Design a customer interview guide for problem discovery

Behavioral Interviews Beat Opinion Interviews

The most common research failure is asking customers what they think instead of what they did. Teresa Torres' continuous discovery writing and the Nielsen Norman Group's research methods taxonomy both emphasize behavioral anchoring: questions tied to a specific past event yield reliable data, while opinion questions ("would you use this?") produce high false-positive signal. The switch interview tradition from JTBD.info gives the best-evidence framing — walk me through the last time.

How the Design a customer interview guide for problem discovery Prompt Works

The prompt structures the guide into five sections with explicit behavioral anchoring in the problem recall section, forces every major question to reference a specific past event, and flags the three questions most likely to produce fake data so the researcher can rewrite them. The "what almost stopped you" and "what else did you try" questions surface forces and alternatives — the highest-signal data in problem discovery.

When to Use It

• You are running first-round discovery for a new feature bet.
• Past interviews produced useless "yes" signal.
• A new PM is starting their first research project.
• You need to validate a problem assumption before committing eng time.
• The team is debating a feature without evidence from customers.

Common Pitfalls

• Asking "would you use this". Answers are 90% yes regardless of intent to buy. Anchor on past behavior instead.
• Leading with your solution. Any mention of your proposed solution contaminates the data. Keep the guide solution-agnostic.
• Skipping the "almost stopped" question. Friction points are the richest product input. Ask what made them almost give up.

Sources

• Continuous Discovery Habits — Product Talk (Teresa Torres)
• Opportunity Solution Tree — Product Talk (Teresa Torres)
• What is Jobs-to-be-Done? — JTBD.info
• Which UX Research Methods — Nielsen Norman Group