Contingency is the least rigorous number in most capital budgets. It is typically set as a percentage of estimated cost — ten percent, or fifteen, or whatever the last project used — and then drawn down against whatever arrives first.
Both halves of that practice are wrong, and the second is worse than the first.
Why a percentage is not a provision
A percentage contingency contains no information about the project it is attached to. It says nothing about design maturity, contracting model, market conditions, or the specific exposures the project carries.
Two projects of identical value can have entirely different risk profiles. One with a completed design, a fixed-price contract, orders placed, and permits granted carries modest residual exposure. One at concept design, with an unvalidated interconnection, no orders placed, and a permitting determination outstanding carries a great deal. Applying the same percentage to both is not estimating. It is rounding.
The alternative is not complicated. Quantify the risk register: for each identified risk, a probability and a cost consequence. Aggregate them, allowing for correlation — risks in infrastructure projects are rarely independent, since a delay usually brings escalation, preliminaries, and extended overheads with it. The result is a provision with a derivation, defensible to a board and testable by a lender.
The drawdown problem
The more damaging error is in how contingency is consumed.
On most projects contingency functions as a general reserve. Something unexpected arrives, contingency covers it, and the balance falls. Nobody asks whether the risk that item related to has now been retired, or whether the remaining balance is adequate for the remaining exposure.
The correct discipline is to link drawdown to risk closure. When contingency is released, the associated risk should be closed on the register and the remaining provision re-tested against the remaining risks.
Contingency and risk should fall together. Where contingency has halved and the risk register has not moved, the project is under-provisioned and does not yet know it.
That comparison — contingency drawdown against risk closure, on a single chart — is the most informative slide in capital project reporting and one of the least frequently produced.
What contingency is not for
Contingency exists to cover the realization of identified risk within the defined scope. It is not a scope reserve, and the distinction matters commercially.
Scope added after budget approval is not a risk event. It is an investment decision, and it should compete for capital on its own merits rather than being absorbed quietly into a provision approved for something else. Projects that fund scope growth from contingency arrive at the final quarter with no protection remaining and a change history nobody can reconstruct.
The same applies to estimating error. If the budget was wrong, that is a re-baselining conversation, not a drawdown.
The escalation question
Escalation deserves separate treatment rather than burial inside contingency.
In markets where equipment and labor costs move materially year on year, escalation is not a risk but a near-certainty, and its magnitude depends on the procurement timeline. A project ordering major equipment eighteen months out has a quantifiable exposure that should be modeled, hedged where possible through early orders or fixed-price commitments, and reported separately.
Blending escalation into a general contingency conceals both. The provision looks adequate because it contains money for a certainty, and the certainty looks managed because it sits inside a provision.
What to ask for
Three artifacts make contingency governable: a quantified risk register with probability and consequence; a contingency balance stated against remaining risk rather than against original budget; and a drawdown log recording, for each release, which risk it retired.
Where those exist, contingency is a control. Where they do not, it is a cushion, and cushions run out without warning.
