A New Approach to Predicting Reliable Project Runtimes via Probabilistic Model Checking

Abstract

For more than five decades, efforts of calculating exact probabilistic quantiles for generally distributed project runtimes have not been successful due to the tremendous computation requirements, paired with hard restrictions on the available computation power. The methods established today are PERT (Program Evaluation and Review Technique) and CCPM (Critical Chain Project Management). They make simplifying assumptions by focusing on the critical path (PERT) or estimating appropriate buffers (CCPM). In view of this, and since today’s machines offer an increased computation power, we have developed a new approach: For the calculation of more exact quantiles or – reversely – of the resulting buffer sizes, we combine the capabilities of classical reduction techniques for series-parallel structures with the capabilities of probabilistic model checking. In order to avoid the state space explosion problem, we propose a heuristic algorithm.