Estimating projects duration in uncertain environments: Monte Carlo simulations strike back

PERT (Program Evaluation and Review Technique), developed in the 1950’s, represented the first attempt to incorporate uncertainty in project scheduling. Despite some weaknesses, it is still widely used in project management mostly thanks to the simplicity of its algorithm in operating on activity network diagrams. Today the increasing complexity of projects requires new techniques and the increasing availability of computer power have not brought project simulation into common usage as expected. Although several reviews assert that simulative approach has already superseded PERT when coping with uncertain environment; the reason why it is not diffused is that simulations require a too long computing time. In this paper we show through an algorithm and experimental results that the computational time, historically the major drawback of Monte Carlo simulations, is definitely minimum thanks also to the computational power available nowadays. We present results of an efficient program made of few lines of code and able to compute the completion time of a network activity diagram with 100.000 activities and about 50.000.000 precedence constraints between them.