Programme Management Monitoring and Controlling: statistical predictive models to improve Estimate at Completion

Programme Management plays a crucial role in successfully executing a group of related projects which together achieve a common purpose in support of the strategic aims of the business. Effective monitoring and controlling activities in the context of a programme are essential to ensure adherence to budget and schedules. Forecasting cost estimate at completion is a fundamental aspect when managing a group of projects in a coordinated way. The thesis focuses on the development and implementation of statistical linear predictive models to enhance cost estimation accuracy and precision in comparison to the most widely used index-based methods, thereby enabling more informed decision-making, proactive cost control, and practical implementation guidance. The research begins with a comprehensive review of existing literature and practices related to project and programme management, cost estimation, and statistical predictive modelling. The already existing approaches and examined techniques are helpful for developing the necessary methodology framework whose dimensions will be applied to a 10-projects porgramme case-study belonging to the software testing IT practice. To this end, the projects under examination have been firstly described and three EVM datasets created to analyze the full programme observations and, in addition, two of its subcategories determined by projects’ division in relation to allocated teams. Firstly, correlation analysis have been conducted to initially assess the relationships between variables and testing PMO’s hypothesis. Next, multiple linear regression analysis are performed. The shrinking and regularizing LASSO selection procedure was used to determine the number of regressors over twelve initial candidate for each of three scenarios, whereas the general linear model tests the overall statistical significance and calculates parameters estimates. Finally, accuracy and precision of the fitted models have been assessed by the Mean Absolute Error and Standard deviation against seven index-based forecasting methods resulting to be always the top-ranked. By looking at the diagnostics of fit, considerations on the three models’ consistency are presented too. The thesis concludes with outlining limitations and hints for future research.