André de Andrade, PauloVanhoucke, MarioMartens, Annelies2024-06-102024-06-1020240254-533010.1007/s10479-023-05269-7http://hdl.handle.net/20.500.12127/7491Forecasting an ongoing project’s actual duration is an essential aspect of project management which received considerable attention in the research community. In studies using Earned Value Management forecasting, it has been argued that the network topology is a driver to indicate the accuracy of these forecasts. However, a new project indicator has been recently defined, i.e. the project regularity, which reflects the value accrue according to the plan. It has shown to outperform the serial/parallel network topology indicator in specifying the accuracy of project forecasts. This paper introduces a novel way to define the project regularity, which provides project managers with an improved indication of the expected forecasting accuracy for their projects. The study is carried out on an empirical database consisting of 100 projects from different sectors, and the results are compared to the academic literature. The experiments show that the new indicator provides a better categorisation compared to the existing approaches. Further, they have shown that the ability of project categorisers to indicate the expected forecasting accuracy is affected by industry sector and project size.enProject ManagementProject ForecastingEarned Value ManagementForecasting AccuracyEmpirical DatabaseProject RegularityEarned Duration ManagementAnnals of Operations Research1572-933858614