30 customer discovery interview questions that actually test something

Most customer discovery interviews fail before they start, because the questions are written to confirm the idea instead of testing it. Below is a working set of customer discovery interview questions, grouped by what each one measures, so you can build a script that gets you evidence and not just encouragement.

Use 8 to 12 of these per interview. Any more and you're rushing; any fewer and you can't cross-check answers.

Before the questions: one screener

Ask one qualifying question at the top, and end the interview politely if the answer rules them out. Something like: "In the last month, how many times did you [do the core activity]?" If the honest answer is "never," they're not your segment, and their answers will only add noise. A good screener is worth more than ten extra interviews; see how many customer interviews you need.

Opening: establish context (3 to 4 questions)

• "Walk me through how you currently handle [the job or process]."
• "Which tools or people are involved in that today?"
• "How did you end up doing it that way?"
• "When did you last do this? Tell me about that specific time."

You want a concrete story here. "Tell me about the last time" gets you one far more often than "how do you usually" does.

Problem: is it real and painful? (6 to 7 questions)

• "What's the most frustrating part of that process?"
• "What happens when it goes wrong?"
• "How often does that happen?"
• "What did that cost you the last time, in time, money, or something else?"
• "Who else is affected when this breaks?"
• "Have you tried to fix it? What did you try?"
• "What made you stop trying, or keep going with the current approach?"

If the answers here are mild ("it's a minor annoyance"), write that down instead of steering around it. When to kill a startup idea covers what a pattern of mild answers means.

Workaround: what are they already doing about it? (4 to 5 questions)

• "What's your current workaround, even if it's a spreadsheet or a manual step?"
• "How long did it take to set that up?"
• "What do you like about it? What breaks?"
• "Have you paid for anything to make this easier? What?"
• "If your workaround disappeared tomorrow, what would you do?"

Money or ongoing effort already spent on a workaround is one of the strongest problem signals you can find. If nobody has one, pay attention to that too.

Willingness to pay: is there budget and intent? (4 to 5 questions)

• "What do you spend today on this, all in?"
• "Who would sign off on paying for a better solution, you or someone else?"
• "What would a solution need to do for that spend to be an easy yes?"
• "When you bought [an adjacent tool], how did that decision get made?"

Ask about purchases they've already made. Hypothetical future purchases tell you very little. There are more willingness-to-pay techniques in willingness to pay survey questions.

Behavior and adoption: would they actually switch? (3 to 4 questions)

• "What would have to be true for you to change how you do this?"
• "What's stopped you from switching before?"
• "Who would need to be on board?"
• "If I gave you a solution today, what's the first thing you'd want to check?"

Close: open the floor (2 questions)

• "What haven't I asked about that I should have?"
• "Who else deals with this that I should talk to?"

That second question turns one research round into a list of people for the next one.

Questions to never ask

• "Would you use this?" Saying yes costs them nothing.
• "Do you think this is a good idea?" This invites politeness.
• "How much would you pay for [feature]?" Hypothetical pricing is noise.
• "Don't you hate when X happens?" This leads the witness.
• Anything that describes your solution before you've established the problem.

Common questions

How long should a customer discovery interview be?

25 to 40 minutes. That's long enough for real stories and short enough that people accept the invite and stay engaged.

Should I record customer discovery interviews?

Yes, with permission. You'll miss things live, and exact quotes matter when you score hypotheses later.

Can I turn these into a survey?

Most of them, yes. You lose the live follow-up, so build the likely follow-ups into the branching or use a tool that generates them. See idea validation survey questions.

How many interviews before I see a pattern?

Usually 8 to 15 well-screened conversations. If everyone says something different, your segment is probably too broad.

Skip writing the script

Hypothis generates a screened, Mom-Test-disciplined question set from your hypotheses, runs adaptive follow-ups on every response, and scores the answers against what you set out to test. It's free during early access. See how it works or create your account.