This study introduces a novel perspective on project network analysis by incorporating network theory and graph analysis to identify network-level indicators for activity-on-node project networks. A key contribution lies in the comparison between real and synthetic projects on the basis of project and network-level indicators, revealing distinct variations. The findings underscore that real projects demonstrate lower flexibility and efficiency than synthetic counterparts, impacting project scheduling and resource allocation. Moreover, the study suggests employing supervised and unsupervised learning classification methods to categorize projects on the basis of indicator values, enhancing project selection and prioritization processes. By bridging the gap between network-level indicators and project aspects, this research enriches the literature by offering fresh insights and resources for project managers to optimize project network structure and performance.