How to Run Customer Discovery Interviews That Tell You the Truth

Most founders talk to customers. Far fewer learn anything from it, because the conversations are designed to confirm the idea rather than test it. Here is how to run discovery interviews that surface real problems, real behavior and real buying intent.

A good discovery interview feels like a conversation about the customer's life, not a pitch about your product.

Key takeaways

• Ask about the customer's past behavior, not their opinion of your idea. People are polite about ideas and honest about what they actually did.
• Do not pitch during discovery. The moment you describe your product, the interview turns into a sales call and the data gets worse.
• Aim for 15 to 20 interviews within one narrow segment before you draw conclusions.
• Write down the three facts that would kill your idea before you start, so you notice them when they show up.

Customer discovery is the practice of talking to potential customers before you build, to learn whether the problem you want to solve is real, painful and worth paying to fix. Steve Blank, the Stanford lecturer who wrote The Four Steps to the Epiphany, built the customer development method around a simple instruction: get out of the building and talk to customers, because no business plan survives first contact with them.

The problem is that talking to customers is easy to do badly. If you ask "would you use this?" you will get a yes, because people like to be encouraging. In CB Insights' analysis of startup post-mortems, roughly a third of failed founders said their company died because there was no market need. Many of them had talked to customers. They simply asked questions that could not give a painful answer.

Why most customer interviews lie to you

Rob Fitzpatrick's book The Mom Test is named after a simple idea: your mother will tell you your idea is wonderful because she loves you, and most other people will tell you the same thing because they are polite. The fix is to ask questions that even your mother could not answer with a flattering lie.

Three kinds of answers are nearly worthless in discovery:

• Compliments. "That sounds great" and "I love it" tell you nothing about behavior. Note them, ignore them and move on.
• Hypotheticals. "I would definitely pay for that" is a prediction about a future self who has never had to open a wallet. People are terrible at predicting their own behavior.
• Generalities. "I usually try to stay organized" sounds like data but describes a self-image, not an action.

What you want instead are specific facts about the past: what happened the last time they hit this problem, what they tried, what it cost them and what they did next.

Common mistake

Opening the call with a demo or a one-line pitch "for context." Once the customer knows what you are building, every answer becomes an attempt to be helpful about your product rather than honest about their life. Save the pitch for a separate conversation once you understand the problem.

Who to interview and how many

Discovery works best when the segment is narrow. "Small business owners" will give you twenty conflicting stories. "Owners of independent gyms with one location and under ten staff" will give you patterns by the fifth call.

Pick the segment you believe has the problem most acutely, then find people who match it. Good sources include:

• Your own network and second-degree introductions, asked for a favor rather than a sale
• Online communities, forums and professional groups where your segment already talks about the problem
• Industry events, trade shows and local meetups where you can book short conversations in person
• Cold outreach on LinkedIn or email with a short, specific request to learn rather than to pitch

15 to 20 The number of interviews in a single segment at which most teams find the same answers repeating, which is the signal to stop and act A common rule of thumb, not a scientific threshold

If you are still hearing new problems after twenty conversations, your segment is probably too broad. Narrow it and start a fresh count.

How to run a discovery interview step by step

1. Write your riskiest assumptions first. Before any interview, list the three or four beliefs your business depends on. For example: this problem happens weekly, the owner handles it personally, and current workarounds cost real money. These become the things you are trying to disprove.
2. Prepare an open script, not a questionnaire. Five to eight open questions are enough. Order them from broad to specific, and leave room to follow interesting threads.
3. Set expectations in the first minute. Say that you are researching how people handle a specific task, that there is nothing to buy and that you want to hear about their experience, including the boring parts.
4. Ask about the last time it happened. "Tell me about the last time you had to chase a late invoice" produces a story. Stories contain the facts you need: frequency, cost, emotion and the tools involved.
5. Dig into the workaround. Ask what they do today, how long it takes and what it costs. If they have never tried to solve the problem, it may not be painful enough to pay for.
6. Ask about money and effort already spent. Have they bought software, hired someone or lost a customer over this? Spending in the past is the strongest predictor of spending in the future.
7. Close with a commitment, not a compliment. Ask for an introduction to someone else with the same problem, a follow-up call or a look at an early version. A real commitment of time or reputation is worth more than any amount of praise.
8. Write notes within ten minutes. Record the facts, the exact phrases the customer used and anything that surprised you. Memory fades fast, and the surprises are the most valuable part.

Questions that work and questions that do not

The difference between a weak question and a strong one is usually a shift from opinion to behavior and from future to past. The table below shows some common swaps.

