Sensitivity analysis in PERT networks: Does activity duration distribution matter?

Abstract

Abstract The Program Evaluation and Review Technique (PERT) is a stochastic network technique developed in the late 1950s. The original model assumes PERT β distribution for the activity durations. Later on, this was criticized by many researchers, and several new distribution types have been introduced, which were believed to be better in modelling the real activity duration distributions or were easier to handle from a mathematical point of view. Introduction of new activity duration distributions was criticized even by Clark—one of the pioneers of PERT—while other researchers were arguing that these new activity distributions could better describe the stochastic distributions of the activity durations and their application helps define the distribution of project duration better. In the course of our research, we have investigated the effect of various activity duration distribution types (PERT-beta, uniform, triangular, lognormal) on the project duration, as well as the impact of the inaccuracy of the activity durations' estimation when performing the PERT three-point estimation. Our basic assumption—that the differences in the distribution of the project duration caused by using different activity duration distributions are not considerable compared to the differences caused by the inaccuracy of the three-point estimations—has been tested on several hypothetical projects and case studies. Four different distributions have been applied, one at a time. The aim has been to clearly show the differences between the results of the application of the selected distributions, therefore the distributions have not been mixed. Monte Carlo analysis has been used to create the probabilistic distributions of the projects. All example projects have proved the basic assumption to be true. The results suggest that from a practical point of view, the accuracy of the three-point estimation is more important than the type of the activity duration distribution. Consequently, the selected activity duration distribution is essentially insignificant. The most important practical consequence of this research is that instead of the time-consuming and costly process of selecting the proper activity duration distributions, planners should devote more effort to adequately determine the activity durations.