When a project is announced as 18% over budget, it reads like an event that happened in the month it was announced. It did not happen that month — it was discovered that month. The overrun itself accumulated over months of small gaps: a change order executed before it was priced, a delivery short of what was invoiced, overtime that never reached a cost code, a bill-of-quantities line quietly running past its contracted quantity. Every one of them was visible in the week it happened, and nobody was looking at it that week.
Key takeaways
- Overruns do not arrive in one piece: they accumulate from small gaps, each of which was detectable in the week it occurred.
- The single largest leak is not mispricing — it is work executed before it is priced, a verbal instruction started before anyone knew its value.
- The procurement gap (ordered vs received vs invoiced) hides differences that only the intersection of all three documents reveals.
- Labour hours that never reach a cost code disappear from the project and surface in company overhead, making the work look cheaper than it was.
- Actual cost means nothing without progress beside it: spending 40% of the budget is either excellent or alarming depending entirely on what was built for it.
An overrun is a late discovery, not a sudden event
On most financially troubled projects, the moment the contractor senses a problem arrives far too late: while preparing a payment application, reviewing a supplier account, or when the accountant asks about a variance too big to ignore. By then the work is built, the materials are consumed and the invoices are paid. Nobody has a decision left to make — only an explanation to give.
The difference between a contractor who controls cost and one who explains it afterwards is not the accuracy of the original estimate. It is the delay between a gap occurring and the gap being seen. An estimate 5% out and caught in month two is entirely recoverable; the same 5% caught in month nine is a booked loss.
Where the money actually leaks
The recurring leaks on construction projects are a short, well-known list, and most of them have nothing to do with how the job was priced:
| The leak | How it happens | Earliest detection point |
|---|---|---|
| Unpriced work | A verbal change order starts on site before its value is approved | The day of the instruction — before the first worker moves |
| BOQ quantity overrun | A line gets built past its contracted quantity unnoticed | At the weekly recording of executed quantities |
| The procurement gap | More invoiced than received, or less received than ordered | At goods receipt, not at invoice matching |
| Hours with no cost code | Labour and overtime paid on the project but never attributed to it | At the daily attendance calculation |
| Material waste and issue | Stock issued with no task attached and quantities nobody can trace | At the moment material leaves the store |
| The cost of delay itself | Extended plant hire and site supervision caused by schedule slippage | At the first deviation on the critical path |
The third column is the important one. Every leak has a moment when catching it is almost free, and another moment — months later — when catching it is just documenting the loss. The system worth having is not the one that computes the overrun accurately at the end; it is the one that pulls the moment of discovery back toward that third column.
Work executed before it is priced
The single biggest leak in contracting is not a wrong price, it is work that began with no price at all. The scene is familiar: the owner or consultant asks for a change on site, the engineer builds it because standing a crew down is worse, and the pricing "gets sorted later". By the time later arrives, the negotiating position has completely inverted — the work exists, and the owner is discussing the value of something they can already see and never signed for.
Work that starts before its value is approved is no longer a change order — it is a claim.
The fix is not to refuse fast execution; that is not realistic on a live site. The fix is to record the instruction at the moment it is given: who asked, when, and the estimated effect on cost and time — even before the final number is agreed. A record dated the day the request happened is far stronger than minutes written three months later from memory.
The procurement gap: ordered, received, invoiced
Three numbers describe any material entering a project: what the purchase order asked for, what the store actually received, and what the supplier invoiced. On a disciplined project the three agree. On every other project no two of them are ever compared, because each number lives somewhere else — the order with procurement, the receipt with the storekeeper, the invoice with accounts.
| Document | Quantity | Value |
|---|---|---|
| Purchase order | 100 t | 250,000 |
| Goods receipt note | 92 t | 230,000 |
| Supplier invoice | 100 t | 250,000 |
| Recoverable difference | 8 t | 20,000 |
That is 20,000 on one material in one month. Reviewing the invoice will never find it — the invoice is perfectly consistent with itself. Only crossing it against the receipt and the order does. That intersection is what a three-way match is, and it is the difference between "we paid what was invoiced" and "we paid for what we received".
