A capital project is usually approved as one number: the estimate plus a contingency of, say, 10%. Both numbers look precise, and neither says how likely it is that the project will be delivered for that amount. A 10% contingency can be generous for a repeat warehouse and reckless for a first-of-a-kind plant on difficult ground.
Quantitative cost risk analysis replaces the percentage with a probability. Each line of the estimate becomes a range, the discrete risks in the register become events with a probability and an impact, and a Monte Carlo simulation adds them up thousands of times. The result is a distribution of the final cost, and the budget becomes a choice of confidence level: P50, P80 or P90.
What the simulation shows
The illustrative project below has a base estimate of 100 million: civil works, mechanical, electrical, engineering, indirect and owner's costs. Each item has a skewed range (it can come in 5% below its estimate, but up to 15% to 30% above), the items share a market factor, and the register holds three discrete risks: poor ground conditions, a permit delay and late long-lead equipment.
Three findings are typical. First, the base estimate is not the middle of the distribution; it sits around the 4th percentile, because the ranges are skewed to the upside and risks only add cost. Second, the distribution has a long right tail: the worst runs are 35% above the estimate. Third, the gap between confidence levels is large. Funding at P50 needs a contingency of about 10 million; funding at P80 needs 15.5 million.
Choosing the confidence level
There is no single correct level; it depends on who carries the overrun and how often the organisation can afford to go back for more money.
- P50 is the expected cost of a single project, appropriate for a portfolio where overruns and savings offset each other, with a central reserve held above the projects.
- P80 is a common choice for a project's own budget: it will be exceeded one time in five, which keeps the pressure on the team without making overruns routine.
- P90 suits projects whose overrun would be hard to fund, such as public infrastructure or a project that is large relative to the balance sheet. Australian federal infrastructure guidance, for example, asks for both P50 and P90 estimates on major projects.
A practical design is to split the two: the project manager controls the contingency up to P50, the sponsor holds the difference up to P80, and the board holds a portfolio reserve. Each level of release requires a reason, which keeps contingency from being spent simply because it exists.
Where the uncertainty comes from
The second output is often more valuable than the first. Two items, civil works and the risk of poor ground, explain 45% of the uncertainty. That is where to spend before the final investment decision: a ground investigation costing a few hundred thousand can remove a risk worth millions of contingency. The simulation turns risk management from a register that is reviewed into a budget that is reduced.
Bartor S.
Founder
