Micellar casein concentrate (MCC) and serum protein (SP) isolate (SPI) can be made from skim milk (SM) using ceramic (C) or spiral wound (SW) polymeric microfiltration (MF) membranes. MCC can be made at different purities (i.e., different SP removal). The purpose of this work was to demonstrate the relationship between values for SP removal and MCC purity. High quality raw SM contains about 3.20% true protein (TP), 2.624% casein, and 0.576% SP and contains about 82% casein as a percent of TP (CN%TP). SP removal (%) with SW and C MF membranes differs, with C membranes having higher SP removal in a 3 stage, 3X MF process at 50°C (72 versus 95% respectively). SP removal is expressed as total Kg SP in permeate removed divided by Kg SP in the original skim multiplied by 100 (measured by Kjeldahl). MCC purity is often expressed as CN%TP. A MF process at 72% versus 95% SP removal in a factory processing 908,000 kg of SM in a day would have a higher yield of 3768 versus 4967 kg of SP per day and a MCC purity of 94.2 and 98.9% CN%TP, respectively. Differences in MCC concentrate purity will have different effects depending on the application (e.g., cheese making versus shelf stable beverages). In retorted or ultrapasteurized beverages, flavor is important. Lower purity of MCC in high-heat-treated beverages may cause more heat induced off flavors due to thermal degradation of milk SP. The lower purity MCC (94.2 versus 98.9%) has a 5 to 6 fold higher SP concentration. The impact of MCC purity is magnified in higher protein beverages as the absolute concentration of the SP increases. Heat induced SP degradation products (e.g., H₂S, methional) have very low sensory thresholds and higher concentrations of residual SP in MCC will be correlated with higher intensities of heat induced off flavors. In addition, the amount of SP in the final MCC will be influenced by thermal denaturation of SP in the pasteurization skim milk before MF, regardless of the type of membrane. Minimum HTST pasteurization will have an apparent increase in CN%TP from 82 to about 83% by Kjeldahl due to SP denaturation. Heat induced covalent binding of SP to casein micelles causes over estimation of MCC purity when measured by classical Kjeldahl methods, so HPLC or SDS-PAGE would provide more correct determination of MCC purity.