A customer just told you "yes, I would absolutely use a tool like that". You leave the call feeling validated. Three months later your launched product gets one signup from that person and zero usage. Customer interview questions that lead the witness are the most common reason founders ship the wrong thing. We have run hundreds of customer interviews across Carriva, the lesson-planning monorepo, and earlier products. Here is the script that works, the questions to delete from your current list, and the warning signs that you are leading instead of listening.
The principle
The customer's job in an interview is not to tell you whether your idea is good. It is to tell you what they actually do today and what makes that hard.
If your interview script asks them to evaluate your idea, you have already lost. They will be polite. They will give you encouragement. None of it predicts behavior.
The whole script that follows is built around extracting facts, not opinions. Past behavior, current pain, real workarounds. Not "would you", "could you", "do you think".
The bad questions to delete first
Open your current interview script. Delete these questions, in order:
"Would you use a tool that did X?"
The worst question in customer development. Hypothetical, leading, polite-answer-friendly. The customer says yes because that is socially easier than saying no to your face.
"Do you think this is a good idea?"
You are asking them to evaluate. They are not qualified. They have not used it. Their opinion is folklore.
"How much would you pay for this?"
People lie about price. Always. They quote the price they want to be the kind of person who pays. Real spending decisions involve a procurement process you cannot simulate in a 30-minute call.
"On a scale of 1 to 10, how interested are you?"
The 1-to-10 scale is uniformly biased upward. People say 7. People who say 7 do not buy.
"What features would you like to see?"
Customers do not know what features they want. They know what problems they have. Asking for feature requests gets you a wishlist of things they would never use even if you built them.
If your script has any of these, delete them now. Replace with the questions below.
The questions that work
A small set, ordered by the moment in the conversation.
Opening: "What did you do today around X?"
Where X is the problem area. Not the solution area. If you are building Carriva, X is "auditing pension records", not "using audit software".
The answer is concrete. Today is fact. They walk you through the day, the steps, the tools, the people they talked to. You are gathering raw material.
If they say "I did not do X today", ask "when was the last time you did X?" and follow that day. If they have not done X recently, you may be talking to the wrong person.
Probing: "Walk me through the last time you did Y."
Y is one of the steps they mentioned. Pick the step you suspect is painful. Ask for the specific most-recent instance. Not the general story. The specific one.
The specific story has details. The general story has rationalizations. Always go for the specific.
Pain extraction: "What was the worst part?" / "What is the most annoying step?"
Open-ended. Their answer tells you where the friction is. If they say "honestly, nothing, it was fine", you may not have a problem to solve. If they say "ugh, when I have to copy data from spreadsheet A to spreadsheet B, it is the worst", you have a candidate problem.
Note that "the worst part" is a real signal. It is hyperbolic in their voice. They will not say "the worst" about something that is not actually painful. People hedge their language unless they care.
Workaround inventory: "How do you handle that today?"
Once you have the pain, ask what they do about it. The answer is one of:
- "Nothing, I just put up with it." (Maybe a real problem, but the customer has no urgency. Watch out.)
- "I have a workaround: a spreadsheet, a script, a person on the team." (Strong signal. They are spending energy on this.)
- "I tried tool X but it did not work because Y." (Even stronger. They have evaluated solutions and found them lacking.)
Number 2 and 3 are gold. Number 1 is a yellow flag. If everyone you interview says "nothing, I just put up with it", your problem may not be valuable enough to pay to solve.
Frequency: "How often does this happen?" / "When does this come up?"
Frequency tells you how big the pain budget is. Once a year is a yellow flag. Daily is green. Weekly is fine. The frequency multiplied by the per-instance pain is roughly your pricing ceiling.
Money question: "Have you spent money to solve this?"
Not "would you", which is hypothetical. "Have you", which is fact.
Answers:
- "Yes, we use [tool] and pay [amount]." (Strong signal. They have a budget. They might switch.)
- "We tried [tool] but it was not worth it." (Strong signal. They have a clear bar.)
- "No, we have not invested in this." (Yellow flag. The problem may not have budget.)
If they have spent money, they will spend more. If they have not, getting them to spend is a longer sales cycle than you probably want.
The closing question: "Who else has this problem?"
For two reasons. First, you want referrals (more interviews). Second, the answer signals whether the customer thinks of this as a personal pain or as an industry pain.
"Just me, my workflow is unusual" → niche of one, hard to build a SaaS for. "Everyone in my role at any company in this space" → broad pain, real market.
The cues you are leading
While you are interviewing, watch for the signals that you are biasing the conversation.
You are pitching
The moment you say "and we are building a tool that does X", you have stopped interviewing. The customer's answers are now colored by what you said. Save the pitch for the end of the call, after the interview is done. Or skip it entirely; this is research, not sales.
You are asking yes/no questions
"Would you use a tool that did X?" gets a yes or a no. Both are useless. Open-ended questions get stories. Stories have signal.
You are filling silences
A pause in the conversation feels uncomfortable. You jump in with a clarification or another question. The customer was about to say something honest. You interrupted them.
