Injuries and fatalities resulting from workplace accidents remain a global concern within the construction industry. While education and training of personnel offer well known approaches for establishing a safe work environment, Serious Games in Virtual Reality (VR) is being increasingly investigated as a complementary approach for learning. They yet have to take full advantage of the inherent data that can be collected about players. This research presents a novel approach for the automated assessment of players data. The proposed method gathers and processes the data within a serious game for instant personalized feedback. The application focuses on close calls and contact collisions between construction workers and hazards like equipment, harmful substances, or restricted work zones. The results demonstrate the benefits and limitations of safety information previously unavailable, or very hard or impossible to collect. An outlook presents work ahead for practical implementation in existing risk management processes.