Supplier Evaluation Using Multivariate Process Capability with Preserved Batch Heterogeneity

Evaluating supplier's process capability often involves multiple correlated quality characteristics. The industrial internet enables buyers to timely access batch‐identifiable process information, thereby explicitly characterizing the pooled sample's batch‐heterogeneity structure and providing a basis for more accurate evaluation methods. This challenges existing approaches that rely on batch‐homogeneity assumptions. We propose an Exact Null Scaling (ENS) bootstrap framework: from observed samples, we extract a batch‐heterogeneity template and, without altering the mixing structure, apply exact scaling along the specification direction to align suppliers’ process capability with the null hypothesis. Supplier comparison then proceeds via a parametric bootstrap with Monte Carlo testing. Theoretical analysis and type I error calibration show that, with homogeneous batches, all methods agree. The classical F‐test tends to be liberal under covariance heterogeneity and conservative under mean heterogeneity. The nonparametric bootstrap is overall conservative. The ENS‐bootstrap consistently maintains the nominal significance level across capability indices, sample sizes, and heterogeneity types. Using exact k ‐scaling, we construct power curves under a fixed heterogeneity template and confirm ENS's advantages in both statistical power and computational efficiency. Finally, a case study on a compressor scroll component illustrates the practical applicability of the proposed method.