handbooksCustomer interviews
Why asking people what they want fails, how to get a specific story instead, and the one-page snapshot that stops your interviews from evaporating a week later.
- Never ask what customers want. Ask them to walk you through the last specific time they did the thing.
- People describe their ideal behavior when asked in general, and their actual behavior when asked about a particular instance. Only the second one is evidence.
- Capture each interview on a one-page snapshot the same day, or the insight leaks away within a week.
- Weekly is the habit. Automate the recruiting or it will not survive a busy month.
The problem with asking
"What would you like us to build?" produces confident answers that do not predict anything. Not because people lie, but because they are being asked to imagine a future self, and everyone imagines a more organized version of themselves than the one who shows up.
Ask someone how they manage their team's tasks and you get their intended process. Ask them to walk you through what happened yesterday and you get the sticky note on the monitor and the message they sent themselves at midnight. The second version is where the opportunities live.
| Instead of | Ask |
|---|---|
| "What features would help you?" | "Tell me about the last time you tried to do X" |
| "Do you find this frustrating?" | "Walk me through what happened after you clicked that" |
| "How often do you use this?" | "When was the most recent time? What happened?" |
| "Would you use a feature that did Y?" | "What did you do instead, last time?" |
Getting the story out
Your job in the interview is to keep the person anchored in one real occasion. They will drift toward generalities, because generalities are easier to produce, and you steer back.
- Start with a specific instance. "Think of the last time you..." Get a date, a place, anything concrete.
- When they generalize, pull back to the event. "You said you usually skim them. What about that one on Tuesday?"
- Follow the timeline. What happened next, and after that.
- Chase emotion. When someone sounds annoyed, relieved, or resigned, stop and ask what was happening there. Emotions mark the places where needs went unmet.
- Say less. Silence gets filled with detail that a follow-up question would have cut off.
Some people will hand you feature requests because they are trying to be useful. Take the request, then ask what happened that made them think of it. The story behind the request is the part you can use.
Snapshots, or it did not happen
Insight from an unrecorded interview has a shelf life of a few days. The habit that fixes this is a single-page snapshot, filled in the same day, one per interview.
“I keep two calendars because I do not trust the sync, so I check both before I book anything.”
Booked a room already taken last Tuesday because the sync had silently failed for three days. Now checks the source calendar by hand before every booking.
“Trust the sync” is not one opportunity, it is a symptom. The real need is a way to tell that a sync has failed, not a better sync.
| Field | What goes in it |
|---|---|
| Who | Name, role, and a photo if you have permission |
| The quote | One memorable line, in their exact words |
| The story | What actually happened, in the order it happened |
| Opportunities | The needs and pain points you heard, in their language |
| Insights | What this changes about what you believed |
The photo and the direct quote are not decoration. Months later, "the woman who kept two calendars because she did not trust the sync" is retrievable in a way that "Interview 14" is not. Human memory works on faces and phrases.
Pick an interview from three weeks ago. If you cannot say who they were and what they were trying to do without opening a document, your capture habit is not working yet.
Make it weekly, make it automatic
The weekly cadence is what separates continuous discovery from research projects. It only survives if recruiting is not a fresh negotiation every time.
- Recruit from inside the product. A small prompt after a relevant action, offering a short conversation, fills a calendar faster than any outbound effort.
- Use the interactions you already have. Support conversations, sales calls, and onboarding sessions all contain people who agreed to talk to you.
- Keep a standing slot. A recurring hour on the trio's calendar that customers book into. Nobody has to schedule anything.
- Interview the same people again. Returning to a customer you already know skips the context-building and gets to the change since last time.
One interview per week, consistently, beats six interviews in a burst followed by two quiet months. The value is in never being far from the customer when a decision arrives.
Go together
The whole trio should be there. The designer hears hesitation, the engineer hears the workaround the customer built, the product manager hears what they already pay for. One person interviews and the others listen and ask at the end.
Sharing notes afterward is not the same thing and never has been. Notes carry conclusions. They do not carry the pause before an answer.
Anti-patterns
- One designated interviewer. The rest of the team receives findings secondhand and argues with them.
- A long discussion guide. A script with twenty questions turns a conversation into a survey read aloud. Bring three areas to explore.
- Interviewing in bursts. Sporadic rounds mean every one starts from scratch.
- Asking about the future. "Would you use this?" is a question about imagination, not behavior.
From interviews to the tree
Each snapshot gives you needs and pain points in customer language. Those go onto the opportunity space, where they get grouped, deduplicated, and compared.