Instead of asking	Ask	Why it is better
Would you use a tool that does X?	How do you handle X today?	Reveals the real workflow and whether a solution already exists
Is this a big problem for you?	When did this last cause you trouble? What happened?	Gets a concrete story with real stakes instead of a rating
How much would you pay for this?	What have you spent trying to fix this so far?	Past spending is evidence; future spending is a guess
Do you like this feature?	Walk me through what you did the last time you needed this	Shows which steps are painful rather than which ideas sound nice
Who else has this problem?	Who else in your company deals with this? Could you introduce me?	Turns a vague answer into a referral you can act on

Pro tip

Use silence. After the customer answers, wait three or four seconds before speaking. Most people fill the gap with the detail they left out, and that second layer is usually where the real problem lives.

Keep the tone light and curious. People share more when they are not being sold to.

How to tell a real problem from a polite one

After a handful of interviews, you need a way to sort what you heard. Look for four signals in each conversation:

• Frequency: the problem comes up weekly or daily, not "sometimes."
• Cost: the customer can name a cost, in hours, money, lost deals or stress.
• Active search: they have already tried something, even a spreadsheet or a hired helper.
• Ownership: the person you spoke to is the one who feels the pain and has the authority to fix it.

A problem that scores well on all four is worth building for. A problem that is frequent but cheap, or expensive but rare, is usually a feature rather than a company.

Real-world example

Before Dropbox built a working product, founder Drew Houston released a short screencast demo in 2007 to test whether people wanted simple file syncing. The video was aimed at a technical audience on the Digg news site, and the company has said the waiting list grew from about 5,000 to 75,000 people overnight. The lesson for discovery is the same one that applies to interviews: a commitment, in this case an email address, says more than a compliment.

Pitfalls that quietly ruin your data

Leading the witness

"Don't you find it frustrating when...?" gives the customer the answer you want. Keep questions neutral. "How do you feel about..." is fine. "Isn't it annoying that..." is not.

Interviewing only enthusiasts

Early adopters who already love the idea will skew your results. Make sure some interviews come from people who were not excited to talk to you, because they look more like the market you will need to sell to later.

Talking too much

If you spoke for more than a quarter of the call, the interview was a pitch. A good target is to listen for at least three quarters of the time.

Hearing what you hoped for

Confirmation bias is strongest for the person who had the idea. Bring a co-founder or team member to every interview, or record it with permission, so someone else can check your reading of what was said.

Signs the problem is real

• The customer tells a detailed story without being prompted
• They already pay for a partial fix or a workaround
• They ask when they can try what you are building
• They offer introductions without being asked twice

Signs you are hearing politeness

• Answers are short, general and positive
• They cannot remember the last time it happened
• They say it is "interesting" or "cool"
• They decline a follow-up call or an introduction

Turning interviews into decisions

Interviews are only useful if they change what you do next. After each batch of five, sit down with your notes and update your assumptions list. Mark each one as supported, contradicted or still unclear. When an assumption is contradicted by more than a few conversations, stop defending it and change the plan.

Three outcomes are common at the end of a discovery round:

• Proceed: the problem is frequent, costly and owned by a reachable buyer. Start building the smallest thing that solves it.
• Narrow: the problem is real for a subset of the people you spoke to. Focus on that subset and interview more of them.
• Pivot: the problem is weak, but a different problem kept coming up. Rewrite your assumptions around that problem and start another round.

Your discovery checklist

• Write down your three riskiest assumptions and what evidence would disprove each one
• Define one narrow segment and list 30 people in it you can reach this month
• Draft six open questions about past behavior, workarounds and money already spent
• Book and run 15 to 20 interviews of 20 to 30 minutes each, with a note-taker on every call
• After every five interviews, update the assumptions list and decide whether to proceed, narrow or pivot
• Collect at least one concrete commitment from each promising interview, such as an introduction or a follow-up

Frequently asked questions

How long should a customer discovery interview be?

Twenty to thirty minutes is enough for most B2B and consumer problems. Longer calls are fine if the customer is enjoying it, but a short, focused conversation is easier to book and easier to analyze.

Should I pay people to take part in interviews?

You can, especially for consumer research where people have no professional reason to help. For business customers, a genuine interest in their problem usually works better, and paid participants sometimes tell you what they think you want to hear.

Can I do discovery interviews by email or survey instead?

Surveys are useful later, when you know what to ask and want to measure how common an answer is. Early discovery needs follow-up questions and tone of voice, which only a live conversation provides.

When is it time to show the product?

Once you have a clear picture of the problem and the workaround, run a second, separate round of calls where you show a prototype and ask for a commitment such as a pre-order, a pilot or a signed letter of intent.

The bottom line

Customer discovery interviews tell you the truth only when you ask about the past, listen more than you talk and refuse to pitch. Pick a narrow segment, run 15 to 20 structured conversations, sort what you hear by frequency, cost, active search and ownership, and let the evidence decide whether you build, narrow or pivot. The interviews that sting a little are the ones that save you a year.