Customer Interviews

---
name: customer-interviews
description: Run user interviews that surface past behavior instead of polite predictions, recruited and synthesized honestly. Use when gathering qualitative evidence for product decisions.
---

# Customer interviews

People are terrible at predicting their behavior and excellent at
being polite. Interviews extract truth only when they ask about the
past, chase specifics, and never pitch: the mom-test discipline.

## Method

1. **Ask about the past, never the hypothetical.** "Walk me
   through the last time you handled X" beats "would you use a
   tool that does X" every time: the second question harvests
   compliments. Chase the specifics: when, what did you do
   next, what did it cost, who else was involved, what did you
   try before (the workaround hunt: see product-discovery
   step 3).
2. **Never pitch in a learning interview.** The moment you
   describe your solution, every answer becomes a kindness.
   Keep solution demos in separate usability/feedback sessions
   with their own rules; a learning interview about the
   problem space mentions your product last or never.
3. **Recruit for the segment, screen hard.** Define who has
   the problem (role, context, recent behavior: "managed a
   team budget in the last 90 days"), screen with behavioral
   questions, and pay fairly. Five to eight interviews per
   segment surfaces the pattern; friends, fans, and whoever
   answers fastest produce warm noise (see
   product-discovery's evidence grades). In B2B, interview
   buyer, user, and admin separately: they disagree in
   load-bearing ways.
4. **Run it as listening with a spine.** A discussion guide of
   topics (not a script), open questions, then silence: let
   them fill it. Follow surprise: "say more about that"
   outperforms your next planned question. Capture verbatim
   quotes (record with consent); your paraphrase already
   contains your bias.
5. **Synthesize across interviews, tagged to evidence.**
   Within 24 hours per interview: key quotes, observed
   behaviors, pains with frequency/severity. Across the set:
   cluster into themes with counts ("6 of 8 rebuilt the
   report manually each month"), separating observations from
   interpretations (see research-synthesis for the general
   method). Themes feed the opportunity map (see
   product-discovery) and experiment designs (see
   ab-test-design for what quals cannot answer).
6. **Close the loop into decisions.** Each round ends with:
   what we believed, what we heard, what changes (a killed
   assumption, a re-ranked opportunity, a new experiment):
   recorded (see decision-journals). Interviews that end in
   a slide deck nobody acts on teach the org that talking
   to users is decoration.

## Boundaries

- Interviews establish existence and texture of problems,
  not market size or preference shares; pair with behavioral
  data and experiments for magnitudes (see product-metrics,
  ab-test-design).
- What people say they do and what logs show diverge; when
  they conflict, logs win on frequency, interviews win on
  why.
- Usability testing (can they operate the interface) is a
  different protocol with tasks and think-aloud; do not mix
  it into discovery interviews and call both "research"
  (see accessibility-review and design-critique for
  adjacent lenses).