What Is Customer Discovery? The Customer Development Model Explained

Customer discovery is the process of testing your startup's assumptions by talking to the people you think you're building for. You write down what you believe about the customer, the problem, and the way you'll make money. Then you go find out which of those beliefs are wrong.

The term comes from Steve Blank's Customer Development model, where it's the first of four steps. It's also the work most founders skip, because building feels more productive than asking. This guide covers the four steps, how discovery works in practice, how to run interviews without fooling yourself, and where it ends and the next step begins.

Where the term comes from

Steve Blank described the model in his book The Four Steps to the Epiphany and later in a series of blog posts he called the Customer Development Manifesto. The idea was a companion to the way companies already built products. Product development had a clear process. The customer side of an early company, Blank argued, had none.

In part five of that manifesto, Blank says the model groups all the customer-related activities of an early-stage company into four steps: Customer Discovery, Customer Validation, Customer Creation, and Company Building. Customer discovery sits at the start because it comes before anything else can make sense. If you haven't found out who has the problem and how much it hurts, there's nothing yet to validate or scale.

The four steps of Customer Development

Blank's own one-line definitions are the clearest way to see how the steps differ. They're summarized here from the same post.

Two things stand out. First, the steps are not a straight line. Blank writes that each step is iterative and that the model assumes it will take several passes through the four steps to get it right. A startup that can't find enough buyers in validation goes back to discovery to work out what it missed. Second, the first two steps are about searching for a business model, and the last two are about executing on one you've found. This matches Blank's definition of a startup as an organization formed to search for a repeatable and scalable business model.

The core principles: hypotheses and getting out of the building

Blank has boiled customer discovery down to a short list of principles. In a 2020 post on doing discovery remotely, he restates them:

• There are no facts inside the building, so get outside.
• All you have are a series of untested hypotheses.
• You can test your hypotheses with a series of experiments with potential customers.

"Get out of the building" is the slogan that stuck. It means your office can generate guesses but not evidence. The same post makes clear the principle isn't about location alone. Blank adds that in-person interviews aren't the only way to test hypotheses, and that video calls work too.

The word hypothesis matters here. A hypothesis is a statement that can turn out to be false. "Small clinics will love this" can't be tested. "Clinic managers with fewer than 10 staff spend more than five hours a week rescheduling appointments by phone" can. If you can't imagine a conversation that would prove you wrong, you haven't written a hypothesis yet.

How customer discovery connects to the business model canvas

Blank pairs customer discovery with the Business Model Canvas. In a 2014 post on Alexander Osterwalder's book Value Proposition Design, he describes the Lean Startup process as three parts: a business model canvas to frame hypotheses, customer development to get out of the building and test them, and agile engineering to build minimum viable products.

So the canvas is the list of things to test, and discovery is how you test them. The two boxes Blank calls the most important are the value proposition and the customer segment, the pair he labels product/market fit. The Business Model Canvas lets you lay all nine blocks out as assumptions. The Value Proposition Canvas zooms in on that critical pair, mapping customer jobs, pains, and gains against what you offer.

Key Facts: Customer Discovery

• Steve Blank's Customer Development model has four steps: Customer Discovery, Customer Validation, Customer Creation, and Company Building.
• Blank describes customer discovery as testing hypotheses and understanding customer problems and needs, in front of customers, by the founders.
• His core principle is that there are no facts inside the building and all you have is a series of untested hypotheses.
• Blank pairs the model with the Business Model Canvas, which he calls the tool to frame all the hypotheses of your startup.
• The NSF I-Corps program requires teams to complete a minimum of 100 potential customer interviews during its seven-week training program.
• The Mom Test book by Rob Fitzpatrick is built on three rules for customer conversations, including asking about specifics in the past rather than opinions about the future.

How customer discovery works in practice

Discovery is a loop, not a one-time event. A common way to run it looks like this.

1. Write down your hypotheses

Start with a canvas or a simple list. Cover who the customer is, what problem they have, how they solve it today, what they'd pay, and how you'd reach them. Then rank them by risk. The assumption that would kill the business if it's wrong goes first. The riskiest assumption test explains how to pick and test that one.

2. Find people to talk to

You need real people in your target group, not friends and family. Where ideas come from shapes who you talk to, so it helps to know where startup ideas come from before you list your first interviewees. Ask for introductions, post in communities where your customer spends time, and use cold outreach with a short, honest message about the problem you're studying.

3. Run problem interviews first

A problem interview explores the customer's world: how they do the work now, what goes wrong, what they've tried, and what it costs them. You're not presenting anything. The goal is to learn whether the problem is frequent, painful, and expensive enough that people already act on it.

