Everyone tells you to validate. Almost nobody tells you what a "yes" is worth. It is worth close to nothing, and you can compute exactly how close.
What A Yes Is Actually Worth
Model it with Bayes. Say 2 percent of people in your segment would genuinely pay for what you are building — that is your base rate before any evidence. Now ask what each signal does to that number, given how often a real buyer produces it versus how often a non-buyer produces it anyway:
— "Would you use this?" answered yes: P(real buyer) goes from 2 percent to 2.4 percent — "Would you pay CHF 50 a month?" answered yes: 3.5 percent — They give you their email: 6.1 percent — They have already built a spreadsheet hack for the problem: 13.3 percent — They book a 30-minute call: 15.2 percent — They pay a CHF 50 deposit: 55.0 percent — They sign a letter of intent: 69.2 percent
Read the first line again. A verbal yes moves you from 2 percent to 2.4 percent — a lift of 1.18 times. You learned essentially nothing, and you walked out of the meeting energised. That gap between what you felt and what you learned is where months disappear.
The reason is not that people lie. It is that saying yes costs nothing, so non-buyers say it almost as often as buyers. A signal only carries information when it is expensive for the wrong person to produce.
Twenty-Four Yeses Equal One Deposit
Put it in one comparison. The likelihood ratio of a verbal yes is about 1.19. The likelihood ratio of a CHF 50 deposit is about 60. Since independent evidence multiplies, you would need roughly 24 independent verbal yeses to move your belief as much as a single deposit does.
Twenty-four conversations. One payment. Same information.
And notice where the good signals cluster: they all involve someone spending something. Money, calendar time, or the effort they already spent building a workaround before you arrived. Opinions are free, so they are worthless. Costs are informative, so they are evidence.
Building Is The Expensive Way To Ask
Founders skip this and build, on the theory that people cannot judge a description. Price both approaches in the only currency you actually have — your time — at roughly CHF 12,000 a month of founder cost:
— Building for 1 month: about CHF 12,000 — Building for 3 months: about CHF 36,000 — Building for 6 months: about CHF 72,000
Against:
— 30 conversations at 30 minutes each: 15 hours, about CHF 1,125 of the same time — 50 conversations: 25 hours, about CHF 1,875
So a three-month MVP costs roughly 32 times what fifty conversations cost — and the MVP frequently produces the same signal, because "I built it and they did not use it" and "I described it and they did not care" are the same finding at very different prices.
A deposit test costs even less: a landing page, a payment link, and a couple of hundred targeted visits. At a 2 percent deposit rate, 200 visits produces 4 deposits. Four people who paid tells you more than four hundred who nodded.
Ask About The Past, Not The Future
This is why the question matters more than the count. Future intent is imagination, and people are bad at it and kind about it. Past behaviour already happened, and it is not negotiable:
— Not "would you use this" but "what did you do the last time this happened" — Not "would you pay" but "what are you paying for today, and how much" — Not "is this a problem" but "show me the spreadsheet or the process you built to cope"
Someone who has already jury-rigged a solution has demonstrated the problem is worth effort. That is the 13.3 percent row above, and it costs you nothing but a better question.
The Swiss Wrinkle
There is a structural reason to distrust encouragement here specifically. Switzerland is a small market with dense networks — the people you can reach early are largely people who know you, or know someone who knows you. That is not a comment on anyone's character; it is arithmetic about who takes your call.
The consequence is that early Swiss feedback is systematically kinder than the market will be, because it is coming from a sample selected for relationship rather than need. The fix is not cynicism. It is insisting on signals that cost the other person something, so that goodwill cannot masquerade as demand.
The Catch
Every number above depends on the assumed base rate and likelihood ratios, which I invented for illustration. Change the base rate to 10 percent and every posterior rises; change the yes-from-non-buyers rate and the whole table shifts. The ordering is robust — costly signals dominate cheap ones — but the specific percentages are a model, not a measurement of your market.
The independence assumption is also generous. Twenty-four yeses from the same friendly network are not 24 independent signals; they are closer to one signal repeated, which makes the verbal-yes column even weaker than shown.
And a deposit is not proof. People do change their minds, deposits get refunded, and some genuinely great businesses could not have collected one early. The claim is narrow: costly signals update your belief far more than free ones, and you should spend your scarce months collecting the former.
Do This
— Before your next conversation, write the question you will ask about the past. If every question is about the future, you are running a poll rather than a test. — Ask for something that costs them: a deposit, a signed intent, a calendar slot, an introduction. Then note whether they pay it. That answer is the data. — Price your next build in months of founder time, then ask what the cheapest experiment is that would produce the same finding. If the answer is a landing page, build the landing page.
This is education, not financial advice. It's general information, not a recommendation for your situation, and it doesn't account for your circumstances. Rules and figures change — verify before you act, and speak to a qualified professional for decisions that matter.
Go deeper at Agent Smith
FNZ keeps this track short and practical on purpose. For the full treatment, see the platform, for running validation without hiring a department.