What your build actually cost you.

The invoice says £180,000. Three phases, thirty pages of scope, a statement of work with a signature block at the end. Everyone in the room reads it as a cost, weighs it against the cash in the bank, decides it is affordable, and signs. It is a good build partner and a fair price for the work described. Nobody in the room does the one piece of arithmetic that would change the conversation, because the invoice is written in the wrong unit.

Pounds are the wrong unit because pounds are replaceable. You can raise more of them. The thing you are actually spending on a pre-seed build is not the cash, it is the share of the company that cash represents, and that share does not come back. So before you sign anything, convert it.

Do the arithmetic in public.

Here is the working, so you can redo it with your own numbers rather than trust mine. You raised £500,000 on a £2m pre-money. Add the cheque and you have a £2.5m post-money. One per cent of the company is therefore £25,000. The £180,000 build is 180 divided by 25, which is 7.2% of the company. Not a line item. Not overhead. Nearly a fourteenth of everything you will ever own, handed over in exchange for a list of screens.

Say it back to yourself in the unit that matters. The question was never can we afford £180,000. The question is would we sell 7.2% of this company for exactly what is written in this statement of work. That is a different question, and it is the honest one. Most founders would haggle for a week over 7.2% of the cap table in a term sheet, then sign it away on a build without converting the figure once.

Why per cent is the honest unit.

Cash you spend is gone but replaceable. The percentage you spend is gone and permanent: it stays sold at every valuation the company ever reaches after it. That is the part the invoice hides. What looks like a £180,000 decision today is a 7.2% slice you are carrying forward for the life of the business. Carry it to a modest Series A at a £20m valuation and that same 7.2% is worth £1.44m. You did not spend £180,000 on the build. You spent the £180,000 and every future pound that slice would have been worth to you.

None of this is an argument against spending equity on a build. Sometimes 7.2% for the right product is the best trade you will ever make. It is an argument for making the trade with your eyes open, in the unit that does not lie to you.

Where the per cent actually goes.

Now the harder part: where the slice leaks. It is almost never the feature you think. On one engagement I watched, the scope had a question in it that got answered both. In month one someone asked whether the product was for one kind of customer or another, and rather than close it, the plan quietly served both. For now. The words "for now" did the damage, because "for now" was never revisited and never closed.

Eleven months later the build was a set of screens trying to be two products at once, because every decision downstream of that open question had to hedge. Then reality arrived, one of the two audiences was chosen, and nine weeks of rework landed, correctly and fairly priced, as a change request. The build partner did nothing wrong. They asked the question, in writing, and built the answer they were given, which was both. The nine weeks were the cost of the consequence. The cause was a single decision, worth nothing to close in month one, that nobody owned.

This is the pattern, and it is almost universal. The invoice can only ever bill for work. It cannot bill for the open question upstream of the work, so the open question shows up later, laundered into a change request, priced as effort when it was really an unowned decision. Search any overrun and you will find it: not a hard feature, a soft "both".

What the per cent bought, versus what the next round asks for.

Here is the part that has changed, and changed recently. Turn the statement of work over and read the deliverables. It is a list of outputs: authentication, an onboarding flow, a dashboard, an admin panel, a billing integration, twelve things you will own. Every one is real and every one is a thing you built. Now hold that list up against what the next round actually asks for, because the bar has moved.

Carta's Q1 2026 data shows the $1m to $2.5m US pre-seed band falling from 24% of rounds to 18%. The soft middle is being squeezed out. What gets a lead now is not a longer deliverables list, it is evidence: a working product plus two to five paying pilots, or one anchor letter of intent, or a real retention curve on a small base. A caveat, honestly: that is US data, and the UK figures will differ. The direction is what transfers, not the percentages. But the direction is unambiguous. The market has stopped funding things you own and started funding things users did.

So run every deliverable through one test before you sign, and it is the most portable thing in this issue. Mark each line O or E. O is an output, a thing that will exist because you paid for it to be built. E is evidence, a thing a real user will do that you could show an investor. "Onboarding flow" is an O. "Ten target users complete onboarding unaided" is an E. "Billing integration" is an O. "Three pilots paying us monthly" is an E. Go down the list and count. If the whole page is O and there is not a single E on it, you are about to sell 7.2% of the company to buy a list of outputs, in a market that funds outcomes.

What the invoice does not price.

The gap the invoice cannot see is not the work. It is the supervision of the work, and it is where the per cent quietly turns into nothing. Four questions, none of which appears on any statement of work, and all of which decide whether the 7.2% bought anything at all:

• Was the scope still correct in month four? Not delivered on time, still correct, given what you had learned since you signed.
• Did anything reach a real user before month nine? Not a board demo, not a friendly beta, a person with the problem using the actual thing.
• Who reviewed the code, by name, against what standard? Or did it ship because the sprint ended and the burndown said done?
• Were the open questions from discovery ever closed? Or is there still a "both" sitting in the plan, compounding into next quarter's change request?

An invoice prices delivery. It is structurally incapable of pricing whether delivery was still the right thing, whether it was checked, or whether anyone with the founder's incentives was watching. That gap is the accountability gap, and no supplier can close it for you, because closing it is the one job that cannot be outsourced. You can buy the build. You cannot buy someone to care about your cap table.

The honest counterweight.

None of this is a case for building in house, and none of it is a case against paying good people to build. Most of the time, buying the build is the right call. An in-house team is its own six-figure decision in per cent, slower to start and harder to unwind, and for most pre-seed companies a strong build partner is exactly correct. The target here is never the craft, and never a named firm. The target is the unsupervised engagement: equity spent on outputs with nobody on the founder's side closing decisions, checking the work, or asking whether the scope is still true.

I can say this because I built and exited an agency at Atomise. I have written those invoices. They were accurate. Every line was work that was really done, at a fair price, by people who were good at it. And accurate is not the same as accountable. The invoice was right about what it billed and silent about everything that decided whether the money became a company. That silence is not the supplier's failure. It is the space the founder has to stand in.

The rules, plainly.

So, the whole thing as things you can actually do, in order:

• Price every phase in per cent before you sign it. Convert the invoice to a share of the post-money and ask whether you would sell that slice in a term sheet for this scope. If the answer is no, the answer is no.
• Search the scope for the word "both". Every "both", every "for now", every "either/or we will decide later" is an open decision. Close each one before the sprint that depends on it, not after, because after is a change request.
• Mark every deliverable O or E. Refuse to start a phase with no E on the page. If the build cannot produce a single piece of evidence a user generated, it cannot produce anything the next round wants.

And then the two questions that make the per cent worth paying, asked out loud before signing and again at every phase gate: what will a real user have done by the end of this that they could not do before, and who on our side owns the decisions this phase depends on? If you cannot answer both, you have not scoped a build. You have scoped a way to spend 7.2% of the company and find out later.

Price it in per cent, close the "boths", demand an E on the page. That is how the most expensive unit you own buys something a user did, instead of a list of things you own. The invoice will still be accurate. You will have made it accountable.