Drug coating is one of the most important processes in the modern pharmaceutical industry. Improving the utilization rate of raw materials (URRM) in the drug coating process is thus important for cost saving and efficiency enhancement. There is little existing research on this topic in the literature of applied statistics. In this paper, motivated by a real dataset collected from a pharmaceutical company in China, we propose to use a novel predictive model that integrates a Bayesian framework with the Gibbs sampling algorithm to characterize the pattern of URRM. Based on certain prior distributional assumptions, the Gibbs sampling algorithm is then applied to sample the posterior distribution of the parameters to obtain more accurate and robust estimation results. By using the proposed method, the drugs can be properly separated into several categories with different patterns of URRM, and the pattern of each category can be properly recognized with selected covariates, which achieves the goals of clustering, variable selection, and regression simultaneously, and provides valuable insights into the improvement of the URRM for drug coating. Numerical studies show that the proposed method works well in practice.