Every firm I audit can show me the contract value of a job. The agreed figure. The thing that got signed.
What they struggle to show me is everything that happened after.
Because a building job is never only the thing that got signed. The client asks for an extra socket. The architect moves a doorway. The ground turns out wetter than the survey promised, and now there’s an extra day of digging out and carting away soil. None of that was in the original price. All of it costs you money. And a surprising amount of it never makes it onto an invoice.
In the trade, a change to the agreed job is called a variation. It’s a plain idea wearing a formal word. You priced the job to do one thing. The client now wants that thing plus a bit, or that thing done differently. The “bit” is a variation, and you are entitled to be paid for it. The trouble is never the work. The work gets done — your people are on site, the client asked, you said yes, the socket goes in. The trouble is the proof.
The yes that never got written down
Here is how it goes missing.
A client walks the site on a Tuesday. They point at a wall and ask if it can move. Your site manager — a capable person trying to keep a job moving — says yes, no problem. The wall gets moved. It takes half a day and a bit of extra material. Nobody writes anything down, because writing things down was not the task that morning. Keeping the job moving was.
Three months later you sit down to put together the final account. That is the closing bill for the whole job: the original price, plus every variation, minus anything that came out. You go to add the moved wall. And the client says: what moved wall?
Now you are in an argument you cannot win. You did the work. You know you did. But the only record is a memory, and a memory is not evidence. The client may not even be being difficult — they genuinely might not remember a thirty-second exchange on a windy site. The burden of proof sits with you, and you have nothing to put on the table.
So you write it off. Or you split the difference to keep the relationship sweet. Either way, real work has just been done for free.
The contract already warned you
If you work under a standard form of contract — and most mid-market firms do — none of this is a surprise. It is written into the rules. The most common form in the UK is the JCT, short for the Joint Contracts Tribunal, the body whose standard building contracts much of the industry signs up to.
The JCT is blunt about variations. A change has to be instructed in writing before it can be valued and added to the agreed price. A verbal yes on site, on its own, does not count. There is even a neat mechanism built in: if someone gives you an instruction out loud, you can confirm it back to them in writing, and if they do not object within seven days, it stands.
That mechanism is quietly brilliant. Almost nobody uses it. Using it means someone, in the moment, stops what they are doing and writes the thing down — and we are straight back to the same problem. The contract is not the gap. The gap is between the conversation on site and any record that it happened.
This is a software problem, not a discipline problem
The usual fix offered here is “be more disciplined.” Tell the site managers to log everything. Send a firm email about paperwork.
I don’t believe in that fix, because I have watched it fail in firm after firm. The site manager is not lazy. They are standing in the mud with a client in front of them and ten other things going wrong. You will not win by asking a busy person to also become a filing clerk.
What works is making the record take ten seconds, on the phone already in their pocket. A variation is not complicated data. It is: which job, what changed, who asked for it, roughly what it will cost, and a photo. That is the whole thing. A purpose-built variations log — one screen, on a phone, that fires a confirmation off to the client by itself — turns the seven-day JCT mechanism from a thing nobody does into a thing that happens by reflex.
Notice what that is and what it isn’t. It is not a sprawling project-management platform with forty other modules you will never open. It is one small tool doing one job: catching the yes before it evaporates. That is the only kind of software I think most firms should buy.
And the money it protects is not loose change. On a firm turning over a few million a year, unbilled variations running at even a couple of per cent of turnover is real profit walking out of the gate, job after job, invisible precisely because it never appeared anywhere to be missed.
Where this doesn’t apply
If your jobs are small, fixed-price, and rarely change once they start — a domestic kitchen, a single shop fit-out — you probably don’t have this problem badly enough to fix with software. A notebook and a decent habit will do. Don’t spend money solving a leak you haven’t got.
And if you already run a tight variations process — written instructions going out the same day, signed off, reconciled at final account without a row — then leave it well alone. The fact that it is boring and it works is the entire point. Nobody should sell you a system to replace something that isn’t broken.
But if the words “what moved wall?” made you wince, the leak is probably already costing you more than the fix would.
If you want to know how much, that is the sort of thing I look at when I run an audit. Get in touch.