Using ‘Cause Association’ & ‘Systemic Mechanisms’ to produce better forecasts
The objective of a Quantitative Cost Risk Analysis (QCRA) is to forecast the cost of an operation or project. Understanding of the key risks, opportunities & uncertainties driving the forecast are an integral part of any exercise. Monte Carlo is the most common method for undertaking a QCRA.
Common practice when using Monte Carlo to forecast Cost is to use a ‘Simple Addition’ approach where the overall cost of risk is the sum is found by adding the cost of each input. By ignoring the the ways in which the inputs are connected serious flaws are introduced into the forecast. This common practice is hard wired in most Enterprise Risk Management (ERM) systems.
Into Risk has developed a framework called CASM (Cause Association Systemic Mechanisms) to simply, quickly and easily incorporate the effect of root causes (Cause Association) and knock-on impacts (Systemic Mechanisms). The framework is explained in our white paper and is supported by an Exel based monte carlo engine that can be downloaded from the same place.
The graph above compares two forecasts based on an analysis of a real Programme of Works for Simple Addition and CASM based Monte Carlo methods. The CASM forecast was produced by downloading Risk Registers from the Client’s ERM system and running them through the CASM framework. This added very little time but significantly changed the characteristics of the forecast.
The Simple Addition forecast is both narrow and symmetrical, whilst the CASM based forecast is broader and asymetrical with a long tail. In other words the CASM forecast indicates that the Programme is more likely to finish close to its budget, but if too many problems are encountered then the cost overruns would be much bigger than suggested by the Simple Addition approach. In the case above this was a much better match with the customers historic experience with projects, and suggested that the contingency levels to be given to the projects should be significantly lower.