Key takeaways
- Finding 15 strangers is harder than asking the questions: The messages that get answered are short, name a specific problem in the first line, and ask for 20 minutes of advice rather than a meeting. Plan on sending several times more than the number of calls you need.
- 15 questions organized by what they uncover: Current behavior, alternatives and workarounds, urgency and willingness to pay, plus the questions you should never ask and how to read polite answers that mean nothing.
- Structure the call in three phases: 60 seconds of context with no pitch, then past-behavior questions, and the last 5 minutes for soft commitment asks. If you talk more than 30% of the time, you are pitching.
- In B2B, interview the end user first: They reveal the real workflow and workarounds. Interview the economic buyer second to map budget, approval chains, and competing priorities.
Good customer discovery questions figure out whether the problem you want to solve is real, painful, and worth paying to fix. Most founders never run customer discovery interviews at all. The ones who do usually do it wrong - they describe their idea, ask "would you use this?", hear "yeah, that sounds great," and count it as validation, which it isn't.
Rob Fitzpatrick wrote The Mom Test to fix this. Even your mom will lie to you about your startup idea if you ask the wrong questions. The right questions are about their life and their spending, not your product.
Will your idea survive the market?
Preuve AI runs 10 agents against live market data and links every claim to a source. Free analysis in 60 seconds.
What Is Customer Discovery Supposed to Uncover?
Customer discovery is not the same as market research. Market research tells you whether a market exists - competitors, demand signals, market size. Customer discovery tells you whether the pain is strong enough to build for.
After 10-15 interviews, you should know:
- Whether the problem is real: Do people describe it unprompted, or only when you bring it up?
- How they handle it today: Current workarounds, tools, manual processes, or spending
- How painful it is: Mild annoyance vs. costs them hours, money, or lost customers
- Whether they would pay to fix it: Based on current spending patterns, not hypothetical answers
- Who the buyer actually is: Job title, company size, decision-making authority
If you finish 15 interviews and cannot answer these five questions with specifics, your interviews were too soft. Go back and ask harder questions.
The Mistake Founders Make Before the Interview Starts
The biggest customer discovery mistake happens before a single question is asked: the founder pitches.
It looks like this: "I am building an AI tool that helps founders validate their startup ideas. Would you use something like that?" The interviewee says "yeah, totally." The founder writes down "strong interest." Nothing was learned, because the question asked for a prediction about hypothetical future behavior rather than a description of actual current behavior.
Humans are poor at predicting their own behavior. They are good at describing what they already do. That is what you should ask about.
Talk about their life instead of your idea.
Before your first interview, write down your three riskiest assumptions. What must be true for your idea to work? Those assumptions become your interview guide. Every question should test one of them.
How Do You Find 15 People to Interview When Nobody Knows You?
Finding 15 strangers to talk about their problems sounds impossible when you have no audience and no brand. It is not, if you keep the message under 75 words and the ask under 20 minutes.
Cold LinkedIn plus email is the fastest B2B channel. Search by job title, company size, and industry on LinkedIn Sales Navigator, export 50-75 contacts, and find emails through Hunter or Apollo. Write a five-sentence email that asks for advice, not a pitch meeting. The messages that get answered share three traits: under 75 words, a specific problem named in the first sentence, and an ask for 20 minutes rather than an hour. I have not measured a reply rate I would publish, and neither has anyone else I can point you to, so budget on volume rather than on a number.
A cold DM template that gets replies:
"Hi [name], I am researching how [role] at [company size] companies handle [specific problem]. Not selling anything. Would you do a 20-minute call this week to share how you handle it today? Happy to share what I learn across the interviews."
Community outreach converts higher but takes longer to set up. Join the Slack groups, Discord servers, Reddit communities, and niche forums where your target audience already discusses their problems. Participate for at least two weeks before asking for anything. Then send a short, specific DM. Community-sourced DMs land better because the recipient has already seen you contribute, not just drop in with an ask.
Warm introductions are the easiest yes, but that is the problem. The people your network connects you to are more similar to you than to your eventual customers. Use warm intros to practice your flow on the first two or three calls, then switch to cold outreach so your sample reflects the real market.
The math: whatever your reply rate turns out to be, measure it after the first 25 messages and size the rest of the campaign from your own number rather than someone else's. Then add a margin for no-shows on top. The User Interviews incentive report (last updated 2023), drawn from more than 20,000 research projects, found that no-show rates fall as the incentive rises. Underpaying does not save you money, it costs you calls.
