The word contingency does a lot of damage. To a board it sounds like slack. To a project manager it sounds like a fund. To an estimator it is neither: it is the expected cost of the things that are known to be uncertain, priced. Of every line in an estimate, it is the one most likely to be attacked in a review, trimmed in a steering committee, and misunderstood by everyone who did not build it.
That is worth fixing, because a project funded without honest contingency does not become cheaper. It becomes a project that runs out of money in month fourteen, and the argument about whose fault that is costs more than the contingency would have.
What contingency is not
Half the battle is vocabulary, which is why AACE International Recommended Practice 10S-90, Cost Engineering Terminology (RP 10S-90), matters more than its modest title suggests. When two parties argue about contingency, one of them usually means an allowance and the other means a reserve. Three things routinely get called contingency and are not.
- Scope growth. If the owner adds a building, that is a change, not a risk realised. Changes are managed through change control against the baseline, not absorbed quietly until the contingency is gone.
- Escalation. Price movement between the pricing date and the midpoint of construction is estimated separately, shown separately, and driven by different mathematics.
- Management reserve. The owner’s own allowance for unknown unknowns, the events nobody has identified, sits above the estimate and belongs to the owner. It is not the project team’s to spend.
Contingency proper covers the known unknowns: the risks and uncertainties that have been identified, assessed, and priced. If an item can be named, ranged, and given a probability, it belongs in the contingency analysis. If it cannot even be named, it belongs in management reserve.
Where the number comes from
The framework is set out in RP 40R-08, Contingency Estimating, General Principles, and the family of method practices that sit under it, including range estimating and expected value methods. The RPs describe the principles; what follows is how the analysis looks on a real estimate.
Two streams feed it. First, every significant line in the estimate carries a range: the quantity could move by so much as design completes, the rate by so much depending on market response. Second, the risk register carries the discrete events: ground worse than the boreholes suggest, a permit arriving late, a vendor slipping delivery. Run together, usually in a Monte Carlo simulation, the two streams produce a distribution of possible outcomes rather than a single number.
Contingency is then a decision, not an output. It is the gap between the base estimate and the confidence level the owner chooses to fund at, commonly P50 for a working budget and P80 for a funding request. The analysis produces the curve; the owner picks the point on it. The full principles, including which method suits which estimate class, are in RP 40R-08 and its companions, available to AACE members at web.aacei.org.
A worked example
Take a base estimate of fifty million dollars, and keep the numbers round so the mechanics stay visible. The line ranges reflect a Class 3 basis: quantities from issued-for-design drawings that could still grow a few percent, rates carrying normal market spread. The register holds a dozen items, of which three dominate: rock where the boreholes were sparse, priced at four million if it lands, with a forty percent chance; a seasonal permit window that costs two million if missed, at twenty-five percent; and a single-source vendor whose slip would cost one and a half million, at thirty percent.
The simulation runs the ranges and the events together. Say the curve comes back with P50 at fifty-four million and P80 at fifty-seven and a half. Funding at P50 means eight percent contingency and an even chance of asking for more money later. Funding at P80 means fifteen percent contingency and a four-in-five chance of finishing inside the number. Neither is wrong. They are different appetites for a second trip to the board, and the owner chooses with open eyes.
Notice what the analysis bought. The contingency is no longer a percentage defended by habit. It is a curve with named drivers, and every dollar of it can be traced to a risk that was identified before the money was requested.
Who owns it and how it gets spent
Contingency that is not managed evaporates. The discipline is simple: the register items that priced the contingency are the same items that draw it down. When a risk lands, the cost is transferred against that item by trend. When a risk retires without landing, its share is released, not quietly reabsorbed. The monthly report shows contingency remaining against risk remaining, and the two should track each other.
What that prevents is the slush fund: contingency spent on scope changes and estimating misses in the first year, leaving nothing for the risks it was actually priced against.
If you cannot list what the contingency is for, it is padding.
Where this goes wrong
The flat percentage by habit. Ten percent because it was ten percent last time, on a project with different ground, a different market and a different schedule. The habit number survives because nobody can challenge what was never derived.
Contingency buried in the lines. Padded quantities here, conservative rates there, then a visible contingency on top of the padding. The estimate carries double contingency that nobody can see or manage, and the bid dies of it, or the budget carries fat that the trend reports can never find.
The reserve confusion. Management reserve presented as contingency, or contingency stripped out because “we already have a reserve”. The two cover different risks and belong to different owners. Show them as separate lines with separate governance, always.
Defending it in the room
The defensible version has a page for every major risk: what it is, how likely, what it costs if it lands, and what is being done about it. When the board asks why fifteen percent and not ten, the answer is a list, not a shrug. Boards do not resent contingency. They resent contingency nobody can explain.
Emerald Group develops contingency from quantified risk analysis and independently reviews contingency set by others, for owners and engineering firms. If your number is heading to a board and the contingency line is the one you are least sure of, get in touch.