A Practical and Accurate Alternative to PERT

Abstract

Many real world projects can be represented by stochastic activity network (SAN) models, where the duration of some or all of the project activities are, at best, known in probabilistic sense. These models are known in the literature as PERT networks and they are analyzed using the PERT procedure. It has been known for many years that the Project Evaluation and Review Technique (PERT) provides inaccurate information about the project completion time. Quite often this inaccuracy is large enough to render such estimates as not helpful. As a result of this inaccuracy, many improvements since the introduction of PERT in 1959 have been developed. However, in spite of this inaccuracy and the many improvements, PERT procedure continues to be taught and presented in most text books on Project Management. This is due, perhaps, to its simplicity, and ease of its application. In this paper a new alternative is developed that addresses the issues of accuracy and practicality simultaneously in analyzing SAN models. The new procedure is based on a more accurate representation of the distribution function of the project completion time. We first show that the project completion time can in certain instances be accurately represented by a normal distribution, but in many other instances it can not. In these other instances we show that the project completion time can be more accurately represented by an extreme value distribution. Hence, SAN is first characterized as to when we can use the normal distribution and when we can use extreme value distribution. In the first case, PERT estimates are accurate and will be used; however, in the second case a new procedure is developed that is easy to use, and results in more accurate estimates of the project completion time and its statistics. In this paper examples are also provided that illustrate the above characterization of SANs and the accuracy and practicality of the new procedure.