Customer discovery is a conversation, not a survey, and the entire method depends on asking questions that reveal what someone actually does rather than what they think they might do. Most founders who "already did customer discovery" made the same mistake: they described their idea and asked if people would use it. Everyone is polite. Nobody lies exactly, but almost nobody tells the truth either, because hypothetical enthusiasm costs nothing to express.
What changed in 2026
- The Mom Test style of interviewing became the default standard, taught widely enough that most accelerator programs and startup courses now teach behavioral questioning over opinion-gathering as the baseline.
- AI transcription and synthesis tools sped up the analysis step. Turning fifteen hour-long interviews into a pattern used to take days; transcription and thematic tagging tools compress that significantly, though the judgment about what the pattern means still has to be yours.
- Recruiting participants got both easier and noisier. Online communities make it simple to find people to talk to, but also easier to accidentally recruit people who are not your real target buyer.
- Discovery increasingly extends into existing customers, not just prospects — checking in with current users about adjacent problems has become a standard second use of the same method.
Good questions vs bad questions
The format of the question determines whether the answer is useful.
| Instead of asking |
Ask this |
Why it works better |
| Would you use an app that did X? |
What do you currently do about X? |
Reveals real behavior, not a hypothetical guess |
| Do you think this is a good idea? |
Tell me about the last time X happened. |
Gets a specific story instead of an opinion |
| Would you pay for this? |
What are you paying for or doing instead right now? |
Reveals actual willingness to spend, not politeness |
| Is this a problem for you? |
How much time or money does this cost you today? |
Quantifies the pain instead of a yes or no |
| What features would you want? |
Walk me through your process step by step. |
Surfaces the real workflow, not a wishlist |
How to run a discovery interview, step by step
- Recruit people who currently deal with the problem, not people who might one day use your product. Existing behavior is the entire point.
- Open with their world, not your idea. Ask them to walk you through the last time the problem came up, in detail.
- Follow up on specifics. "What did you do next," "how much did that cost you," "what did you try before that" — specifics are where the signal is.
- Do not mention your product idea until the interview is essentially over, if at all. The moment you pitch, you start getting polite feedback instead of honest information.
- Write up notes within an hour, while the conversation is fresh. Tag recurring phrases and workarounds across interviews as you go.
- Stop around fifteen to twenty conversations once the same two or three problems and workarounds keep repeating. More interviews after that mostly confirm what you already know.
Discovery tells you whether a real problem exists. It does not tell you whether people will pay for your specific solution — for that, pair it with how to build an mvp in 2026 once you have a pattern worth testing.
Common mistakes
Pitching instead of listening. The instant you describe your idea, you have traded a discovery interview for a sales pitch, and the data is no longer reliable.
Interviewing friends and family. They want to encourage you, which makes their feedback almost uniformly positive and almost uniformly useless.
Asking about the future instead of the past. Hypothetical questions get hypothetical, overly generous answers. Past behavior is harder to fake.
Stopping after three or four conversations. A handful of interviews is a handful of anecdotes, not a pattern. Push through to at least a dozen before drawing conclusions.
FAQ
How many customer discovery interviews should I do?
Fifteen to twenty is a reasonable range for most early-stage ideas, enough to see a pattern without stalling indefinitely on research instead of building.
Should I record the interviews?
With permission, yes. Reviewing your own notes after several conversations blur together is unreliable; a recording or transcript lets you catch details you missed live.
What if everyone I talk to says something different?
That is itself a signal — either your target segment is too broad, or the problem is not as sharp or common as you assumed. Narrow the segment and try again.
Is customer discovery the same as market research?
Related but different. Market research often uses surveys and existing data at scale; customer discovery is small-sample, qualitative, one-on-one conversation focused on real behavior.
Where to go next