ProdSight: A data-driven decision support system for improving software engineering team productivity using software quality indicators

Modern software development generates vast workflow traces through issue-tracking systems, yet managerial decision-making relies on aggregate indicators that obscure lifecycle inefficiencies. Existing approaches remain either purely descriptive or predictive without structured decision support. This paper presents an AI-based Decision Support System empirically validated on 936,943 Apache JIRA issues spanning 2000–2024. Using 23 workflow-aware features and chronological train-test splitting, CatBoost achieved the strongest performance (mean absolute error (MAE) = 807.77 hours; coefficient of determination (R²) = 0.716). By integrating predictive modeling, uncertainty quantification, and actionable risk signaling, the system transforms software analytics into a scalable framework for proactive workflow optimization.