Ask three people what construction contingency percentage a scheme should carry and you will get three numbers, all of them round, none of them derived from anything. The figure is usually inherited from the last project, adjusted slightly for how the speaker feels about this one.
That approach survives because it is quick and because it is rarely tested until it fails. When it does fail, it fails in a specific way: the allowance turns out to have been covering something nobody had written down, and there is no basis on which to argue about it.
A percentage is an output, not an input
A contingency should be the sum of identified risks, each with a value and a likelihood, expressed afterwards as a percentage for convenience. Built that way it can be explained to a board, defended to a funder, and reduced line by line as each risk resolves.
Built the other way round, as a percentage applied to a total, it can do none of those things. Nobody can say what it covers, so nobody can say when any of it is free. It therefore either sits untouched until the end of the job, which is expensive, or gets spent on the first thing that goes wrong, which leaves nothing for the second.
Separate design risk from construction risk
These behave differently and belong in different places. Design development risk covers growth as the drawings mature: details that were not shown, quantities that were approximated, specifications that were described generically. It is at its largest at the start and should reduce at every design stage.
Construction risk covers what happens once the design is fixed and work is under way: ground conditions, weather, access, coordination between trades. It does not reduce with design maturity. It reduces as the work is physically completed.
A single combined allowance conceals both movements. Split, the two tell a useful story: design risk falling as the drawings improve, construction risk falling as the building goes up. When design risk is not falling, that is information a board can act on.
Build the register before the number
The list comes first. Ground conditions where no intrusive survey exists. Statutory diversions with no confirmed date. A planning condition not yet discharged. A specification item with a single supplier. An adjoining owner matter. A design element the team has not yet resolved.
Each item gets a description, a value if it occurs, and an assessment of how likely it is. The value is the part people avoid, because it requires a judgement they may later be held to. That judgement is exactly what the client is paying for.
Where a risk cannot be valued at all, it is not a contingency item, it is a project decision that has not been taken. Saying so is more useful than pricing it at a comfortable number.
Refurbishment changes the arithmetic
On new build the design is the main uncertainty. Equally, on a refurbishment the existing building is, and it is uncertain in ways that no amount of design work resolves. Structure, services, hazardous material and compliance triggers stay unknown until the fabric is opened up.
That is why a percentage borrowed from a new build scheme is particularly dangerous on refurbishment work. The risks are different in kind, not just in size, and they retire on a different timetable: not as the design matures but as the survey programme completes and as opening up progresses on site.
The practical consequence is that survey work is the cheapest contingency reduction available. Every risk converted into a measured item is a risk removed from the register. That relationship is set out further under refurbishment and reuse.
Who holds it, and who can spend it
A contingency inside a contractor's price is not the client's money and will not come back. A contingency held by the client, outside the contract sum, is available and can be released. The two are frequently confused in conversation and the difference is significant.
Where a scheme is funded, the funder will usually want the client-held contingency identified separately in the drawdown schedule, with rules about what it can be spent on and who approves each release. Agreeing that before the first drawdown avoids an argument at the fourth. The lender view of this sits under development monitoring surveying.
Authority matters as much as quantum. A contingency that any project manager can spend without reference to anybody is not a contingency, it is a budget increase that has already happened.
Release it deliberately
The point of a built register is that each line can be closed. The ground investigation comes back and the abnormals line reduces to the measured figure. The planning condition is discharged and its allowance is released. The single source item is ordered at a known price.
Each release should be recorded in the cost report with a reason. Over the life of a scheme that produces a visible narrative of risk reducing, which is a far better report to a board than a static figure that never moves until it suddenly does.
It also protects against the opposite failure, where contingency quietly disappears into ordinary change without anybody noticing that the protection has gone. Monthly discipline on this sits under cost control and variations.
What this means for you
If your contingency is a round percentage, ask what it covers. In the event that nobody can answer in a list, the allowance is not protecting you from anything specific and it will not survive its first serious test.
Rebuild it from named items. The exercise takes a day or two on most schemes and it changes the conversation permanently, because from that point on the contingency is a document rather than a number, and documents can be argued with, tested and improved.
Not sure whether your contingency is adequate?
Send the cost plan and the risk register if one exists. We will tell you what the allowance is actually covering and where the gaps are.