Bayesian Point Estimation and the PERT Scheduling of Stochastic Activities

Conventional PERT procedures frequently introduce undesirable and often indeterminant biases into the derived time statistics of large-scale complex projects. An alternative scheduling procedure BPERT is developed, employing an activity-based time estimation loss structure and a cost-minimization criterion. Bayesian point estimates are formulated for beta-distributed activity duration times minimizing the potential losses of misestimation. Viewed as certainty equivalents, these time estimates are then aggregated to yield a single project completion time. “Crashing” is introduced, and us implications for BPERT examined. An illustrative example contrasts PERT and BPERT.