A Bayesian learning approach to advance the reliability of LeAgile project portfolio management

This dissertation contributes to a growing body of research on Project and Portfolio Management for the Information System’s LeAgile projects, which require continuous planning, delivery, and improvement. It introduces a Bayesian learning approach to continuously assess the performance of the LeAgile project and portfolio. The principle of Pasteur’s quadrant is used to realize a highly practical solution, which extends the existing wisdom on LeAgile continuous planning. Binomial and Exponential distributions are applied to respectively model the number of tasks completed and time to completion in LeAgile projects; the Bayesian learning technique is implemented to continuously update the models and the performance measures such as project baseline and reliability. The accuracy of the Bayesian approach is compared with the traditional approaches using real case SharePoint data. Specifically, the continuously predicted new baseline, reliability, and project performance at both project-level and portfolio level are compared among 569 similar tasks of five projects. The results suggest that the evolving Bayesian baselines can generate a more realistic measure of performance than traditional static baselines. Similarly, Bayesian reliability estimation generated a more realistic metric to continuously plan and measure the performance of evolving LeAgile projects and portfolios. This research suggests accurate performance estimation can be achieved by continuous learning from immediately prior and continuous evolution of baselines. Furthermore, the continual learning approach considers the cumulative effect of all past experiences of each task to achieve continuous project reliability and performance prediction. This study provides a practical performance prediction tool for decision making, enabling LeAgile projects and portfolios to be better managed in the continuously changing environments of today.

Embargo status: Restricted until 06/2022. To request the author grant access, click on the PDF link to the left.