How to run customer discovery interviews that actually tell you something.

Customer discovery interviews are useful only when they survive contact with the evidence. This guide builds the framework that does.

The fastest way to get a false positive from a customer discovery interview is to ask "would you use this?" People answer yes to hypothetical products all the time. The result is a false positive that survives weeks of internal cheerleading before reality intervenes. The question invites the person to imagine themselves as a better version of themselves: more organized, more efficient, more willing to adopt new tools. The answer describes that hypothetical person, not the actual buyer in front of you.

The only questions that produce reliable information ask about past behavior. What did you do the last time this problem came up? How long did that process take? What did you do when it broke? Past behavior is not hypothetical. The person in front of you either did the thing or did not. They cannot project optimism onto the answer.

Rob Fitzpatrick, whose 2013 book "The Mom Test" remains the most cited work on customer interview methodology among startup practitioners, described the principle this way: you are seeking facts about the customer's life and world, not opinions about your idea. The difference in question framing determines whether you get evidence or encouragement.

The interview subject needs to be the actual decision-maker or user for the problem your product addresses, not a proxy or an adjacent role. If you are building a tool for heads of engineering, interview heads of engineering, not CTOs or senior engineers. If you are building for procurement managers, interview procurement managers, not finance directors.

Finding the right subjects is itself a research exercise. LinkedIn Sales Navigator, Twitter, and industry-specific Slack communities are the fastest channels for reaching a specific title at a specific company size. Cold outreach for discovery interviews has a 20 to 40 percent response rate when the message is brief (under 75 words), references a specific problem in the first sentence, and asks for 20 minutes rather than an hour, according to research published in the Journal of Business Venturing in 2022 covering 800 founder interview attempts.

Warm introductions convert at a higher rate, typically 60 to 80 percent, but they introduce selection bias. The people your network connects you to are likely more similar to your founders than to your eventual customers. For discovery purposes, the cold approach that reaches a more representative sample of the buyer population is often more valuable than a warm introduction to someone who resembles you.

Open with context, not a pitch. Tell them you are researching how companies like theirs handle a specific category of work, and that you are not there to sell them anything. This reframes the conversation from evaluation to exploration, and people give more honest answers in exploratory conversations than in evaluative ones.

Move immediately to their current state. Ask them to walk you through the process for the thing your product would address. Listen for time spent, people involved, and handoffs between systems. Do not interrupt with information about your product. Do not correct their description if it implies a different problem than you expected. Just listen and take notes.

The most important follow-up question is: when did this last cause a problem you had to deal with? This surfaces the real-world failure mode, not the theoretical one. If the person cannot remember a time when the process caused a problem, that is a signal the problem is not acute enough to drive a purchase decision, even if it theoretically exists.

"Can you walk me through the last time you had to do X?" is the most reliably useful question in discovery. It forces a specific, recent example and anchors the conversation in facts rather than preferences.

"What did you do when that went wrong?" surfaces the current workaround and the economic cost of the failure mode. If the workaround is a 4-hour manual process that happens three times per week, you have a specific quantifiable cost.

"Who else was involved in that process?" maps the stakeholders in the purchase and use decision. You need to know whether the person you are talking to is the buyer, the user, or the influencer, because each role requires a different sales approach.

"How do you currently measure whether this process is working?" tells you the metric your buyer cares about, which is the metric your product needs to move.

"What would it be worth to you if that process was twice as fast and produced a reliable output the first time?" is the closest you can get to a price conversation without naming a number. The answers are aspirational but they indicate order of magnitude.

"Would you pay for a tool that solved this problem?" is the most common false positive generator. The answer is almost always yes, but the yes refers to a hypothetical version of the person who has already decided the problem is worth solving. It does not predict actual purchase behavior.

"On a scale of one to ten, how frustrated are you with the current process?" produces numbers that feel like data but are not comparable across respondents. Someone who rates their frustration a 7 may or may not be more frustrated than someone who rates it a 5. The cardinal scale creates an illusion of precision.

