Customer discovery: what it is and how to run it

Customer discovery is the early work of testing what you believe about your customers before you commit to building for them. You write down who you think has the problem, how they handle it today, who would pay and how they buy, and then you talk to people outside your company to find out which of those beliefs hold up.

The term comes from Steve Blank's customer development method, where customer discovery is the first step. Blank's shorthand is that there are no facts inside the building, so you have to get outside. In a post on founder beliefs, he describes the purpose of discovery as using evidence from outside the building to confirm, reject or change your hypotheses, starting with who the customers are, what they need and who pays.

A worked example

Suppose you want to build a tool that helps property management companies track whether their maintenance vendors have current insurance certificates. Your beliefs might be:

• Property managers chase expired certificates by email and spreadsheet.
• A missed certificate has caused real trouble, such as an uninsured vendor on site.
• The person who feels the pain is a property manager, but a regional director or procurement lead approves new software.
• Companies with a few hundred units or more care most.

Each line is a guess you can check. Discovery means talking to property managers, the people who approve vendors and the people who would sign the contract, and asking what happened the last time a certificate expired.

Write down what you believe first

Blank warns about two failure modes. Teams with no stated beliefs go out, collect a long list of feature requests and hand it to engineering without any way to choose between them. Teams with strong beliefs who never talk to customers end up settling arguments by who is loudest.

So before you schedule anything, write your beliefs as short statements someone could prove wrong. "Property managers hate compliance" cannot be proven wrong. "The last expired certificate was found by accident, after the vendor had already done work" can be.

Decide whose account you need

In business markets, the person who does the work, the person who picks a tool and the person who pays are often different people. Plan to hear from each. Our guide to interviewing B2B buying roles covers what to ask each of them, and choosing a first customer segment helps if you are unsure which kind of company to start with.

Early customers are hard to reach when you have no product and no network in the industry. The finding customers guide lists places to look. For the property example, a directory page such as procurement professionals in property management shows the kinds of people who choose and manage vendors.

Ask about what already happened

Discovery interviews work best when they focus on recent, specific events. GOV.UK's interview guidance makes the same point: focus on stories and real examples, not on how things should happen. Ask about the last time something happened, not whether a feature would be useful. "Tell me about the last vendor whose certificate expired. How did you find out?" gives you facts. "Would you use a tool that tracks certificates?" gives you politeness.

YC's talk on how to talk to users covers running these conversations with both current and potential users. Keep solution talk for later. The difference is covered in problem interviews versus solution interviews. For ready-made wording, use the customer interview questions or the 30-minute discovery script.

Look for commitments, not compliments

Praise for your idea is weak evidence. Stronger evidence is something the person has already done or is willing to do: they built a spreadsheet to track this, they paid for a partial fix, they introduce you to the person who approves vendors, or they agree to a pilot. Blank writes that the best validation you can get is an order.

If nobody has tried to solve the problem and nobody will commit time or money, that is useful information too. It may mean the problem is real but not urgent enough to pay for.

Update your beliefs after each batch

After every few conversations, go back to your list. Mark each belief as supported, contradicted or still unclear, and note which conversation told you so. Change the next round of questions and the people you invite to match what is still unclear. The interview analysis guide and the decision memo template help you turn notes into a decision.

Discovery does not end with a number of interviews. It ends when you can predict much of what the next person will tell you and you have enough evidence to make your next decision, whether that is building, changing the customer you target or dropping the idea. The interview count guide discusses how to judge that.

Getting your first conversations booked

Start with the belief you are least sure about and the role that can answer it. If your network does not reach those people, describe the role in a people search. Instant Expert finds people whose work matches your question. You review them, it sends the invitations, and you pay for each call that gets booked. Plan your first round with the research plan guide before you send the first invitation.