In production environments where finished products (FPs) share identical specifications, grouping them under a master reference has traditionally been a manual and empirical process. However, this method lacks analytical rigor and overlooks critical trade-offs between operational efficiency and inventory risk. Initial efforts focused on clustering FPs based on shared technical characteristics, but subsequent analysis revealed an essential dilemma: full aggregation minimizes safety stock requirements due to consolidated demand profiles but simultaneously increases scrap rates due to mismatches in width utilization. Conversely, pursuing minimal scrap leads to fragmented master structures, causing safety stock inflation. To address this duality, a multi-objective optimization model based on the NSGA-II genetic algorithm was developed. This approach dynamically explores the trade-off frontier between total scrap and safety stock, generating a Pareto set of optimal configurations. Each solution enables supply chain planners to balance operational efficiency with inventory resilience, moving beyond manual decision-making towards data-driven strategies. The model delivers quantifiable benefits by identifying configurations that reduce waste while stabilizing inventory levels and is fully scalable to complex industrial scenarios. Ultimately, this project replaces subjective clustering with an intelligent optimization framework, demonstrating a significant step forward in supply chain design and operational excellence.