Should you pay interviewees? For B2B professionals, the 2023 User Interviews incentive report recommends $100 per hour as a baseline for remote sessions. For early-stage customer discovery where you are interviewing peers or other founders, a $20 coffee gift card or a promise to share your findings is often enough. The risk with paid incentives is not that they corrupt the signal. A University of Surrey analysis found that payments can actually reduce non-response bias by including people who value their time too much to participate for free. The real risk is attracting people motivated by the payment rather than by the problem. If you pay, screen harder: ask about the problem in your recruiting message and reject anyone who cannot describe it before the call.
The friends-of-friends trap. If all 15 interviewees came from one Slack community or one LinkedIn connection's network, your sample is correlated. Aim for at least three independent recruiting channels. Mix cold outreach and communities, plus referrals from interviewees themselves (question 15 in the list below handles this).
How Do You Run a 30-Minute Customer Discovery Call?
Book 30 minutes so nobody feels rushed, but plan to use the 20 you asked for. Here is a time budget that keeps the conversation focused without turning it into a survey.
Minutes 0-1: Set context. Tell them you are researching how people in their role handle [problem area], that you are not selling anything, and that there are no wrong answers. This framing matters. People give more honest answers in exploratory conversations than in evaluative ones.
Minutes 1-5: Current state. Ask them to walk you through how they handle the problem today. Listen for tools, time spent, and handoffs between people or systems. Do not interrupt with information about your product.
Minutes 5-15: Follow the thread. Use 5-7 questions from the list below, starting with current behavior and working toward urgency and spending. Follow up on anything specific. The best insights come from follow-up questions you did not plan.
Minutes 15-20: Commitment. Ask for a referral (question 15), ask if you can follow up when you have something to show (question 14), and thank them. If they offer to pay for early access or introduce you to a colleague unprompted, that goes in your notes in capital letters.
Recording vs. live notes. Record if they consent. A phone on the table with a transcription app (Otter, Grain, or your phone's built-in recorder) means you can make eye contact instead of typing. Always ask first: "Do you mind if I record this so I can focus on listening instead of taking notes?" If they say no, take notes by hand and write up full notes right after the call, not an hour later.
Solo vs. paired. An interviewer plus a dedicated notetaker is the setup worth defaulting to when you have the headcount. The interviewer keeps eye contact and follows the conversation while the notetaker captures verbatim quotes and flags moments to revisit. If you are a solo founder, record and transcribe instead. Running both the conversation and the notes degrades both.
The 30% rule. In a customer discovery interview, the interviewee should be talking at least 70% of the time. If you are over 30%, you are pitching, not discovering. The customer talks while you listen, probe, and take notes.
15 Customer Discovery Questions That Actually Work
These are organized by what they uncover. You do not need all 15 in every interview - pick 5-7 based on your riskiest assumptions going in.
Questions About Current Behavior
These establish what the person does today. Current behavior is the most reliable predictor of future behavior.
1. "How do you currently handle [problem]?"
Opens the door without leading. If they have a detailed answer, the problem is real. If they shrug, it is not painful enough.
2. "Walk me through the last time you dealt with this."
Forces specifics. Generalizations hide the truth. A real story reveals the actual workflow, frustrations, and time spent.
3. "How often does this come up?"
Frequency matters. A painful problem that happens once a year is different from one that happens daily. Daily pain drives purchasing.
4. "What have you tried so far to fix it?"
If they have tried nothing, the pain is not strong enough. If they have tried 3 tools and none worked, you found a gap.
5. "What do you like about your current approach?"
Surprising answers here. People tolerate bad solutions because they are familiar. Your product needs to beat the switching cost, not the current solution's quality.
Questions About Existing Alternatives and Workarounds
These reveal your real competition: not the product you think you compete with, but the behavior you need to replace.
6. "What tools or services do you use for this right now?"
Maps your competitor landscape from the buyer's perspective. Often reveals competitors you missed in desk research.
7. "What do you dislike most about how you handle this today?"
Direct path to unmet needs. The complaints they volunteer are your feature priorities.
8. "Have you looked for a better solution recently?"
If yes, they are an active buyer. If no, they have accepted the status quo. Active buyers convert faster.
9. "Why did you stop using [previous tool]?"
Churn reasons from competitors are gold. They tell you what to avoid and what to guarantee in your positioning.
Questions About Urgency and Willingness to Pay
These separate "nice to have" from "need to have." The hardest and most important questions in customer discovery.
10. "How much time or money do you spend on this per month?"
Quantifies the pain. If they spend $0 and 0 hours, they do not care enough to pay you either. If they spend $500/month on a workaround, your pricing has a ceiling and a floor.
