High-Mix/Low-Volume (HM/LV) semiconductor fabs operate under ever-changing product portfolios and frequent recipe change, where maintaining yield requires flexible and adaptive process control and operations management. In traditional practices, qualification remains static, process control is reactive, and process optimization relies on human expertise. In contrast, the emerging trends tend to make qualification dynamic, control predictive, and optimization automated and data-driven. This work leverages end-to-end data integration across Electrical Wafer Sort (EWS), Statistical Process Control (SPC), and Fault Detection and Classification (FDC) domains (EWS -> SPC -> FDC) to provide a unified process view. Building on prior research that revealed SPC-EWS interdependencies (i.e., critical process steps or unit operations), FDC data are now integrated into a hierarchical framework to identify necessary and sufficient process parameters that require adjustment for yield improvement and to guide a recipe-level optimization framework constrained by mathematical consistency and physical plausibility for individual products. The proposed approach is assessed based on real-life data, and the resulting industrial benefits are explicitly discussed.