4. Run solution interviews later

Once the problem looks real, a solution interview tests your proposed answer. You show a sketch, a mockup, or a description and watch how people react. Keep the order straight. Showing the solution too early biases everything after it, because people start judging your idea instead of describing their lives.

5. Look for patterns, then update

After a handful of conversations, review your notes. Which hypotheses held up? Which didn't? Change the hypothesis list, narrow the segment, and run the next round. When a pattern keeps repeating across people in the same segment, you're learning something real. This is the same learn-and-adjust cycle behind the Lean Startup method.

Interview pitfalls, and what The Mom Test says to do

Most bad discovery isn't a lack of effort. It's a lack of reliable data. People are polite, and they want to help. Rob Fitzpatrick wrote The Mom Test as a hands-on handbook for exactly this problem, with a focus on how to avoid biased feedback and how to tell whether someone will really buy. The name comes from the idea that your mother will tell you your business idea is great, even if it isn't.

A reader's summary of the book lists three rules for passing the test:

1. Talk about their life instead of your idea.
2. Ask about specifics in the past instead of generics or opinions about the future.
3. Talk less and listen more.

The same summary notes a related point about compliments: they cost the customer nothing, and even sincere ones pull the conversation toward your idea instead of their workflow. And it suggests asking how the customer currently handles the problem and what alternatives they've already looked at, because if they haven't looked for a fix, they're unlikely to look for or buy yours.

Here's how those rules change the questions you ask.

A few habits help. Keep the conversation on the customer's last real experience. Take notes on what they did, not what they said they might do. End each interview by asking who else you should speak to, and whether there's something you should have asked. And treat a compliment, a vague "send me a demo," or a promise to follow up as noise unless something concrete follows, like a pilot, a calendar slot, or money.

Customer discovery at scale: the NSF I-Corps program

A good picture of discovery as an institution is the National Science Foundation's Innovation Corps, or I-Corps. NSF describes it as hands-on training where participants get direct experience in customer discovery, which it defines as talking to potential customers, partners, and other industry stakeholders. The results let the team judge the commercial potential of their technology.

The program is built around interviews. NSF's page for national team applicants says teams must complete a minimum of 100 potential customer interviews during the seven-week training program. A separate NSF page for accepted teams says participants should plan on at least 15 hours a week and that the focus of the program is customer discovery.

Treat these as the requirements of one specific program for scientists and engineers, not a rule for every startup. A software founder doesn't need 100 interviews to learn that a problem is real. But the shape is instructive. The expectation is that evidence comes from many conversations in a short window, and that the team keeps going until the same patterns repeat. Steve Blank, who helped teach early I-Corps classes, says his own classes ask for 100 customer or beneficiary conversations in 10 weeks.

Customer discovery vs customer validation

The two terms get mixed up constantly, so it helps to keep them apart.

Discovery is about the problem. Validation is about the sales model. You can finish discovery with strong conviction that a problem exists and still discover in validation that no one will pay enough, or that the sales process doesn't repeat. When that happens, Blank's model sends you back to discovery. That's not a failure of the method. It's the method working as designed.

The two steps also tie to different questions around problem-solution fit, which asks whether your idea addresses a real problem, and later to the work of building a minimum viable product to test it with real users.

Common customer discovery mistakes

1. Interviewing the wrong people. Friends, colleagues, and fellow founders are easy to reach and poor sources of evidence about a market you haven't defined.
2. Pitching instead of asking. The minute you describe your product, the conversation shifts from their life to your idea.
3. Leading with the solution. Running solution interviews before you understand the problem produces polite feedback on the wrong thing.
4. Collecting opinions instead of behavior. What people did last month is stronger evidence than what they say they'd do next quarter.
5. Counting compliments as validation. Enthusiasm without a commitment is a weak signal.
6. Talking to too few people. Three conversations can feel like a pattern. Keep going until the answers start repeating inside one segment.
7. Never changing the hypothesis list. If your canvas looks the same after 20 interviews, you either learned nothing or weren't listening.
8. Doing it alone. Having a second person in the room helps catch things you wanted to hear.

What to do after discovery

When discovery works, you come out with a sharper picture: a narrow customer segment, a problem they already spend time or money on, and a few hypotheses that survived contact with reality. If you'd like to frame the customer's goals more explicitly, Jobs to Be Done is a useful lens for the interview notes.

The next step is to test whether people will pay and whether you can sell repeatedly. For where that sits in a company's life, see the idea stage of a startup, the earliest part of the journey, where most customer discovery happens.

Related reading

• The idea stage of a startup
• Where startup ideas come from
• Problem-solution fit
• Minimum viable product
• Lean Startup method
• Riskiest assumption test
• Founder-led sales
• Business Model Canvas
• Jobs to Be Done