Price the phase, not the project.

There is a sentence a founder cannot say in month four of a build, and the reason has nothing to do with courage.

The sentence is: the plan we agreed in month one is wrong. Not late, not badly built, not anybody's failure. Wrong, in the sense that you have learned things since January that could not have been known in January, and the thing you are nine weeks into building is no longer the thing that should be built. Almost every founder I sit with has some version of that sentence in their head by month four. Almost none of them say it out loud, because the arrangement has nowhere to put it.

Say it on a status call and it lands as an accusation, because the only people in the room are the people who have done the work. Say it in writing and it becomes a change request, which is a document with a price at the bottom. So the sentence stays where it is, the build carries on to the plan, everybody behaves impeccably, and eight months later you launch a careful and well made answer to a question you stopped having in March.

Why nobody in the room is wrong.

It is worth being precise about whose fault this is, because the honest answer is nobody's.

A fixed project price is a good instrument. It moves the estimate risk, the staffing risk and the it-took-longer risk off the founder and onto the supplier, which is exactly where a company with eighteen months of runway wants them. It makes the number in the board pack real. It lets a company with no engineering leadership buy engineering without also having to manage it. Somebody quoting nine months of work at a fixed price is absorbing real risk and charging for it, and they should. I have quoted plenty of them. The instrument is not the problem.

The problem is what the instrument requires. A fixed price can only be held if the scope is held, because the price was calculated from the scope. So the supplier defending the plan in month four is not being obstructive. They are being professional, and if they cheerfully absorbed every change you thought of they would be running a worse business, and eventually a business that was no longer there in month nine when you needed it. That is the mechanics of N°40 arriving in real life: hold all three corners and you have also, quietly, fixed the plan.

Which leaves the founder holding the only unpriced item in the whole arrangement, which is the knowledge that the plan has gone stale, and no legitimate way to spend it.

A phase is a unit of commitment.

The alternative is not open ended time and materials. That hands every risk back to the founder and is worse. It is to keep the fixed price and change what it is fixed to.

Fix the price per phase. Six to eight weeks, scoped properly, priced properly, with the same protections on both sides that a project contract carries. The one difference is that a phase does not end on a delivery. It ends on a decision. You are not buying a build. You are buying the next eight weeks of a build and the right to look at what they produced before you buy the eight after them.

Say what that costs, because it costs something. Here are illustrative numbers, shown working, so you can redo them with your own. A build quoted as a single project comes to £180,000 over about eight months. The same scope broken into four phases of eight weeks comes to £49,000 a phase, so £196,000 if you run all four. That is £16,000 more, just under nine per cent. The premium is real and it is not padding. Each phase carries its own scoping and its own quote, and a supplier who cannot commit your team eight months out cannot price as though they could.

Now look at what the nine per cent bought. Under the project contract you committed £180,000 on one signature. Under the phase contract, if the gate at the end of phase two says this is not working, you have spent £98,000 and you stop. The £16,000 is not an administration fee. It is the price of an option to stop, three times, and it is the only line in either contract that gives you one.

What a gate has to contain.

All of which is worthless if the gate is decorative, and most gates are. A meeting at the end of eight weeks where the work is walked through, everybody agrees it was built, and phase two starts on Monday has a gate's furniture and none of its function. It asks the one question whose answer was never seriously in doubt.

A gate that does something asks five questions, in this order.

• What evidence did this phase produce? Not what shipped. Evidence is something a person outside the building did or said that you did not already know. A list of features is an account of your own activity, and eight weeks of activity was never in question.
• What decision is on the table because of it? A gate with no decision in it is a report. If the honest answer is that nothing has changed and nothing needs deciding, that is a finding too, and a suspicious one this early in a build.
• Who makes it? One name. Not the team, not the steering group, not we. A decision with three owners is a decision that gets carried politely to the next gate by everybody at once.
• What would make phase two not happen? Write it down while you still do not know the answer. A gate that has never been capable of producing a no is not a gate, and the wrong moment to discover that is the gate where you needed one.
• What does phase two cost? Fifth. Not first, not alongside. Fifth.

The order is not presentation, it is the mechanism, and it is the part of this I would defend hardest. Ask what the next phase costs first and you have opened a negotiation, and everything after it, including the evidence, gets read as material for that negotiation. Ask it fifth, once the evidence is on the table, the decision is named, somebody has put their name against it and you have both said out loud what would stop this, and the number is being quoted against a decision that has already been made. The same five questions. Two completely different meetings.

I have made a shorter version of this argument before, and said that a phase gate needs two questions. That was not enough, and I would rather supersede it than quietly restate it. Two questions can establish that evidence exists and that somebody owns the decision, and still fall over, because neither of them fixes when the price gets asked. If the number is on the table while everyone is looking at the evidence, the questions stop mattering.

