0
the number of future-tense questions allowed in a customer interview. "Would you use it?", "would you pay?", "would you trust it?" — all out. The only tense allowed is the past.
The question I kept asking
My first conversations with founders about vibetec went like this: I talked about the mini-accelerator for a couple of minutes and asked whether they would join. The answers were friendly. "Interesting." "I'd consider it." "You should also have a follow-on programme." I felt better after every call and knew nothing more.
The customer discovery workshop at Startup Wise Guys Founders Club gave that mistake a name. The lecturer, Marina Shutova, had participants read out their first interview question and rewrote each one with the same rule: no question in the future tense, everything into the past. I listened to the recording later and did the same to my own questions.
Why "I would use it" is not evidence
Marina's distinction is simple. Validation is what feels good. Evidence is what changes the next decision. "Cool idea" says nothing about whether the problem is frequent, expensive or urgent, or who holds the budget. "I would use it" says nothing about what the person does today, what they have tried, and what would have to happen for them to change their behaviour.
Strong evidence contains three things at once: behaviour, context and consequence. Weak: "the manual work is annoying". Strong: "every Friday two people copy data from three spreadsheets by hand, before the monthly report." The second sentence has who, how often, what and why. The first has nothing.
A future-tense question always produces the first kind of answer. People don't know what they would do. They do know what they did last time. That is why the rule is so strict.
Eight questions that got rewritten
- Would you use this? — Tell me about the last time this happened.
- Would you use an AI assistant? — When did you last have to do this by hand?
- Would you trust an AI recommendation? — Tell me how your team last decided what to prioritise.
- Would a dashboard like this help? — Tell me about the last time you had to collect evidence by hand for a report or an audit.
- Would you pay for this? — What have you paid before to solve this?
- Which features do you want? — What problem would that solve for you?
- Do you like the idea? — What is frustrating in the current process?
- You should have an integration with X. — What breaks when they are not connected?
The last line is a trap, not a question. When the respondent proposes a feature, the temptation is to write it on the list. Marina calls it the feature trap: a suggestion is not demand. Ask what breaks today, and only then do you know whether there is a problem.
The same goes for "did you know you have to…". If eight out of ten don't know about a new requirement, that is evidence they have managed without it so far. Urgency comes from the deadline, not from your product.
My own banned question
For vibetec it sounded like this: would you join a four-week programme for €5,000 where AI builds your idea into a working product? All three bans in one sentence: future tense, price and pitch.
Rewritten: tell me how your last attempt to build an MVP went. Who did it, how long it took, what broke, what you paid for it. If the person has not tried to build anything in the last six months, they are not a respondent, they are an acquaintance.
The workflow question also changes who to talk to. I was looking for people who might join. The people to look for are the ones who have already tried: with a freelancer, an agency, or Lovable on their own. They know what fails in the current solution. Those who haven't tried only know an opinion.
Don't show the product
The second rule that hit me: the interview is not about you. One participant was tempted to show the respondent the platform, because "you get more information that way". Marina's answer: you don't want to know their reaction to your product, you want to know what their reality looks like. Showing the product is a later experiment, not discovery.
A bad interview leaves you reassured. A good one leaves you informed. If you pitched for eight minutes and asked "would you use it", you got the first.
Four words when everything else slips your mind
- Workflow. How it runs today, who does it, how often.
- Alternative. What has already been tried, what was paid for it. Without it you don't understand the competition.
- Consequence. What happens when it goes wrong. Without it you inflate the urgency.
- Parties. Who uses it, who feels the pain, who pays. Without it the buying cycle goes unnoticed.
In B2B one customer is several people. The manager knows the budget but not the workflow. The junior knows the pain but not the buying process. Start with whoever owns the workflow. The budget owner is a bonus.
And at the end of every interview, one question that is in neither the past nor the future: who else should I talk to?
What I do differently now
The vibetec interview guide no longer has a single "would you…" question. I don't mention the programme until the respondent has told me about their last attempt to build. I look for respondents by the criterion "has tried to build an MVP in the last six months", not by "might be interested".
Marina's recommended minimum is three interviews, a strong result is five. Honestly: I have not done them yet. I'll write again when the first five are done and the quotes are in.