Part VI · Chapter 14 · Founding · lesson 2

What people say and what they pay for

11 min

The conversation went brilliantly. She said it was a real problem, that the current tools are terrible, that she would definitely use something like this, and that you should send her a link when it is ready.

You have learned nothing. Not almost nothing — nothing. Every sentence in that paragraph is compatible with her never spending a unit of currency with you, and none of it cost her anything to say.

People are accurate about their past and unreliable about their future.

Ask what someone will do and you get a prediction made by a person who wants the conversation to go well, about a version of themselves who is more organised than the one who will actually be there.

Ask what they did — last month, last year, the last time this went wrong — and you get a fact. It happened, it can be checked, and it was decided by the real person under real conditions.

Why the pleasant conversation is the dangerous one

Three things are working against you in the room, and none of them involve anyone lying.

Politeness. Telling someone their plan is poor is unpleasant, and most people will decline to do it, especially to someone visibly invested. The cost of enthusiasm is zero and the cost of honesty is a difficult ten minutes.

Hypothetical distance. A future purchase is not a purchase. It is a pleasant thought, unattached to the budget, the approval, the switching cost or the other nine things competing for the same money.

Your framing. The question you ask contains the answer you want, and listeners cooperate. Describe your solution first and everything afterwards is a reaction to it rather than a description of their situation.

Predict

Which of these questions produces the most reliable information?

Would you use a tool that did this?How much would you pay for something like this?Walk me through the last time this went wrong. What did you do?What features would you want it to have?

The five questions that produce facts

Each of these points at something that already happened.

What do you do about this today? Establishes that the problem exists and that they are doing something about it. If the answer is nothing, you are looking at an irritation rather than a problem.

When did it last happen? Frequency, without asking for a rate. Last week and last March are different businesses.

What have you already tried? The graveyard of previous attempts tells you what the alternatives are and why they failed, which is the specification for your version.

What did that cost — in money, or in someone's hours? This is the ability-to-pay filter, answered with a historical number rather than a hypothetical one.

Who decided to spend that? Names the buyer. Very often it is not the person in front of you, and finding that out in month one rather than month nine is worth the entire conversation.

Not the same thing

Stated preference

What someone says they want, would use, or would pay, when asked.

Measured in: Enthusiasm — nods, adjectives, follow-up emails
For example: Nine people out of ten in a room saying they would find this useful.
Chase it alone and: It correlates with how likeable you are and how well you described the idea, both of which you control, and neither of which is a fact about the market.

Revealed preference

What someone has actually already given up — money, hours, attention — to deal with this problem.

Measured in: Currency spent, hours logged, tools bought, people hired
For example: A company paying someone two days a month to reconcile this by hand.
Chase it alone and: It is backward-looking, so a genuinely new category has no history to reveal, and you can miss a real market by insisting on evidence that cannot exist yet.

Which is which? Put each one on a side.

Eight people signed up to a waiting list
A department has been paying a contractor to patch this every quarter for two years
Someone said they would pay 50 a month for it
Someone built their own version in a spreadsheet and maintains it themselves

Something people say

“Validate the idea by asking enough people whether they would use it.”

Why it spreads: It is genuinely better than asking nobody, it produces a number that can be put in a document, and it usually returns an encouraging result — which makes it feel like it worked.
The part that is true: Talking to the people who would use the thing is exactly right, and skipping it is a far more common and more expensive mistake than doing it badly.

The error is in the question rather than the activity. A yes from someone who has risked nothing is not validation; it is a courtesy, and collecting a hundred of them produces a hundred courtesies.

The repair is small: change every question from the future tense to the past tense, and stop describing the solution until the end. What you want out of the conversation is a story about something that already happened, with numbers and names in it.

The strongest version of validation is not a question at all. It is an ask — for a payment, a deposit, a commitment of someone's time, an introduction to the person who holds the budget. Anything that costs the other party something converts an opinion into evidence.

Talking to the wrong people

Two sampling errors do most of the damage, and both feel like diligence.

Talking to people who like you. Friends, former colleagues, anyone who wants you to succeed. They are the easiest to reach and the least able to tell you anything, because their answer is partly about you.

Talking only to people who already have the problem badly. If you find the ten most affected companies and they all agree, you have learned that ten unusual companies are unusual. The question you needed answered was how common they are, and the ten cannot tell you.

Check

After twenty conversations, everyone agrees the problem is real and nobody has committed anything. What is the most likely reading?

The problem is real and sits below the line where anyone acts on itThe idea needs to be explained betterTwenty is too small a sampleThe wrong people were asked

Worth remembering

People are accurate about their past and unreliable about their future, so every useful question is in the past tense.
Enthusiasm costs the respondent nothing, which is exactly why it carries no information.
The five questions worth asking are what they do today, when it last happened, what they already tried, what that cost, and who authorised it.
Revealed preference means something scarce was given up — money, hours, a hand-built workaround.
Agreement from twenty people with commitment from none is a finding, not a communication failure.