How to Make the Project Data Practically Relevant?

Given that many methods for project scheduling, risk analysis, and forecasting rely on distributions that represent uncertainty in activity durations, this chapter seeks the most realistic distribution to capture this uncertainty. Through the so-called calibration procedures, the chapter analyzes empirical project data, not only to verify the appropriateness of assuming a lognormal distribution but also to estimate the parameter values for this distribution as accurately as possible. While the chapter shows that an existing calibration procedure from the literature can be employed, it also highlights significant improvements obtained through a combination of human input and statistical automation. Experiments conducted on a dataset of a set of empirical projects reveal the potential of these novel calibration procedures for both further academic research and practical applications. The original idea to calibrate projects did not come from our research group, but it did inspire us to work on it ourselves. This chapter is based on three studies performed by different members of the Operations Research & Scheduling group. First and foremost, the original calibration procedure was validated using empirical project data in the study titled “Empirical perspective on activity durations for project management simulation studies,” published in Journal of Construction Engineering and Management. Next, the calibration procedure was extended to human partitioning in the study “Fitting activity distributions using human partitioning and statistical calibration,” published in Computers and Industrial Engineering. Last but not least, the procedure was extended to a powerful statistical partitioning in the study “A statistical method for estimating activity uncertainty parameters to improve project forecasting,” published in Entropy.