Cost risk analysis is, perhaps, the most common application of the Monte Carlo simulation technique in project management. Usually, a cost risk model is built using lines items of cost defined at a relatively high level of the cost breakdown structure. The planned values for these lines of cost are then substituted with uncertainty estimates (typically defined by probability distribution functions like those illustrated below). The model is then augmented with risks from the project risk register which are additional potential sources of cost. The structure created is likely to be similar to that illustrated in the figure below-right, taken from the APM’s PRAM Guide. Some projects may make the simplifying assumption that uncertainties in the planned lines of cost are negligible in comparison to the potential impact of risks. In this case, the first section of the model would be removed and the purpose of the analysis of risks only would be to estimate the financial reserves required over and above the planned costs.

The Monte Carlo process analyses the cost risk model by randomly simulating the effects that a mixture of good and bad luck may have on the overall cost. Iterations in which there is a disproportionate amount of bad luck result in relatively high costs and, similarly iterations in which there is a run of good luck result in much lower costs. The result is that the simulation output should include all the possibilities that are, realistically, possible. The output is usually displayed as an S-curve.

Monte Carlo cost risk analysis is particularly useful at strategic points during a project when managers need a realistic understanding of the eventual project cost. They can also be used to distinguish between targets (by which a realistically low cost outcome might be incentivised) and expected costs or commitments. A number of major risk-mature organisations use this type of analysis routinely as part of their governance processes. For example this type of information might be used in evidence to support a major project go / no go decision. However, it is important that any analysis required to support such decisions is conducted to a high standard.

Performing cost risk analysis to a high standard is likely to involve a number of challenges. Typically, these include:

• Structuring the cost risk model so as to avoid duplications of risk impact, whilst ensuring that all major sources of risk are included
• Discriminating between sources of cost outcome uncertainty and sources of risk events
• Making realistic estimates for the uncertainties (most estimates tend to be too narrow)
• Making realistic probability and impact estimates for risks
• Including appropriate degrees of correlation within the model

The problem of model structuring is often overlooked. A symptom of weak risk analysis is a cost risk model formed by importing every cost-related risk from the risk register. This structure assumes that the overall risk is equal to the sum of the parts. Whilst this may be a reasonable simplifying assumption for simple projects, it quickly becomes irrational on the type of large or complex projects for which Monte Carlo analysis is often considered to be worthwhile. A typical problem is that risks that impact on both time and cost can result in duplications in the cost risk model. For example, if the project is exposed to so-called “marching army” risks concerned with the retention of parts of the project team for longer than planned, then the inclusion of such estimates in multiple risks may result in duplications during some simulations. The time / cost variances area at the bottom of the cost model figure above shows lines related to this type of effect. A similar approach may be needed for other time-driven risks such as liquidated damages liabilities.

More detailed advice on structuring cost risk models can be obtained from papers that can be downloaded from this site or from the APM’s PRAM Guide.

Pertmaster and @RISK for Excel are both suitable tools for cost risk analysis. Further details on these and other risk tools can be found at www.risktools.com.

Contact: martin.hopkinson@hvr-csl.co.uk