Process quality control of the herbal medicines (HMs) still hits technical bottlenecks due to their complicated multi-step pretreating. Nuclear magnetic resonance (NMR) spectroscopy could serve as a powerful technique able to obtain comprehensive information of chemical compounds in the manufacturing process of HMs by a single analytical test. In this study, ethanol precipitation, a critical pretreating process of an especially widely used HM preparation, was conducted with different process variables, including ethanol-material ratio, apparent ethanol concentration and moisture concentration of the raw material, 40 lab-scale experiments of ethanol precipitation were orchestrated by a full factor design with additional central composite circumscribed experiments. Then, 16 metabolites were recognized as the most critical potential chemical markers in the process by H-1 NMR spectra and orthogonal partial least squares-discriminant analysis (OPLS-DA). Finally, multiple linear regression (MLR) was applied to evaluate the impacts of process variables and their interaction on the chemical markers and explain the consistency variation of the process intermediates. The results promote the further understanding of ethanol precipitation process and open up possibilities to develop a more effective process quality control methodology of an HM, through providing a more in-depth and improved pharmaceutical process understanding and characterization.