How to tell if a fixed-price job is losing money before it finishes

A fixed-price job rarely goes wrong at the end. It goes wrong in week three. Nobody finds out until the final account is agreed, and by then the money has gone and the only question left is who absorbs it.

Knowing how to tell if a job is losing money before it finishes means reading leading indicators rather than outturn. A leading indicator is a number that moves before the profit does. Labour hours burnt against percent complete. Committed material cost against the take-off. Work instructed on site but never priced.

Each of those turns weeks ahead of the final account. Each is visible on a Tuesday afternoon from information the business already holds, provided somebody is asked to go and look at it.

Why does a job's margin only appear at the end?#

Because most firms measure spend, and spend on its own says nothing.

A cost report tells you that forty-eight thousand pounds has gone into a job budgeted at a hundred and twenty. That is not a signal. It becomes one only when you set it against what has physically been built. Most contractors do not pair those two numbers until the job is closed and invoiced.

The gap is expensive and it is normal. KPMG's 2015 Global Construction Survey found that only 31% of projects came within 10% of their budget, which tells you how routine the drift has become. Routine is not the same as survivable.

None of this is unique to construction. The same argument runs through professional services. Written for studios rather than sites, the guide to designing profitable projects for creative agencies makes it in the language of margin targets and scoping. Decide the number you are protecting at the quoting stage. Then measure against it while the work is live.

By the time a job finishes, the decisions that cost you the money are months old, taken by people who have since moved to other sites and recorded in a WhatsApp thread nobody thought to keep. You cannot unbuild a wall. You can only price the next one better, which is a long way from getting paid for this one.

How can you tell if a job is losing money before it finishes?#

Compare three things every week: hours booked against percent complete, committed material cost against the take-off, and variations instructed against variations priced.

That is the whole method. None of it needs new software, a dashboard or a quantity surveyor on staff; what it needs is a percent complete figure the site manager will say out loud and stand behind a week later. That is the part that fails.

Indicator	What you compare	How often	What it changes
Labour burn	Hours booked against percent complete	Weekly	Gang size and sequencing
Material drift	Committed cost against take-off	Weekly	Ordering, waste, rework
Unpriced variations	Instructed against priced	Weekly	Paperwork and cash
Preliminaries burn	Weeks used against weeks allowed	Monthly	Programme

Run all four and you will know, in twenty minutes a week, which of your live jobs is in trouble.

Labour burn is the first number to move#

A fixed-price job is won or lost on labour. Buildtrayd puts labour at around 60% of total project cost for most contractors, and higher again for firms that self-deliver rather than sublet. It frames the timing plainly too: an overrun caught in week two is a correction, while the same overrun found at closeout is a loss. Nothing that large drifts quietly for long.

The check itself is a ratio, not a report.

Take the hours booked to the job. Divide by the hours you estimated. Then compare that percentage against how much of the work is standing. Half the hours gone with a third of the build complete is not a rounding error. It is a forecast, and the forecast is that this job ends under water.

Two causes account for most of it. The first is sequencing: a trade turning up before the one in front has finished, so the gang spends its morning working around stacked boards, a scaffold in the wrong place and somebody else's wet screed. The second is unbooked time, where the hours go to the job in reality but land somewhere else on paper.

Both are fixable in week three. Neither is fixable in week eleven.

Material cost drift, and the order nobody priced#

Committed, not invoiced. That is the distinction that decides whether this indicator works.

A purchase order raised on Thursday is money spent, whatever your accounting system thinks about it on Friday. Firms that track invoices instead of commitments run a month behind their own costs, and on a ten-week fit-out a month is most of the job.

Then watch the small ones. A second skip. An extra delivery charge because the first drop was short. A pack of boards replacing the pack that was cut wrong.

That last category has a name and a price. The Get It Right Initiative's research report puts the direct cost of errors that result in defects at 2.5% to 10% of turnover for Tier 1 contractors. Around 5% is the typical figure. Its wider estimate is blunter still: the total cost of error runs well above a tenth of what construction costs in the first place. Rework is the line item nobody puts in the take-off.

Unpriced variations are margin you have already spent#

This is the third indicator. On fit-outs it is the worst.

A variation instructed on site is work you have already paid for. Labour, materials, plant, all of it. Until it has been priced and accepted in writing, it is not revenue at all. It is a gift.

So count them weekly, in two columns. Instructed on the left, priced on the right. When the left column has run longer than the right for three weeks, the margin on that job has already moved. Nobody has told the client.

The discipline travels. Pricing a change against what it costs you applies to a site variation exactly as it applies to charging for extra revision rounds in a studio. Both start in the same place. Know your own cost before you name a fee, instead of negotiating it once the work is done.

What margin are you actually protecting?#

Whatever margin a healthy contractor runs, typically 8% to 12% net. Foundation Software puts that range against a construction industry returning 6.3% pre-tax net income on revenue in 2023. Those are thin numbers. They are also what makes the weekly check worth doing.

Work it through on a job. Take a $250,000 fit-out priced with a 10% margin, so $25,000 of profit sitting on $225,000 of outlay. Say labour is $115,000 of that. Run it 20% above estimate and the overrun alone is $23,000.

So the profit is gone. The job still looks busy, still looks fine on site, and still gets talked about as a good one until the final account lands.

What do you do when a signal fires?#

Act in the week you see it, not at the next monthly review. Three moves cover most cases.

• Re-forecast to completion. Take the current burn rate, run it forward to the end of the programme, and write down the outturn you land on. A forecast outturn is a number you can argue with. A percentage overspend is not.
• Price every open variation before Friday. Instructed work that has not been priced is the cheapest margin to recover. The client has not yet decided it was free.
• Change the labour plan, not the hours. Adding hours to a fixed price adds cost. Resequencing, a smaller gang, or pulling a following trade forward changes the outcome instead of funding it.

What you should not do is wait for better information. The week-two picture is rough and it is enough. The week-eleven picture is precise and it is a post mortem.

This is the loop Korrel is built around. A proposal carries deliverables, risk flags and milestone billing, with cost and sell prices set line by line. Convert that proposal into a project and time and expenses book against those same lines, so the actuals sit beside the original estimate while the job is still running rather than months after it completes.

Where to start this week#

Pick the job you are least sure about. Not the worst one. The one you could not describe confidently if someone asked you over the phone, because that uncertainty is itself the symptom.

Then do three things. Get an honest percent complete from the person doing the work. Pull the hours booked to date. List every variation instructed since the last valuation, and mark which ones have a price against them.

Twenty minutes, once. If the hours are ahead of the build and the variation list has unpriced rows, you have your answer. You also have weeks of programme left in which to resequence the trades, price the outstanding instructions and tell the client something they can still do something about, which is the entire point of looking early.