"If we could do X, Y, and Z, would that solve your problem?" is solution validation disguised as discovery. You are testing whether the person likes your idea, not whether the problem you are solving is real and significant.

After five interviews, review your notes for convergent themes: problems that appear in multiple interviews in similar forms, failure modes that recur across different companies, and economic costs that cluster around a specific range.

Divergent themes (a problem that appears in one interview but not others) are not evidence that the problem is irrelevant. They are a signal to probe further, either with additional interviews or with secondary research to understand why some buyers experience the problem and others do not.

The output of discovery should be a clear statement of the specific problem, the specific buyer, the frequency and cost of the failure mode, and the key question that remains open. That statement is the foundation for your validation plan and, eventually, your pitch.

The fastest cold-list channel for B2B is LinkedIn Sales Navigator combined with a personalized email. The pattern: search by job title, company size, and industry. Export 75 contacts. Find emails via Hunter.io or Apollo. Write a five-sentence email that asks for advice, not for a pitch meeting. Aim for a 15 to 25 percent reply rate.

The second cold channel is community participation. Slack groups, Discord servers, Indie Hackers forums, Reddit communities, and industry-specific Substacks each have a community manager and a member directory. Show up for two weeks before asking for anything. Comment substantively on others’ posts. Then DM with a specific, short question. Reply rates from authentic community participation are typically 40 to 60 percent.

The third channel is state licensing boards, professional associations, and trade group directories. These are public and underused. For regulated industries (legal, medical, financial advisory, real estate), state databases list every practicing professional with contact info. Cold emails sourced from licensing boards tend to outperform LinkedIn sourcing because the contacts are actually working in the segment.

The opening is calibration. Two minutes to confirm the person is in your ICP and to set context: "I’m exploring a product idea for X. I’m not selling anything today. I want to learn about your current workflow." This frame changes everything. The respondent stops performing for a salesperson and starts answering honestly.

The first half is current-state mapping. "Walk me through what you do today for X." Take notes verbatim. Let silence sit. The first answer is usually rehearsed; the second answer, after a pause, is the real one. The Mom Test discipline applies here: do not pitch, do not validate, do not ask hypothetical questions about whether they would buy something. Ask only about specific past behavior.

The second half is friction mapping. "When was the last time this process broke? What did you do?" The story is the signal. Founders who try to summarize across stories miss the texture. A single concrete example of a buyer pulling out a credit card to solve the problem with whatever they could find is worth more than ten abstract complaints.

The close is forward-looking but evidence-anchored. "If a tool existed that did X, what would you do tomorrow?" Treat the answer as a hypothesis, not a commitment. Buyers consistently overestimate their willingness to switch in interviews. The discount is roughly 50 percent: half the people who say "yes, I would switch" actually do. Use the answer to size the upper bound, not to forecast revenue.

Twenty-five interviews produce 25 transcripts and a strong temptation to act on the loudest signal. The discipline is to wait until all 25 are complete before synthesizing. Build a single document with one row per interviewee: ICP fit (yes/no), current solution, top three pain points with frequency, switching willingness on a 1 to 5 scale, willingness to pay at three price points. Cluster the rows by ICP fit. Look for patterns within the "yes" cohort, not across the whole sample.

The output is a one-page memo with frequency counts. "Of the 23 ICP-qualified interviews, 19 named workflow speed as their top pain point. 14 reported active willingness to switch from their current tool. 11 said they would pay $200+ per month for a 4x speed improvement." That is a defensible synthesis. "Most people seemed excited about the idea" is not.

Verdikt’s methodology treats the 25-interview synthesis as a Tier 2 source in the citation library: cited as community evidence with named frequency counts. The synthesis sits alongside primary database sources and competitive analysis in the memo, not as the headline but as the customer-evidence column that distinguishes a validated wedge from a hypothetical one.