“Would you use this?” is the wrong question

You have half an hour with somebody who might be a customer. What you ask in that half hour decides whether you walk away with evidence or with a pleasant feeling, and the difference between the two is almost entirely a matter of which direction your questions point.

Questions that point forward ask a person to predict themselves. Questions that point backward ask them to report something that already happened. Same person, same thirty minutes, completely different quality of answer.

Two questions, five minutes, different answers

Put these side by side, because the contrast is the whole article.

“Would you use this?” To answer, somebody has to imagine a future version of themselves encountering your product in circumstances neither of you can specify. They will imagine that person generously, because people imagine their future selves as more organised and more decisive than their present ones. They will also want to be encouraging, because you are sitting there having clearly worked hard. The answer will be warm and it will not predict anything.

“What are you using for this now?” To answer, somebody has to remember. The decision already happened, with real money or real time behind it, and it happened before you turned up, so there is nothing to be kind about. If the answer is a spreadsheet, a colleague, or nothing at all, you have just learned what you are genuinely competing with, which is frequently not what you assumed.

One question asks for a forecast. One asks for a fact. Only the second one survives being checked.

Why a prediction about yourself is not evidence

It is worth being clear that this is not about people being unreliable narrators in some general sense. The failure is specific.

When somebody predicts their own behaviour, they are reasoning about intention, and intention is real. What intention leaves out is everything that actually decides the outcome: the competing priority that turns up that week, the existing tool that is already installed and already understood, the switching cost that looks trivial in the abstract and turns out to be an afternoon, and the fact that most problems people describe as significant are ones they have been living with comfortably for years.

None of that is visible from inside the imagined moment. So the prediction is offered honestly and is missing the variables that matter.

There is a second problem stacked on top. You are not a neutral interviewer. You built the thing, the other person knows it, and the social cost of telling a builder their idea is not for you is high enough that most people will not pay it. This does not make them dishonest. It makes them polite, which is a thing you would also be.

The questions that point backward

Here is the practical set. All of them ask about something that has already happened.

What are you using for this today? Establishes the real alternative, including the very common answer of nothing.

Walk me through the last time this came up. A specific instance with a date. Frequency claims without an instance are usually estimates, and estimates of frequency run high.

What did you do about it? The workaround tells you what the problem is worth. Somebody who built a spreadsheet has told you more than somebody who says it is a major pain point.

What have you tried before, and why did you stop? Abandoned tools are the most candid part of any interview. People remember precisely what made them quit.

Have you paid for anything in this area? What, and how much? Past spending is the strongest available proxy for future spending, and unlike willingness to pay it is a fact.

Who else was involved in that decision? Frequently the person in front of you is not the person who chooses, and finding that out early saves months.

What to stop asking

A short list, and every item on it is a question that feels productive.

Would you use this? Covered above. It is the default question and it is the weakest one available.

Would you pay for this? Worse, because the yes feels like commercial validation. Nobody has spent anything by answering it.

How much would you pay? This asks somebody to price a thing they have not used, in a category they may not know, with no reference points. The number that comes back is not information.

Do you think this is a good idea? This asks for an opinion about you.

What features would you want? Reasonable much later, once somebody is using the thing. Early, it invites a person with no context to design your product, and the answers will be a list of things they have seen elsewhere.

Reading what you get back

Two habits make the difference between notes and evidence.

Separate what they did from what they said. Write those in two columns. The did column is your evidence. The said column is context, and it is worth keeping, but it does not get a vote on its own.

Count how many people you spoke to, and notice who they were. Six conversations, all with people who already know you, is six data points from one corner of the market. That is not nothing, and it is not a market read either. We wrote about that sampling problem separately, because it is the most common way a diligent founder ends up confidently wrong.

If you can run these calls, run them. A founder with access to ten real potential buyers and the discipline to ask backward questions will learn more in a week than any tool can tell them, and we would rather say that than pretend otherwise. The problem MarkLens actually solves is the situation before that: no audience to call, no list to email, and a need to read what the market is doing from the outside. When you do have people to ask, ask them. Just ask them what happened, not what they think will.