Project schedule estimation continues to be a tricky endeavor. Stakeholders bring a wealth of experience to each project, but also biases which could affect their final estimates. This research proposes to study differences among stakeholders and develop a method to aggregate multiple estimates into a single estimate a project manager can defend. Chapter 1 provides an overview of the problem. Chapter 2 summarizes the literature on historical scheduling issues, scheduling best practices, decision analysis, and expert aggregation. Chapter 3 describes data collection/processing, while Chapter 4 provides the results. Chapter 5 provides a discussion of the results, and Chapter 6 provides a summary and recommendation for future work.

The research consists of two major parts. The first part categorizes project stakeholders by three major demographics: “position”, “years of experience”, and “level of formal education”. Subjects were asked to answer several questions on risk aversion, project constraints, and general opinions on scheduling struggles. Using Design of Experiments (DOE), responses were compared to the different demographics to determine whether or not certain attitudes concentrated themselves within certain demographics. Subjects were then asked to provide activity duration and confidence estimates across several projects, as well as opinions on the activity list itself. DOE and Bernoulli trials were used to determine whether or not subjects within different demographics estimated differently from one another. Correlation coefficients among various responses were then calculated to determine if certain attitudes affected activity duration estimates.

The second part of this research dealt primarily with aggregation of opinions on activity durations. The current methodology uses the Program Evaluation and Review (PERT) technique of calculating the expected value and variance of an activity duration based on three inputs and assuming the unknown duration follows a Beta distribution. This research proposes a methodology using Morris’ Bayesian belief-updating methods and unbounded distributions to aggregate multiple expert opinions. Using the same three baseline estimates, this methodology combines multiple opinions into one expected value and variance which can then be used in a network schedule. This aggregated value represents the combined knowledge of the project stakeholders which helps mitigate biases engrained in a single expert’s opinion.