%20scilla.studio.svg)
The deal closed on a Thursday, and the whole company celebrated. A logo everyone recognised. A number big enough to matter. Six weeks of chasing, finally paid off. People high-fived in the channel, the founders bought a round, and not one of them noticed it was the day the company stopped becoming a product company.
This company does not exist, or it exists many times over: a consultancy that got good at building software, raised money to stop selling its hours and start selling a product, then kept selling its hours. Here is what I think. For a consultancy trying to become a product company, the most dangerous thing it can do is close deals that need custom engineering.
The company raised on one hypothesis: that it could take a single product, aimed at one kind of customer, solving one clear problem, and prove that enough of those customers exist, will buy it, and will keep buying it. That is the whole bet. The runway, call it nine months, is not really cash in a bank. It is nine months of work from the few people who can build and test the product, a handful of engineers and a product lead, with no second team in reserve.
A consultancy that just raised money has a problem that does not look like a problem: it already knows how to make money. The sales motion is muscle memory. Collect leads, take the meeting, listen for the opening, offer whatever the customer seems to want. It works, and always has. It makes payroll this quarter, not in some hypothetical next year. Against a product bet that might not pay off for a year, the deal in front of you always looks better.
It shows up in two places, the selling and the building. In the selling, a good rep listens to each prospect and tailors the offer to what they hear. It looks like discovery. In practice the company shows a slightly different product to every buyer, and never tests the same one twice. Steve Blank defined a startup as an organisation searching for a repeatable, scalable business model. A custom pitch for every prospect is neither, so nothing it closes proves the model exists. In the building, the promise the rep made to win becomes the roadmap the engineers now owe one client.
Each custom build spends, say, a month of engineering, and it comes out of the exact people whose job was to prove the product works. So the deal that looked like traction delivered a one-off feature to one client, paid for with the only work that could have proven the product. The team is fully booked, and no closer to knowing whether the product works. A pipeline full of custom deals looks busy and proves nothing.
Burning money is the plan: a funded startup is supposed to lose money every month, and that is what the raise was for. So the board is not worried about the loss. They want to know what each month of it buys. Custom revenue does not answer that. It leaves the loss the same size and only changes what the money bought: a month of custom delivery buys a happy client and nothing the next raise is priced on.
So what is the next raise priced on? Retained, recurring revenue from the customer the company was aimed at. The repeatable, scalable model again, the thing a custom pitch can never be. David Sacks gave this a number, the burn multiple: how much cash you burn to add one dollar of new recurring revenue. Burn two million to add a million, and your burn multiple is 2. Burn five for the same million, and it is 5. Lower is better: a low number means the market is pulling the product out of your hands, a high one means you are pushing it onto the market. Investors keep rough bars in their heads. Around 2 is normal for an early company, under 1.5 is what a Series A wants to see, under 1 is excellent. And before they run that math, serious investors strip one-off services and custom-integration revenue out of the recurring-revenue figure. So the revenue this essay keeps warning you about is gone before the number is even calculated. It counts as zero.
The last trap is the subtle one. Say the company gets disciplined about recurring revenue and starts signing monthly contracts. Is that the signal? Not by itself. A consulting retainer recurs too. The old business, wrapped in a subscription, bills every month and shows up on the board slide as "recurring revenue" while staying a services firm underneath. The signal is narrower: recurring revenue for the standard product, bought as it ships, by customers who would not need it rebuilt for them first. One number tells the two apart. Product software runs gross margins around 70 to 80 percent. A services business runs closer to 30 to 50, because every dollar of it takes human hours to earn. If your recurring revenue carries services margins, you already know what it is.
These deals are how we learn, the reps say. Every client conversation is a signal about what the market wants, and we are getting paid to collect it. Why would you turn that down?
Because a sales rep collects signal for a different reason than a product team, and the two come apart where it matters. A rep is trained, correctly for the job, to listen for the opening: the objection to handle, the worry to soothe, the one feature that turns a maybe into a yes. That is signal in the service of persuasion: it tells you how to change the offer to win this deal. What it cannot tell you is whether the market would buy your standard product unchanged. Those are different questions, and the sales conversation is built to answer the first.
The test: after fifty of these conversations, can you name the one product all fifty buyers would have bought as it shipped, with no custom work? If the answer is no, each needed something a little different, then you were collecting persuasion signal, and all it confirmed was that you should keep customising.
So, one question for every deal: is this sale worth a month of our one shot at the product? Price it in engineering months, the only budget that can prove the bet, and most custom deals fail on the spot. The ones that pass move every future customer toward the same standard product.
Two numbers keep you honest, because a services business cannot fake either one for long. The burn multiple asks whether the market is pulling or you are pushing. Net revenue retention, whether the revenue from customers you already have grows or shrinks, tells you if the ones who bought will keep paying. Under about 110 percent, something is wrong. You can see where healthy retention sits at benchmark.scilla.studio before you talk yourself into believing yours is fine. Both numbers only count the standard product, bought by the customer you aimed at. Neither one counts the heroic custom deal you closed on a Thursday.
And that company? It spends month nine the way it spent the first eight. Busy, booked, and looking like a success, with a pipeline any consultancy would envy and not one piece of the evidence the next raise needed. None of the deals looked like a mistake. That is what made them one.