Train yourself to wait 4 to 6 seconds after they finish speaking. The most useful answers we have ever gotten came after a long pause where the customer thought about what they actually meant.
You are nodding too much
Nodding signals "yes, that is what I want to hear". The customer escalates to give you more of the response you appear to want. Stay neutral. Take notes. Do not react to specific answers.
Mom Test, briefly
The classic reference here is The Mom Test by Rob Fitzpatrick. It is the single best book on customer development. The thesis is that customer development is the practice of asking questions whose answers you cannot game by being polite. Ask about facts, not opinions. Past behavior, not future intention. Specifics, not generalities.
We have applied this method consistently. We covered our specific application in our Mom Test customer development writeup, with the script we use for Carriva discovery interviews.
The script we actually use
This is the one we run for B2B discovery interviews. About 30 minutes.
- Background (3 min): "Tell me about your role and your day-to-day."
- Today's work (5 min): "What did you do today around [problem area]?"
- Specific recent example (10 min): "Walk me through the last time you did [task]."
- Pain points (5 min): "What is the most annoying part of that?"
- Workarounds (5 min): "How do you handle that today?"
- Money (2 min): "Have you spent money to solve this?"
- Referral / wrap-up (1 min): "Who else has this problem?" + closing.
No pitch. No "what would you like in a tool". No "would you pay for X". The whole script is fact-extraction.
What to write down during the call
A two-column note format:
| Quote | Implication |
|---|---|
| "I export to Excel because the system does not let me filter the way I need." | Pain: filtering. Workaround: Excel. Strong signal. |
| "It is fine, we do it once a quarter." | Frequency: low. Yellow flag. |
| "I tried [tool X] but it was clunky." | They have evaluated. They have a bar. |
The quotes go on the left. Your interpretation goes on the right. Separating the two is critical because your interpretation will shift over time as you understand the domain better. The verbatim quote does not.
After 5 to 10 interviews, you will see patterns in the quotes column that you can act on. Patterns in the interpretation column are mostly your own bias.
When the customer leads you
Sometimes the customer wants to talk about something other than what you came to ask about. That is the best signal of all.
If a customer keeps redirecting to a different problem ("yes, but actually the worst thing about my job is..."), pay attention. Their attention is the truth. Your interview script is your best guess.
We pivoted Carriva once based on three interviews where the customer kept redirecting from "audit accuracy" (what we were probing) to "onboarding new advisors" (what they actually cared about). We added the onboarding feature. It became one of our strongest selling points.
The customer who derails your interview is doing your discovery work for you. Follow them.
How many interviews
There is no magic number, but a rule of thumb:
- 5 interviews: you start to see your assumptions wobble.
- 10 interviews: you see clear patterns.
- 20 interviews: you have a real picture of the problem space.
- 50 interviews: you understand the domain at a level that lets you build with confidence.
We did 30 interviews before launching Carriva, 50 by the end of the first six months. Each interview cost us 30 to 60 minutes. The cumulative cost was real. The cumulative learning was decisive.
If you have done fewer than 10 customer interviews and you are about to ship a product, slow down. Do another 10 first. The week of delay will save you a quarter of building the wrong thing.
How to find customers to interview
A separate practical concern. Real channels we use:
- Cold email or LinkedIn message to people in the role you are targeting. Be honest: you are doing research, not selling.
- Existing networks. The first 5 should be people who already know you and who you can ask candidly to be candid.
- Domain communities. Subreddits, Discord servers, professional associations.
Do not interview your friends without warning them about the bias. Friends are kinder than the market.
We documented the broader sourcing pipeline in our why we built Carriva writeup; the discovery process that led to Carriva was almost entirely interview-driven.
Connecting interviews to product decisions
Interviews are inputs. The decisions are yours. Do not pretend the customer told you what to build. The customer told you what is hard. You decide which hard problem you can build a sustainable product around.
A specific real example. We did 20 interviews early in the lesson-planning project. The dominant pain was "I spend Sunday night planning lessons". We could have built any of: a planning tool, a lesson library, a curriculum-mapping tool, a peer-collaboration tool. We picked a planning tool because it best matched our team's strengths. The customer did not tell us to pick that. They told us where the pain was.
This is also relevant to picking a niche in a saturated market. We covered that strategy in our niche SaaS in saturated markets writeup. Interviews tell you where the pain lives. Strategy tells you which slice of the pain is worth your time.
What we would test first
If you are sitting with a customer interview script right now:
- Delete the words "would", "could", "should", and "do you think" from your script. Replace with "did", "do", "have", "tell me about".
- Add three "walk me through the last time" questions. Specifics over generalities.
- Add a money question about past spending, not future hypothetical.
- Plan for 4 seconds of silence after every question. Do not fill it.
- Run 5 interviews. Read your notes back. The patterns will surface.
Customer interview questions are not magic. They are discipline. The discipline is asking about reality, not about your idea.
TL;DR
Past behavior over future intention. Specifics over generalities. Money spent over money offered. Open-ended over yes-no. Save the pitch for after the interview is done. Customer interview questions framed correctly are the cheapest research method any founder has access to. The cheapness is a trap; you have to do the discipline to get the value.
