In 2023 I spent a weekend selling a bot that did not exist. Six people paid $30 for a one-week pilot. By Monday five were left, $150 net, 18 hours of work. $8.33 an hour against a $100 target. I shut the line and refunded.
The method exists. I do not put it first. In most niches ten deep talks and a manual concierge are enough. Selling a promise is for the moment earlier stages could not kill the hypothesis, and burying another quarter of build already hurts.
You take live money for a promise to ship in two to four weeks. In 48-72 hours you decide: build or refund. In B2C that reads as fraud even if you are honest. One miss can close a market. What you test is willingness to pay for a promise today. Retention, LTV, real CAC, the operations of a live product are not in the picture.
When you should not even sit down
Four doors. If one is closed, I do not touch the method.
Talks already happened: ten interviews, one pain in different words, you know what they use, who decides, what the budget is. The legal floor is accepted: an offer, instant refund, you can live with late-delivery penalties and a fraud statute. The limit of the method is said out loud: two payments are not a business, LTV is not visible from here. You can spend two or three days and $300-1,000 if the till stays at zero.
Without ten interviews where different people named the same pain in their own words, this is a lottery.
The legal floor for consumers. A real offer, not the first template on the web. A refund policy. A date you commit to deliver. A clause that says "if the experiment fails" may not stand in court. Late delivery penalties. Fraud statutes if you cannot show good faith. Refund on first request, at once, no theatre. One complaint or chargeback eats the weekend and more on top.
You know that two or three payments are not a business. Early buyers are often the most impatient. You do not set strategy on one weekend.
You can spend two or three days and $300-1,000 with no guarantee. If this is the last money, it is a casino.
Stop list: long enterprise and public contracts, fintech and medicine, retention subscriptions and marketplaces, hardware you cannot build in a month, a segment you have not spoken to.
Where it can still work: a short cycle, a simple digital product, a B2B pilot with no payment in 48 hours, a pitch-and-price test on an audience you already know.
Safer almost always: 15-20 interviews, preorders without a charge, a manual concierge for three to five people, no-code in a week. You go to a promise when those four gave no signal and you can carry the risk.
What the method sees, what it does not
It sees whether anyone will pay now, and which phrases kill the deal.
It does not see whether they return, what a person costs at scale, what it costs to run a live product, whether the model lives. Low retention looks like success in the first days and dies in a month. Weekend ad CAC is not CAC a quarter later. A promise is often ten times cheaper than the operation. That story is common.
People who pay in 48 hours often leave just as fast. If the whole market is those people, what follows is chaos, not a company.
Monday to Thursday
Three to five talks on one script. Five minutes on the last time the pain got them. Three minutes: what they use now and what it costs. Two minutes: they name a price for the future thing, not "something". Five minutes: what has to happen for them to buy tomorrow.
Red flags: "interesting, not now", "I will think", they cannot name a number. Three out of five like that: you do not go to the weekend.
In two or three hours I look at three to five competitors, prices, channels, bad reviews. If there are no competitors, the idea was more often buried than you are first.
Hypothesis in one cell:
[Segment] will pay [price] for [what], because [pain].
Success = [N] payments with budget ≤ $[X] and ≤ [Y] hours per client.
Cannot fill it: do not start.
Stack for an evening: a landing, a form into a sheet through Zapier or Make (a direct Sheets write fell over on me), Stripe, UTM. More than six hours on plumbing is engineering itch. Friends are counted apart: they are buying you. Manual work over an hour and a half per client kills the economics before launch.
Until Monday the promise, the price and the segment do not move. Headlines and channels can. Want to change the hypothesis: that is already a miss, and it needs a name.
Funnel in one table: view, started the form, submitted, started payment, paid, call. At each step: source, time, which text they saw, price. Card transfers with "I will send a receipt later" is already a shadow shop, not an experiment.
Every client hears the same sentence: the product is not there yet; if demand holds you get the result in two to four weeks; if not, I refund at once. People who need it "right now" leave. The rest at least know what they are buying.
Friday to Sunday
A test lead walks the whole chain, a one-dollar payment clears, the auto-reply is up. The first two hours you sit on the channel.
