The Customer Interview Guide: Asking Questions That Can't Be Lied To
The customer interview is the cheapest validation tool and the most misused. A badly run interview produces approval theater, not data: you pitch your idea, the other person politely says "nice, I'd use that," and you mistake it for evidence. A well-run interview gets the customer talking about their past behavior, not your idea.
The Core Rule: Don't Pitch Your Idea Ask About Their Life
People can't predict their future behavior, but they know what they did in the past. Hence the interview's golden rules:
| Don't ask (future/hypothetical) | Ask (past/behavior) |
|---|---|
| "Would you use an app like this?" | "When did you last hit this problem? What did you do?" |
| "How much would you pay for this?" | "What have you spent so far (money/time) to solve this?" |
| "Do you think it's a good idea?" | "How do you handle it today? What's the worst part of that method?" |
| "Would you want feature X?" | "Walk me through the last time you needed X step by step" |
The current workaround (Excel, WhatsApp, paper, a consultant, nothing at all) is the interview's most valuable discovery: it reveals both your real competitor and the customer's real payment bar. If no workaround exists, the problem is probably not as urgent as you think.
Finding Participants and Sampling
- Who: Exact matches to your target segment definition mixed profiles produce mixed signals. 10-15 interviews per segment is the typical saturation point where patterns start repeating
- Where: Wherever the segment gathers (communities, LinkedIn, associations, the outer ring of your network). Close friends and family are the worst sample, thanks to politeness bias
- The invitation: Not "evaluate my idea" but "I want to learn about your X process I won't sell you anything." A learning invitation gets accepted at a multiple of a sales invitation
The Interview Flow: A 30-Minute Structure
- Warm-up (3 min): Role and context "how does the X process work in your job?"
- Problem discovery (15 min): The last concrete incident: "When did it last happen? Then what? Who was involved? How long did it take?" dig the story chronologically
- Current solution and cost (7 min): "How do you solve it now? Why that method? What did you try and abandon?"
- Severity signal (3 min): "When did this problem last keep you up at night / claim a budget line / force a hire?" a problem with resources allocated to it is a real problem
- Close (2 min): "Who else should I talk to about this?" snowball sampling + a signal about the segment's network structure
Record (with permission), give note-taking to someone else or write up afterward taking notes during the interview kills the listening.
Turning Findings into Decisions
Don't let interview data remain an anecdote pile: within 10 minutes of each interview, produce a standard summary (problem severity 1-5, current workaround, money/time spent, verbatim quotes). After 10-15 interviews, build the pattern table: the share ranking the problem in their top 3, the shared workarounds, the repeated words (your marketing language is born here). Compare against the hypothesis threshold you wrote in advance: if you said "10 of 15 will rank it top-3" and got 4, the answer is a pivot signal not more interviews.
FAQ
People won't accept interview requests what do I do?
Check the invitation frame: nobody gives time to anything that smells like sales; "I want to learn, I'm not selling anything" + a time commitment (20-30 min) + something in return (a findings summary, an industry comparison, a small gift card) measurably lifts acceptance. Channel matters too: warm community introductions, mutual-connection referrals or the segment's physical venues (trade shows, meetups) convert far better than cold email. In B2B, doors open ten times more easily for a founder "doing research" than for a seller.
Everyone loves my idea in interviews, but nobody uses the product why?
The classic politeness bias + wrong question symptom: saying yes to "do you like it?" is free; using the product costs something. Two corrections: cut the idea pitch from interviews entirely (problem and past behavior only), and test micro-commitments that demand a price instead of words pre-registration, a waitlist email, a small deposit, a pilot slot on their calendar. Interest that evaporates when a price appears is interest that never existed.
After how many interviews can I start building?
It's a signal threshold, not a number: if within 10-15 same-segment interviews (1) the same problem consistently ranks near the top, (2) people already spend money/time on it, and (3) there are concrete complaints about current solutions move to the next validation step (landing page, prototype, pre-sales), not straight to a full product. Interviews are problem evidence, not solution evidence: the "problem validated, let's build" leap is validation's most expensive mistake.
Should I do the interviews myself, or can I outsource them?
Early on, the founder must do them it's non-delegable: the interview's value isn't in the notes but in the pattern forming in the founder's mind and the direct contact with customer language. Thirty agency-run interviews teach less than ten the founder runs personally. When scale demands it (other geographies, many segments), outside help is fine but the founder should have personally done the first 5-10 in every segment.