This thesis documents the redesign of supplier quality performance evaluation system at Ariston Group. It first clarifies the existing KPI and governance landscape, then examines the legacy Vendor Rating and Business Warning models, highlighting weaknesses related to evaluator dependence, coarse scaling, and volatility under low exposure. The redesign removes evaluator dependent artifacts by excluding the count of 8D reports, focuses on transactional indicators, namely Parts Per Million and Weighted Supplier Incidents, stabilizes skewed metrics through logarithmic transformation, and applies Bayesian shrinkage to temper noise for low-volume suppliers. The resulting Vendor Rating is more discriminative, intepretable, and statistically more stable, while the Business Warning logic becomes more proportionate and fair.