This study presents a two-phase framework for optimizing retail workforce management. First, we develop machine learning models to forecast store foot traffic, incorporating temporal features, local events, and seasonal patterns. Second, we leverage these predictions to build a mathematical model for shift planning that addresses real-world retail constraints including employee availability, skill requirements, labor regulations, and service levels. We implement this model using Google OR-Tools' CP-SAT solver, which efficiently handles the complex constraint satisfaction problem. Our integrated approach enables retail managers to dynamically adjust staffing levels based on predicted demand, reducing both overstaffing during slow periods and understaffing during peak hours. The solution accommodates various scheduling constraints while producing practical shift plans. This research bridges predictive analytics and operational decision-making in retail workforce management, offering both methodological contributions and practical value through improved operational efficiency.