Complex, ill-defined mixtures of natural origin are often used as nutrients in the production of biological products through microbial fermentation. Product yields are affected by variation in these natural products. It was desired to examine near-infrared spectroscopy as a rapid screening tool for qualifying raw material lots. Specifically, the characterization of yeast extract was investigated. The model system consisted of a Merck & Co., Inc., microbial fermentation process. Cell mass and specific product yields are dependent upon variations in the yeast extract used in the medium. Partial least-squares regression on the second-derivative spectral absorbances of various yeast extracts in the ranges 1150-1380, 1554-1826, and 2100-2300 nm resulted in the development of models with multiple correlation coefficients of 0.99 for cell mass yields and 0.96 for specific product yields in large-scale fermentations. These models could also be used to predict cell mass yields in 15 L batch fermentations and specific product yields in 2-L shake flasks.