lower your bar for initial outreach. Cold LinkedIn messages, Reddit posts asking for help, Twitter DMs. Getting first interviews is hard. Once you have a few, ask for referrals and it snowballs.

When should I reveal my idea?

At the end of the interview, if at all. You want their unbiased perspective on the problem first. If you're testing a specific solution, you can describe it late in the conversation and gauge reaction.

How do I avoid confirmation bias?

Actively seek disconfirming evidence. Ask "What would make this not work for you?" Assign someone else on your team to play devil's advocate. Reread interviews looking for negative signals you might have dismissed.