How to conduct better customer interviews
Practical tips for running customer interviews that surface real insights instead of polite agreement.
Most customer interviews fail before they start — not because the questions are bad, but because the approach is wrong. Here's how to run interviews that actually surface useful insights.
Start with a clear research goal
"We want to understand our users better" is a wish, not a goal. A good research goal is specific and points at an action: "Why don't trial users convert to paid plans?", "Why do mid-market companies churn within 90 days?", "What are the top three pain points in onboarding?"
A useful test: if your goal doesn't suggest a specific action you'd take with the results, it's too vague. It also determines your questions, your participants, and how you analyze — get it right first.
Ask open-ended questions about the past
The biggest mistake is asking leading or closed questions. Instead of "Do you find our product easy to use?" try "Walk me through the last time you used our product." Open-ended questions let people tell their story; closed questions just confirm your assumptions. Start prompts with "Tell me about...", "Walk me through...", or "Describe..." to invite narrative, where the richest insights live.
Just as important: don't ask about the future. People are terrible at predicting their own behavior, so "Would you use this?" and "How much would you pay?" produce unreliable answers. Ask about past behavior instead — "Tell me about the last time you had this problem." If someone says "I'd definitely use that," it means almost nothing; "Last week I spent three hours doing exactly that in a spreadsheet" is a strong signal. The same applies to pricing: don't ask what they'd pay, ask what they pay for their current solution.
Follow up on what matters
The best insights come from follow-ups, not your script. When someone says something surprising or emotional, dig deeper — "Tell me more about that," "What do you mean by [their word]?", "Why was that important to you?" Watch for signals: hesitation, emotion, unexpected vocabulary, and contradictions between what people say and what they did. This is where AI-powered interviews excel — a well-designed AI interviewer detects these signals and probes consistently, something even experienced humans miss when focused on their script. And be comfortable with silence: people fill pauses with their best insights.
Interview the right people
Five interviews with the right participants beat fifty with the wrong ones. For churn, talk to people who recently cancelled, not loyal customers — churned customers tell you what broke. For acquisition, talk to recent sign-ups while the decision is fresh. For usability, talk to people matching your target profile, not power users who've adapted to your quirks. For pricing, talk to people who evaluated but didn't buy. The wrong participants don't just give you less useful data — they give you misleading data.
Keep it conversational
An interview should feel like a conversation, not an interrogation — the respondent should do 80% of the talking. Acknowledge answers ("That makes sense") without leading, and reference what they said earlier ("You mentioned X — tell me more"). Don't correct or educate: if they misunderstand your product, their misunderstanding is the insight. And mirror their language — if they say "dashboard," don't say "reporting module."
Watch especially for the gap between words and behavior. A customer might say "I love your product" while logging in monthly. When you spot a gap, explore it gently: "You said you value X, but it sounds like you rarely use it — help me understand that?"
Synthesize as you go
Don't wait until the end to analyze. Within 30 minutes of each interview — memory degrades fast — write down the most surprising things you heard, emerging patterns, and quotes in the respondent's own words. You don't need 50 interviews: most studies reach saturation, where new interviews stop revealing new insights, between 5 and 12 participants. If your last three just repeated earlier themes, you have enough.
The goal is understanding, not validation. If every respondent confirmed your hypothesis, something went wrong. Good research challenges your assumptions — the most valuable interview is the one that makes you rethink your roadmap, not the one that confirms it.