Why Monte Carlo simulations of project networks can mislead
Monte Carlo simulation of project networks is a standard project-modelling technique. However, much of this analysis is inadequate, as project managers always take action to recover late-running projects, which is ignored in most models. This paper shows the importance of this omission (which generally gives unreasonably wide probability distributions) and discusses simple and easily coded models of project-management actions. The paper also notes a second flaw, explaining why risk-analyses rarely predict catastrophic overspends that sometimes occur, namely the inability to capture feedback loops resulting from chains of causality from management actions. The need to recognize these as part of the modelling and then take steps to avoid them is described.