Customer discovery, the founder's guide to talking to users

Customer discovery is the first stage of Steve Blank's Customer Development model, the structured practice of testing whether the problem you imagine your users have actually exists in their week. It is not sales discovery, which qualifies a buyer against a known offer, and it is not market research, which sizes a category from the outside. It is the founder, in conversation, hunting for evidence that a hypothesis is wrong.

Why founder-led discovery beats outsourced research#

Agencies can run interviews. They cannot do what a founder does between two interviews: hear something unexpected on Tuesday, change the roadmap on Wednesday, and book a call on Thursday to test the new theory. The feedback loop only closes when the person asking the questions also writes the code.

Signal density is higher in your own chair. You catch the half-second pause before someone answers a pricing question. You notice that three founders in a row used the same odd phrase to describe their workflow. You feel, viscerally, when you are boring someone. None of that compresses into a deck.

For roughly the first hundred conversations, the founder is the instrument. Hand it off after that, not before. A discovery report read by the CEO is a poor substitute for the CEO doing discovery.

The Mom Test, three habits worth internalizing#

Rob Fitzpatrick's The Mom Test fits on an index card. The premise is that even your mother will lie to you about your idea, not from malice, but from politeness. The fix is procedural.

• Ask about the past, not the hypothetical.:

  What someone did last Tuesday is data. What they would do next quarter is wishful narration. Anchor every question in a real, recent episode.

• Look for behavior, not opinion.:

  Find out whether they opened a tool, paid for a workaround, hired a contractor, or built a spreadsheet. Behavior is a paid receipt. Opinions are free.

• Listen before you pitch.:

  The moment you describe your idea, you turn an honest interview into a polite one. Save the pitch for the last five minutes, or for a separate call entirely.

Fitzpatrick's other useful move is to push back gently when someone gets enthusiastic. Compliments are a warning sign. Ask what they have actually done about the problem in the last month. If the answer is nothing, the enthusiasm is theater.

Three killer questions worth memorizing#

You can run a useful discovery call with three questions and the discipline to shut up after each one.

These three questions are not a script. They are a spine. Improvise around them, but come back to them every call so you can compare answers across conversations.

How many interviews to run#

There is no magic number, but there are useful zones. Treat them as rules of thumb, not quotas.

The underlying principle, borrowed from qualitative research, is theoretical saturation. You keep interviewing until new conversations stop producing new categories of insight. Saturation, not a calendar, tells you when a research cycle is done. For the mechanics of running each call well, see user interviews.

How to recognize you can stop#

Three signs show up in the same week when you are approaching saturation.

First, you can predict the next answer before the person finishes the sentence, and you are right more often than not. Second, your notes start repeating themselves. The new transcript looks like the last one with different names. Third, the surprises shift from content to intensity. You stop learning new problems and start learning which existing ones hurt the most.

When this happens, do not declare victory and stop. Change the segment. Talk to the people who said no, the ones who churned, the adjacent role you have been avoiding because they are harder to book. Saturation in one segment is the signal to widen the frame, not to close the notebook. Real stop conditions look like this: you can describe the buyer's week in their own words, you can predict which features they will dismiss, and you have a workaround inventory long enough to price against.

Founders who want to keep doing customer discovery beyond the first 20 calls without burning the calendar can use an AI-moderated tool like Diaform to run them asynchronously, then read structured summaries instead of re-watching recordings. See Diaform for founders for how early-stage teams use it pre-PMF.

One founder-mode thought to close on#

Discovery is the part of the job that does not look like work. There is no commit, no deploy, no metric moving on a dashboard. The temptation to skip it and go build is constant, and it gets stronger the more technical you are. Resist it. The cost of writing the wrong product for six months is larger than the cost of forty hours on the phone, by a factor that is hard to overstate. The founders who reach product-market fit are not the ones with the best ideas. They are the ones who were willing to sit through one more uncomfortable conversation than everyone else.