A 15-minute call: what hurts, what they use, what the hole costs, the offer, pay or refuse with a reason. I do not move the price on the call.
By Sunday you do not need "lots of clicks". You need five talks with people who have a budget, a catalogue of objections in their words, who decides, a pain with a number. Three payments with no idea why is worse than zero payments and five identical objections.
The right first customer is from the segment, has budget for the full product, bought the pain, not the author. Green: they name "we lose 15% of orders to slow replies" themselves, they name a budget ceiling, they ask about the limit instead of "can you add another button", they will live with a crooked first version. Red: only friends pay, "I will try at this price", a three-page requirements list, a week of silence.
Benchmarks. B2C: 200-400 clicks, 6-12 applications, 2-4 payments, the price covers ads and at least an hour of work. B2B: 30-50 contacts, 15-20 replies, 5-7 talks, 2-3 pilots. Next to that, always the alternative rate for the same hours. If it sits far above the fact, the hypothesis is already sick.
Stop: 300-500 clicks, 20-30 talks, zero payments. Only people you know. Silence. More than two hours of hands per person.
The case I am writing from
Hypothesis: a startup CTO will pay $30 for a week to close FAQ with a bot. Five talks before launch: three from my network, two from LinkedIn. I did not ask how many hours they would put in. I did not ask about the stack. I sold a Zendesk replacement, not a layer on top.
Friday 18:00, Tilda and Stripe. At 19:30 the first payment: "can we start tomorrow?" I had promised Monday, sat down for an hour and a half that evening. Saturday: a second payment, another "urgent", Google Sheets died on concurrent writes, two leads turned up in the evening in Stripe spam. By evening six payments, four of them warm. A cold one asked about Jira. I did not know the API, promised, burned three hours on Sunday, failed, refund. Four onboardings of two hours each: every FAQ structure different. I forgot to log two people, rebuilt them from email.
Monday: five still in, $150, 18 hours, $8.33 an hour. One already writing that the bot chokes on hard questions. I shut it. Refunded four with an explanation. They said thanks. One asked to hear about the next ideas.
In the recordings, one picture. All six asked about an integration. Five of six were ecommerce on Shopify or WooCommerce. The objection: "we already have Zendesk, why another box". They did not want to change the conveyor. They wanted the typical (status, return, shipping: the same 80%) answered by a bot, and the hard stuff sent to Zendesk.
New cell: a ready Shopify plugin, 20 answers written in advance, OAuth in three clicks, $99 a month. Next, five talks with shop owners, not another weekend on the same fantasy.
The weekend burned two days and showed a hole I would otherwise have spent two months building.
Two sheets I fill on Sunday before the head splits.
Hypothesis, clicks, applications, talks, cold payments apart from warm ones, revenue, ads, hours, rate per hour, three client phrases verbatim, a decision from the three shelves, a new cell if there is one.
And next to everyone who paid: own segment or not, sane expectations or magic, bought the pain or the author. If more than half bought the author, the signal is weak even at six payments. That is what I had: four of six bought the acquaintance.
Monday
Three shelves, not "it took off / it did not".
Two or three cold payments from your segment, napkin economics hold, objections are specific: 10-15 talks with people who already paid, a small MVP for 3-5 people in two to four weeks.
Zero or one payment, but "let us meet Monday" and one repeating objection: you change the cell, five talks, a rerun in a week. You do not build the product before the rerun.
Zero payments on 300-500 clicks and twenty talks, silence, economics fail even in a fairy tale: you write it down and close. Pushing the decision to Wednesday means looking for an excuse.
If you continue, week one tears off the tape: real analytics, forms and pay in code, mail without a manual send, any step longer than half an hour is automated or thrown out. Every three live hypotheses is a week or two of engineering on the plumbing. Otherwise growth will crush you with the same sheet that died on Saturday.
Napkin formula: acquisition plus your service time under 30-50% of what the client brings over a life. Does not hold: price, cost, or stop.
Most readers should not touch this method. If there is doubt, start with talks. Selling a promise is the last shelf, not a habit for every Saturday.