3.4 Using mathematical models
In certain cases it may be useful to aggregate the impact from multiple risks, with unrelated root causes, each with their own impact and probability. A common reason for doing this is on large projects to look at both potential schedule and financial impacts, and ensure that sufficient funds or contingency (time and money) are put aside to deal with the risks.
The approaches used to assess these risks can be quite advanced and are beyond the scope of this course. They are usually performed by experts using specialised computer packages. The information below outlines common approaches that you may come across.
Analysis type | Description |
---|---|
Monte Carlo | A mathematical model (sometimes called a stochastic model) based on a repeated random sampling to obtain a numerical value. Results are numbers. ‘Simulations’ are then brought together to give a statistical probability of a particular outcome. Results are often presented as ‘p’ values, denoting how statistically likely a particular outcome is. |
Strategic Review Analysis (SRA) or Programme Evaluation and Review Technique (PERT) | This is an approach to understanding the likely variance to a schedule. Like the Monte Carlo, this is a mathematical simulation. Each task is typically give three values (often called a three-point estimate) where the values relate to Shortest, Longest and Expected time to complete. Multiple ‘iterations’ are then run to present the statistical probability of a particular outcome. |
Expected Monetary Value (EMV) | EMV is calculated by multiplying the impact of the risk by its likelihood. For example, if the impact of a risk is £10m and the likelihood is 50% then the EMV is 10 × 0.5 = £5m. This can be helpful if there is a group of independent (unrelated) risks where the impacts and likelihood are similar. This approach is often used for calculating contingency funds for projects. Care should be taken, however, where there is a small number of risks, where the risks are inter-related and where there is a number of large, but unlikely risks. Like most risk modelling it is recommended that you engage an expert. |
It should be remembered that as with all calculations, these models are only as good as their inputs. Moreover calculations, if assumptions are not made clear, can be misleading. Often end users may not understand what a ‘p’ value is. Furthermore, the shape of the distribution may be important to understand, and, finally, people may believe that the output is more trustworthy simply because it has been put though a complicated model, irrespective of the quality of the underlying information.