How do I run a customer discovery interview?

Ask about past behavior, not future intentions. 'Would you use this?' is a bad question — people lie to be polite. 'Walk me through the last time you had this problem' is a good question — behavior reveals truth. Keep interviews under 45 minutes, record them with permission, and look for emotional language (frustration, urgency, embarrassment) as the real signal.

What questions should I ask in a customer discovery interview?

The five most useful questions: (1) Walk me through the last time you had this problem. (2) What did you do about it? (3) What do you wish existed? (4) How much does this problem cost you in time or money? (5) Who else feels this pain in your organization? Avoid leading questions and hypotheticals.

How many customer discovery interviews do I need?

You need enough to reach saturation — the point where new interviews stop surfacing new patterns. For most B2B SaaS products, this is 12–20 interviews across your primary candidate segments. If you're hearing the same language and pain points repeated by interview 15, you have enough signal to move forward.

How to Run Customer Discovery Interviews That Give You Real Signal

Customer discovery interviews tell you what you want to hear unless you design them not to. Most founders come out of 10 interviews more confident in their original hypothesis than when they started. Research by Cindy Alvarez shows 85% of discovery interviews are invalidated not by the answers but by the questions — leading questions that make agreement the path of least resistance for a polite stranger.

Why this happens

Founders are optimists. That’s a useful quality for building companies and a harmful one for running interviews. When you believe in your product, you phrase questions so agreement feels natural. “Would this save you time?” almost always gets a yes. “Tell me about the last time you solved this problem and what you paid for it” gets you a story, which is what you actually need.

The other failure mode is asking about the future. Humans are notoriously bad at predicting their own behavior. “Would you pay for this?” tells you almost nothing. What someone did last month — the workaround they built, the tool they paid for, the consultant they hired — tells you whether the problem is real and what they already believe it’s worth.

What to check first

Before your next round of interviews, audit the questions you’re planning to ask:

Are any questions phrased as hypotheticals (“would you,” “could you imagine,” “if we built”)?
Are any questions asking them to evaluate your idea rather than describe their experience?
Do your questions allow a one-word answer — “yes,” “no,” “maybe”?
Are you planning to demo before asking about the problem?

If you answered yes to any of these, rewrite those questions before the interview starts.

How to fix it

The Mom Test rule is the simplest framework: only ask questions you couldn’t answer by talking to your mom. Your mom loves you and will validate anything you’re excited about. A stranger will too, unless you make it impossible.

Three rules that hold across every interview:

Only ask about the past, not the future.
Only ask about their life, not your product.
Listen for emotional language — frustration, urgency, embarrassment. Calm answers mean low stakes. Emotional answers mean you’re in the right territory.

The five questions worth asking in every session: Walk me through the last time you had this problem. What did you do about it? What do you wish existed? How much does this cost you in time or money? Who else feels this pain on your team?

Twelve to 20 interviews is usually enough to reach saturation — the point where you’re hearing the same language and the same pain points on repeat. At saturation, you have enough signal to move.

Remove the guesswork

Even well-run interviews have a ceiling: you can talk to 20 people, but you can’t talk to 200 in a week. RightAudience runs 100+ synthetic buyer simulations across your candidate segments, scoring each on purchase intent, willingness to pay, and conversion likelihood. It surfaces the patterns that would take months of manual interviews to find — and it removes the confirmation bias that comes with founders running their own discovery.

See how RightAudience works