The catch.
The billing model changes the pressure, but it does not decide where the pressure points.
Here is the counter in one sentence. A fixed price aims a supplier at the edge of the scope.
Watch what a fixed-price supplier does when the margin starts to die. They do not work faster. They read the contract again.
Not to invent a variation. They cannot: a change order needs the client to have asked for something first, so the supplier is not manufacturing scope, they are waiting on it. A vague scope is what makes the wait pay. You ask for the thing you always meant, the supplier reads the same sentence and sees a change, and now the two of you are arguing about what was bought rather than building it. That argument lands in week nine of a project that was already late in week six. An hourly supplier rarely bothers with any of it, because there is nothing to defend. They bill you for week nine and move on.
So the billing model settles nothing by itself. What settles it is whether anybody in the room can say what done means precisely enough to write a price against it. Get that and the fixed price is honest. Miss it and the fixed price is worse than hours, because now you have bought the argument as well as the delay.
Name the finish line.
Name the finish line: write down the one observable event that ends the project.
Ask for it in the room and the first answers are almost always states. When it is stable. When UAT passes. When the client signs off. When the snag list is closed. Anything you cannot watch happen can be argued about for another six weeks.
Some states convert. "UAT passes" is weak until it means these named tests, in this environment, by this date, accepted by this person. Written that way it is observable and it will carry a price. What matters is whether somebody can see the thing happen without calling another meeting.
A finish line is real when it passes four checks.
- It is an event. "The doors open at 08:00 on 4 September and the first visitor sees their host's name on the screen." You could point a camera at that. "The system is stable" will not survive contact with a supplier who wants another fortnight.
- Somebody outside the delivery team notices it happen. A customer, an operator, finance, compliance, a room full of invited guests. If the only people who would know the date moved are the people building it, it is an internal date, and internal dates move.
- Moving it has a visible cost. Money, credibility, compliance exposure, a dependency further down the line that stalls. Name who feels it. If nobody feels it, you do not have a deadline, you have a preference.
- Somebody can accept it. Name the person with authority to sign the thing off. Where the work genuinely needs multi-party acceptance, and regulated work often does, name the chair and put the acceptance window in the plan. A committee is a legitimate control. It only hurts you when nobody has costed the time it takes.
Four checks, one page. The client can run them without you, and you should hand them over and let them.
What the checks buy you.
Run them before anybody talks about price, because the result tells you which contract is even available.
All four pass and a fixed price becomes writable. You are pricing to an event with a date and an owner, so the scope has a natural edge and both sides are looking at the same edge. The supplier quoting it is taking real risk, and the risk is bounded by something out in the world rather than by a paragraph in an appendix.
If check one fails, stop. You cannot price a state. Any number you write against one is a guess with a lawyer attached.
Check two fails and the date will move whatever the contract says, because moving it costs nothing. A fixed price on a movable date means the supplier absorbs your indecision, and they will price for that. Badly.
Check three fails and you have discovered that the deadline came off a slide and nobody has to answer for it.
Check four fails and the finish line is real but further out than the plan admits. That one can be fixed: move the date to include the acceptance window, name the person with authority, and the price becomes writable again.
This is the step that gets skipped. The billing conversation is easier to have than the finish line conversation, so it happens first, the fixed price gets written against a scope nobody tested, and everyone is surprised in month three.
What it looked like at Canada Square.
The clearest version I have seen was KPMG's 12th Floor Welcome Lobby. The previous partner had missed the deadline and spent the budget. That is usually when we get called.
Apply the four checks to it now and they all pass. The event was the lobby opening and the firm hosting its first guests. People outside the delivery team noticed, because the guests were the entire point of the room. Moving the date cost the firm credibility with people it had already invited. And somebody could stand in the lobby on the morning and say yes, this is open.
That is what made the rest of it tractable. The question stopped being how good the lobby could be and became what had to be true for invited guests to be received in that room on that morning. Everything outside that answer could wait, and a good deal of it did. The lobby opened on time.
The checks were not the clever part. They made the clever work possible, because they stopped the scope moving while we were trying to finish it.
When there is no finish line to name.
Sometimes the honest answer is that there isn't one, and then hourly is the right shape and a fixed price is the trap.
Research work qualifies. If you are finding out whether a thing is possible at all, the finding out is the work, and nobody can name the event in advance without pretending. Platform work qualifies. A team building an internal service that a dozen products will eventually use has no single moment where it is done, and forcing one produces a fake milestone that everybody games by the second quarter.
Early discovery qualifies too. If you genuinely do not know what you are building yet, buy some hours, buy some senior attention, and go and find out. That is what hourly consulting is good at, and it is a different job with a different shape.
The mistake is running a rescue that way. A rescue has a finish line by definition, otherwise it is just a project with a bad mood. So when the checks fail across the board and the work still has to happen, the first job is governance rather than delivery: get an event named, get an owner against it, and make the cost of moving it visible to the person who will pay for it. That work is unglamorous and it is usually three meetings. Do it first anyway. Everything you buy afterwards, hours or fixed price, gets priced against it.