One addition for a first phase specifically. Somewhere in that first gate, what the product actually is has to be settled rather than assumed, including the surface it lives on and whether the thing you demonstrated is the thing you intend to sell. Most plans treat that as given on day one. It is a phase one decision.

One phase, worked through.

Here is an invented example, so that I can tell you how it ended.

A company selling rota and compliance software to independent care providers. Phase one, eight weeks, £49,000, and the plan said it would produce the rota builder and the shift swap flow. Both were built. Both met their acceptance criteria exactly, which is the first thing worth noticing, because the standard the work was measured against was written by the people doing the work and did not mention a care manager anywhere in it.

The gate did not ask whether the rota builder was finished. It asked what evidence eight weeks had produced. Six care managers had run one real month of their own rotas through it rather than the demo account, which is the only version of that sentence worth writing down, since a demo can be entirely true and still be evidence about nothing your customer will ever touch. Four of the six never opened the shift swap flow. They swap shifts on WhatsApp, they have done for years, and they were not going to stop. All six, unprompted, tried to use the thing to prove something to an inspector, which it did not do at all.

So the decision on the table was not whether to carry on. It was whether this is a rota product with compliance attached or a compliance product with a rota inside it, which are two different companies. The founder's name went against that one, not the team's. What would stop phase two was written down: if two of those managers would not show the inspector facing output to an actual inspector, phase two does not get built. And the gate named who owned the switch away from the spreadsheets these managers still run their weeks on, because a gate that prices the next phase without naming who owns the switch has priced half of it.

Only then did anybody ask what phase two cost.

Phase two became the compliance record. The shift swap flow was not deferred politely, it was dropped. Under the project contract that flow would have been finished and polished in month five, because it was on the plan and the plan was the price, and nothing in that arrangement was capable of noticing that four people out of six had walked around it.

The honest cost of working this way.

Two objections, and one of them is correct.

The first is that this is less efficient, and it is. Four scoping conversations instead of one. Four quotes to write and four to read. Senior people pulled off the build for a week, four times, to think about what happens next. That is genuine overhead, it belongs in the decision, and anybody telling you it is free has not run one.

The second cost sits on the supplier's side, and naming it honestly explains most of the resistance you will meet. A phased engagement means they cannot fill the pipeline eight months ahead. Their forecast gets shorter, their staffing gets harder, and that is a real business cost rather than an attitude. It is most of the nine per cent. A good supplier will tell you exactly that and quote for it, which is the answer you want to hear.

The wrong ones say something else. They will call the phasing inefficient while meaning that a nine month signature is the most valuable asset on their books, because the efficiency argument is a better argument than the real one. You can usually separate the two in a single question: ask what a gate would have to show for them to recommend not doing phase two. Somebody who works this way has a specific answer ready. Somebody who has never had the thought will tell you it will not come to that.

Who owns the gate.

Which leaves the question this whole month has been circling.

A founder arrives at a build with two questions. Can this be built, and should this be built. Everybody in the room is paid to answer the first. Nobody is paid to answer the second. That is not dishonesty and it is nobody's character flaw. It is a seating plan. The people who can answer whether it can be built sit on the supplier's side, because that is where the skill lives, and the gate exists because of where people sit rather than because anyone is behaving badly.

So the gate cannot be owned by the supplier. Not because they would lie, but because you would be asking them to weigh evidence about whether the next eight weeks of their own revenue should happen. Nobody marks their own homework well, and the better they are, the more uncomfortable a position you have put them in.

And it cannot be owned by the founder alone at eleven at night, with a demo they watched and a deck they were sent. Not for want of intelligence. Because they have not done it before, they are the most invested person in the room, and the specific skill of looking at eight weeks of work and saying that is activity rather than evidence is one you acquire by having been wrong about it a few times.

Which means somebody has to sit on the founder's side of the table and own the first four questions before the fifth one gets asked. That is the argument in N°49, arriving in a specific chair on a specific Tuesday. The person can come from several places: a non executive who has run a build, an experienced product person for a day a week, a technical co-founder if you have one and they are not the party being measured. The job title is not the point. The point is that they are not paid by the outcome of the decision, and that they have watched a gate go wrong before.

I built and exited an agency at Atomise, and I have sat on the supplier's side of enough of these to know that the chair I am describing is usually empty and that the work is better when it is not. N°62 made the same point from the revenue side of the invoice, where the number nobody on the supplier's side is paid to check is the one you charge your own customers. This is the cost side. Same missing chair, same eight weeks, different invoice.

So take the plan you have signed, or the one sitting on the table waiting for a signature. Find the eight week mark. Write down two things: what evidence that date is supposed to produce, and whose name goes against the decision it leads to. If the honest answer is a list of features and no name, what you are holding is a delivery date. Turning it into a gate costs one conversation, and it is a great deal cheaper to have that conversation now than in month four, when you will be having it about £180,000 you have already committed.