ModelRisk needs to be installed in order for the model to work.
An example of a Monte Carlo simulation risk analysis model for cost contingency calculation. Published September 2026.
Technical difficulty: 2 Techniques used: Monte Carlo simulation in Excel, correlation with copulas, discrete risk events ModelRisk functions used: VoseModPERT, VoseCopulaMultiNormal, VoseRiskEvent, VoseSimTable
Build the cost estimate as a Monte Carlo model: give each uncertain line item a three-point (PERT) range, correlate the items that move together, add discrete risk events with their probability and impact, and simulate. Contingency is the chosen percentile of the simulated total minus the base estimate — here, a P80 contingency of about €1.96m on a €12.0m base.
This page hosts the exact model behind the worked example in our guide How to calculate cost contingency. The article quotes a dozen figures from a 400,000-sample simulation of this model; downloading it turns every one of them from "trust us" into "check us".
A small capital project: ten line items, each with a three-point PERT range, summing to a base estimate of €12.0m at their most-likely values. Six of the ten — earthworks, civils, steel, electrical, piping, temporary works — share a common market/productivity driver and are correlated at ρ = 0.7 through a Gaussian copula (VoseCopulaMultiNormal). The ordered mechanical package carries a deliberately narrow range. Two identified risks are not ranges on any line item, because either they happen or they do not: ground conditions worse than surveyed (30% chance, €0.4–2.5m, most likely €0.9m) and an approval delay pushing earthworks into winter (15% chance, €0.2–1.1m). Both are modelled with VoseRiskEvent, so each appears as a single bar in the tornado chart instead of splitting into probability and impact.
A VoseSimTable switch runs the model twice in one simulation session: Sim 1 exactly as published, and Sim 2 with the six driver items independent — changing nothing else. The comparison is the point: correlation is worth about €0.30m of the P80 contingency, roughly 15% of it, and a model that ignores it undersizes the reserve by that amount while looking identical on the surface.
Run it yourself: the workbook ships preset — two named simulations, 100,000 samples, a fixed seed shared across both — so just click Start and the Results sheet fills in next to the published figures; expect small Monte Carlo differences. How many samples are enough is a question with a real answer: see how many Monte Carlo iterations you need. The article's €150k ground-survey mitigation scenario is left for you to explore — vary the ground-conditions risk's probability and impact and watch the P80 move.
How do you calculate cost contingency in Excel? Build the cost estimate as a Monte Carlo model: give each uncertain line item a three-point PERT range, correlate the items that move together, add discrete risk events with their probability and impact, and simulate. Contingency is the chosen percentile of the simulated total minus the base estimate — in this model, a P80 contingency of about €1.96m on a €12.0m base.
What is in this example model? The exact model behind the worked example in our cost contingency guide: ten line items with PERT ranges, six of them correlated at ρ 0.7 through a Gaussian copula representing a common market and productivity driver, and two risk events modelled with VoseRiskEvent. A VoseSimTable switch reruns the identical model with the six items independent, so you can see what correlation is worth.
Why does the base estimate sit at only the P18? Because the ranges are right-skewed — costs have more room to overrun than to underrun — and the risk events only ever add cost. Summing most-likely values ignores both effects, so the €12.0m base has an 82% chance of being exceeded. That asymmetry, not pessimism, is why contingency exists.
Do I need ModelRisk to open the model? The workbook is an ordinary Excel file, but the simulation functions need ModelRisk. The fully functional 15-day free trial is enough to run it, check every published figure, and adapt the model to your own estimate; the free Basic edition also opens it.
Run it on your own estimate — free 15-day trial