11. "What happens if you do nothing about this?"
Tests urgency. If the answer is "nothing bad, really," the problem is not urgent enough to drive a purchase. If the answer involves lost revenue, wasted time, or missed opportunities, you have urgency.
12. "If a solution existed that did [core value], what would it be worth to you?"
Weaker than observing actual spending, but useful for framing. Compare their answer to what they currently spend. Big gaps mean they are guessing.
13. "Who else in your company would need to approve buying this?"
B2B essential. Reveals the decision-making chain. If the person you are interviewing cannot buy without approval, you need to understand who can.
14. "Can I follow up with you when we have something to show?"
Soft commitment test. A "yes" with their email is a signal. A vague "sure, whenever" is polite deflection.
15. "Do you know anyone else who deals with this problem?"
Referrals are the strongest signal. If they introduce you to a colleague or friend, they consider the problem worth solving. Also a free way to fill your interview pipeline.
Skip weeks of manual research
Get complete market research, sourced proof, competitor map, and pricing data for your idea instantly.
Questions Founders Should Never Ask
These questions feel productive but produce useless data. Every one of them invites the interviewee to tell you what you want to hear.
| Bad question | Better question | Why |
|---|---|---|
| "Would you use a tool that does X?" | "How do you handle X today?" | Hypothetical vs. actual behavior |
| "Do you think this is a good idea?" | "What have you tried so far to fix it?" | Opinion vs. evidence of action |
| "How much would you pay for this?" | "How much do you spend on this now?" | Fantasy budget vs. real spending |
| "Wouldn't it be great if...?" | "What is the hardest part of this for you?" | Leading vs. open-ended |
| "Do you have this problem?" | "Tell me about the last time you dealt with..." | Yes/no vs. narrative with details |
The pattern: bad questions ask people to predict or evaluate. Good questions ask them to describe and recall. Predictions are unreliable; descriptions are data.
How to Interpret Polite Answers That Mean Nothing
The hardest part of customer discovery is not asking the questions - it is interpreting the answers honestly. Most people are polite, and polite answers feel like validation while containing almost no usable information.
| What they say | What it means | Signal strength |
|---|---|---|
| "That sounds really cool" | They are being polite. Nothing was learned. | None |
| "I would definitely use that" | Hypothetical intent. Costs them nothing to say. | Weak |
| "Send me a link when it launches" | Mild interest. No commitment. | Low |
| "Can I sign up for the beta?" | Active interest. They are willing to invest time. | Medium |
| "Can I pay for early access?" | Real commitment. Money on the table. | Strong |
| "Let me introduce you to my colleague who handles this" | Referral = social capital spent on your idea. | Strong |
The signal strength ladder: compliment < opinion < time commitment < money commitment < referral. Only the last three count as validation.
In B2B, Do You Interview the User or the Person Who Pays?
Both, but the order matters.
Interview the end user first. The person who touches the workflow every day reveals the real pain, the workarounds, and the time lost. They usually cannot tell you about budget or approval chains, but they can tell you whether the problem is worth solving and how badly it needs fixing.
Then bring in the economic buyer. The person who signs the check tells you about budget, competing priorities, and what a purchase decision actually looks like at their company. They rarely describe the daily workflow, because they do not do it.
Interview the buyer first and you end up optimizing for what sounds good in a budget proposal. Interview only the user and you build something they love but their boss will never approve.
In B2B, there is often a third role: the internal champion. An internal champion is the person inside the target company who brings your product to the economic buyer's attention. Sometimes the user and the champion are the same person, sometimes not. Question 13 in the list above ("Who else would need to approve buying this?") maps this chain. Then go interview that person too.
What If Your Interviews Contradict Your Desk Research?
Good. It means your interviews surfaced information your desk research missed.
When interview findings and market data conflict, check two things before deciding which to trust. Is your interview sample biased? If all 12 interviewees came from one community, they may share an uncommon perspective, and you should widen the sample before overriding your desk research. And is your desk research stale? Market data ages fast, and the competitors you mapped three months ago may have pivoted or died.
I wrote a separate guide on scoring the evidence you have to decide which signals deserve your trust. The short version: if your interviews produced strong signals on the commitment ladder (money, referrals, time investment) and your desk research is directional (market size estimates, trend data), weight the interviews more heavily. Real spending on a workaround tells you more than any TAM slide.