The practical requirement is that recording a receipt has to be easy enough to happen at the gate as the truck is unloaded. A goods receipt entered three days later from memory loses its value as an independent record, because it usually gets copied from the very invoice it is supposed to be checking.
Hours that never reach a cost code
Labour is usually the largest cost line on a construction project and the most easily lost in the accounts. A worker moved to another site for three days, overtime agreed verbally, a crew standing idle waiting on a delivery — all of it is paid time. If it is not attributed to a project and a site, it vanishes from the project cost and reappears in company overhead.
The consequence is worse than one wrong figure: the work looks cheaper than it was, and the next tender gets priced on that number. A documentation error becomes a pricing error, and a one-off loss becomes a recurring one.
- Attendance captured with its coordinates and its work site, not a timesheet filled in on Thursday.
- Overtime recorded on the day it happens, with approval, rather than reconstructed at payroll.
- Material issued from the store against a task, not merely deducted from a stock balance.
- All of it attached to the project that consumed it, so the cost lands where it was incurred.
The number that warns you early
Every gap above eventually shows up in just two numbers: what we spent, and what we built for it. Having spent 40% of the budget is neither good news nor bad news on its own — it becomes news when placed beside percent complete. 40% spent against 50% complete is excellent; 40% spent against 25% complete is an overrun heading for 60% if the trend holds.
That is exactly the cost performance index (CPI) in earned value analysis: earned value divided by actual cost. Below 1 means every unit of currency spent bought less than a unit of work. Its value is that it produces a signal in month two or three — while there is still time — rather than in the closeout review.
How this works in muqawil
Cost control is not a separate module in muqawil. It is what accumulates when every other module records into one place — because the overrun itself accumulates the same way.
- A bill of quantities with lines, quantities and rates, and progress recorded against the lines rather than against the project as a whole.
- Change orders whose effect is recorded against the baseline live at decision time, with the person who raises one separated from the person who approves it.
- A three-way match across purchase order, goods receipt and invoice, with receipts recorded from the site.
- Attendance with coordinates and a work site, turning into hours and then into payslips — each shift a run claims stays attached to its site, so the cost reaches the project that incurred it.
- Material issued from each site's own stock against a task, with a moving average unit cost per item.
- Earned value (CPI/SPI) computed from the same recorded data, not from a separate sheet someone maintains by hand.
The order of that list is deliberate: the indices sit at the bottom because they say nothing unless everything above them is being recorded as it happens. A cost performance index computed from incomplete data is a precise number describing an incomplete reality.
Frequently asked questions
What is the single biggest cause of cost overruns in construction?
Work executed before it is priced and approved. A wrong tender rate shows up early and can be managed; work already built without approval turns into a negotiation after it has already been paid for in materials and labour.
How often should actual cost be reviewed against planned?
Monthly at minimum, and weekly on high-value lines. A quarterly review finds the variances after the window for doing anything about them has closed.
Does a three-way match really stop overpayment?
It stops one specific kind: paying for quantities that were never received. It will not reveal a rate that was too high in the first place — that is the job of bid comparison at purchase, not matching at payment.
What if the budget is a single lump-sum figure with no lines?
The comparison still works, but bluntly: you learn that you are over, not where. Breaking the budget into lines is what turns "we are 18% over" into "we are over on the MEP package specifically".
Is the cost performance index enough to judge a project?
No. It is read with the schedule performance index: a project underspending because it has effectively stopped will show a healthy cost index and a terrible schedule one. The two numbers together are the reading.
How do we stop the same overrun repeating on the next project?
By making actual cost reach its correct lines during execution. The next tender is priced from the last project's cost, so if hours or materials were lost into overhead, the next tender inherits the same error at a larger scale.
Find the variance in its own week, not in month nine
Record the bill of quantities, change orders, procurement and attendance in one place in muqawil, and let actual cost reach its own line as it happens.