Halvor bottles cold brew for office kitchens in Ghent and sells a 2 litre bottle for 11 euro. The beans in it cost 3.10, the bottle and cap 0.80, the label 0.15, and its share of the delivery round 1.60. That is 5.65 of direct cost, so 5.35 of every bottle is left over, which on 2026 costs is a gross margin of 49 per cent. Gross margin is price minus direct cost, divided by price, and until that number exists there is nothing to plan with.
The second division is the one that decides whether the business exists. His fixed costs run whether or not anybody orders: 640 a month for time in a shared kitchen, 45 for insurance, 310 for the van, which is 995. Break even in units is fixed costs divided by the margin on one unit, so 995 divided by 5.35 is 186 bottles a month. He sells to seven offices taking four bottles a week, which is 126 bottles a month. He is not close, and before he did the division he thought he was nearly there, because the money coming in was going up every month.
The line people leave out is the third one in the first paragraph. Delivery is a direct cost, and leaving it out is the usual error: it feels like something the van is for rather than something the bottle is for, and 1.60 a bottle is 30 per cent of the margin as it stands in 2026. A model built without it shows a break even around 130 bottles, which is roughly where Halvor is, which is exactly why the mistake is so hard to notice from the inside.
The van is a decision, and it moves the finish line
Of the three fixed costs, one is a signature rather than a fact. The kitchen time is what it is and the insurance is not optional, but the van is leased at 310 a month and the alternative is hiring one for the two delivery days, at roughly 180. That takes fixed costs to 865 and break even to 162 bottles, which is 24 bottles closer, more than a whole office. A fixed cost you can turn off is not the same risk as one you signed for, and the difference is worth knowing before the signature rather than during a slow quarter.
Neither number is a forecast and neither is a target for a good month. They are the point at which the business stops eating money, which is the least ambitious question available and the only one that has to be answered first. Getting from 126 bottles to 198 bottles is four more offices on a round that already passes their doors, which is a sales problem with a known size, and a known size is what all of this arithmetic is for.
The order
Six stops in order, and not one of them carries a date
The order is real. The timing is not, and anybody who tells you the second stop takes a quarter is describing their own business rather than yours.
- A problem somebody already pays to work aroundNot a product. A cost being absorbed today01
- 02Ten conversations, and the count of the noesBefore anything is built, and with the reasons written down
- One unit, priced, with its costs listedMargin per unit, or there is nothing to divide03
- 04The first ten who payMoney, not encouragement, and in the room if possible
- The admin that makes it a businessRegistration, insurance, terms of sale, records05
- 06The part that stops being youWhere the hours stop being what is sold
No dates, on purpose. The stops are ordered by what each one makes possible, and every published timeline for this is one person's own history sold as a schedule.
The stall is building it before anybody has said they will pay
Here is where people stop, and it is comfortable, which is the danger. Building can be done alone, at a desk, at whatever hour suits, and every day of it produces something visible. Asking ten people to buy cannot be done alone and produces, on most days, nothing but noes. So the label gets designed, the site gets built, the recipe gets refined for a fourth time, and the question of whether anybody will hand over 11 euro stays open for a year.
The test is a receipt. A customer who has not paid has not validated anything, and the reason is not cynicism about people, it is that saying a thing sounds great costs nothing and is the polite answer. Halvor's first six conversations were warm and produced no orders; his seventh was lukewarm and produced a standing order, and the difference was that the seventh was asked for money in the room.
The other half of the test is the one nobody keeps. Count the ones who said no, and why. Four of Halvor's first ten refusals were about the price and six were about the delivery day, and those are opposite problems with opposite fixes. Without the count, ten noes are a mood; with it, they are the most specific piece of market research available and it cost nothing to collect.
What to open in 2026
A one page unit sheet in Google Sheets or Excel, both free in a browser, is the tool this whole page is about, and it replaced the number in your head, which is always the best case and never includes the delivery. One row per cost, one cell for the price, two formulas underneath. The reason to keep it in a sheet rather than a note is that the beans change price and the answer has to change with them, which is a thing a formula does and a memory does not.
Stripe payment links, or a card reader for the ones done standing up, replaced the bank transfer and the promise attached to it. There is no monthly fee on a link and it takes a cut of each payment, which is a direct cost and belongs in the 5.65 rather than in a complaint. What makes it worth setting up in week one rather than month six is that it turns the validation test above into something that can happen in the room, while the person is still interested.
A model drafts the customer conversation and the terms of sale now, which is the change since 2023, and it replaced the blank page in front of both. Every major assistant covers it on a free tier. What it cannot do is have the conversations, and it is worth being blunt about the failure mode: asked to help validate an idea it will produce an excellent list of questions and a plausible summary of a market, and none of that is a person handing over money.
The half nobody photographs
The business a customer sees, and the rest of it
The proportion is the claim. Two rows are the business as it is described at a party, and the other five are the ones that decide whether it is still there next year.
- A bottle on the office counterThe whole of what is being bought
- The label, and the name on itThe part everybody calls the business
- The margin on one bottleEverything under here is divided by it
- Delivery, which is a direct costThe line most models leave out
- Fixed costs that run whether or not anybody ordersThree lines, one of which is a signature
- Registration, insurance and terms of saleArrives as a deadline rather than as a choice
- The office that pays in sixty daysProfitable and out of money in the same month
The five rows below the line are not obstacles to the business, they are the business. The two above it are the only part anybody enjoys describing.
Six clusters, and the order is real while the timing is not
Finding one worth the years is where the problem gets separated from the product, and it is the shortest cluster and the one people spend longest on. How money comes in, and what it costs to make is the two divisions above, and it is the cluster that decides whether the rest is worth reading. The hardest ten is the stall, named and worked.
The finance a small business actually needs is smaller than it sounds and mostly concerns one fact, which is that profitable and out of money is a single ordinary month when a customer pays in sixty days. The unglamorous half is the one nobody photographs and the one that arrives as a deadline rather than a choice. Getting past the point you are the bottleneck is where this topic hands over to getting work done through other people, and the handover is abrupt: the skills that got Halvor to 186 bottles are not the ones that get anybody past his own two delivery days.
Somebody selling their own hours rather than a thing that keeps selling is reading selling your own skill, directly, where the same arithmetic runs on a rate instead of a bottle. The question of who to sell to before any of it is spent belongs to the decisions that come before any channel. All of it sits under running work, getting paid, and the software it happens in.
There is a cheaper move than four more offices, and no cluster below will make it for him. Going from 11 to 13 euro takes the margin from 5.35 to 7.35 and break even from 186 bottles to 136, which is close enough to reach this quarter, and it is a smaller change than finding four offices. Whether the seven he has would pay it is a fact about them that no page can supply, and the only way to find out costs him the answer.