What to Do After 10 to 15 Interviews
After your interviews, score your findings. Here is a simple tagging system:
For each interview, tag the responses:
| Tag | Meaning | Example |
|---|---|---|
| PAIN | Described the problem unprompted or with emotion | "I spend 3 hours every week manually pulling competitor data" |
| WORKAROUND | Has a current solution, even if imperfect | "I use a spreadsheet and check G2 once a month" |
| SPEND | Currently pays money or significant time | "We pay $200/month for SimilarWeb but only use 10% of it" |
| MEH | Acknowledged the problem but showed no urgency | "Yeah, it is annoying but I have bigger priorities" |
| POLITE | Generic positive response with no substance | "That sounds like a great idea" |
Decision framework:
- 8+ PAIN or SPEND tags out of 15: The problem is real and painful. Build.
- 5-7 PAIN/SPEND tags: Promising but unclear. Narrow your target audience and do 5 more interviews with a tighter definition.
- Fewer than 5 PAIN/SPEND tags: The problem is not painful enough for this audience. Pivot the audience, reframe the problem, or move to a different idea.
- Mostly POLITE tags: Your questions were too soft. Go back and ask harder questions about current behavior and spending.
Combine this with your market validation data. Desk research tells you whether the market exists. Interviews tell you whether the pain justifies building. You need both. A free idea validation scan covers the desk-research half in about 60 seconds, so your interview time goes to the half no dataset can answer.
Customer Discovery Mistakes That Waste Your Time
1. Interviewing friends and family
They will tell you what you want to hear. Always. Confirmation bias is strongest with people who care about you. Interview strangers who match your target customer profile.
2. Running the interview like a survey
A rigid list of 20 questions read in order produces shallow answers. Start with 2-3 open-ended questions. Follow the thread. The most useful insight usually comes from a follow-up question you did not plan.
3. Pitching instead of listening
If you talk more than 30% of the time, you are pitching. Customer discovery means the customer talks. You listen, probe, and take notes. Save the pitch for after you have validated the problem.
4. Stopping after 3 interviews
Three interviews is anecdotal. Ten is the floor. Patterns start emerging around interview 8-10, and stopping at 3 because the first three were encouraging means you are confirming your bias, not testing your assumptions. my startup validation guide covers how to combine interview data with desk research for a complete picture.
5. Skipping desk research before interviews
Walking into interviews without knowing your competitors, the market size, or existing demand signals means you will ask worse questions and miss follow-up opportunities. Run a free Preuve scan first, or follow my step-by-step product idea validation guide. Use the competitor and demand data to inform better interview questions.
FAQ
How many customer discovery interviews should I do?
10 to 15 interviews is the minimum for pattern recognition. After 10 interviews, you should see recurring themes. If every conversation surfaces a different problem, either your target customer definition is too broad or the problem is not painful enough. After 15 interviews with clear patterns, you have enough to make a decision.
What is the difference between customer discovery and customer validation?
Customer discovery is about understanding the problem: who has it, how they deal with it, how painful it is. Customer validation is about testing your solution: does your product solve the problem well enough for people to pay? Discovery comes first. You cannot validate a solution if you have not confirmed the problem.
Should I tell interviewees about my product idea?
Not until the last 5 minutes, if at all. The moment you pitch, the conversation shifts from honest discovery to polite encouragement. Ask about their current behavior, workarounds, and spending first. If you mention your idea, do it at the end and watch for commitment signals (asking for a beta, offering to pay) rather than generic praise.
Should I pay people for customer discovery interviews?
It depends on the audience. For B2B professionals, the 2023 User Interviews incentive report recommends $100 per hour as a baseline for remote sessions, and its data shows no-show rates falling as the incentive rises. For early-stage founders interviewing peers, a $20 coffee gift card or a promise to share findings is usually enough. The main risk of paying is not signal corruption but attracting people motivated by the payment rather than by the problem. Screen harder if you pay: ask about the problem in your recruiting message.
What if people say they love my idea but will not pay?
That is the most common outcome in customer discovery, and it is the entire point of doing interviews. Enthusiasm without commitment means the problem is not painful enough to pay for, or your positioning is off. Dig deeper: ask what they currently spend on workarounds. If the answer is nothing, the pain is not strong enough.
Should I record customer discovery interviews?
Record if the interviewee consents. Ask before starting: "Do you mind if I record so I can focus on listening?" A phone-based transcription app lets you maintain eye contact instead of typing notes. If they decline, take handwritten notes and write up full notes immediately after the call. Waiting even an hour degrades recall significantly.
Vincent
5 years in B2B growth, building Preuve AI in public. 82% of ideas it scores aren't ready, the point is finding out in 8 minutes, not 3 months.
Follow on X →
Building is expensive. Validation is free.
Run your idea through 10 AI agents before you write a line of code. Every claim source-linked.
