The aim of this study was to investigate the association between Mozzarella cheese yield (MCY) and variables of milk composition, processing and the recovery of whey constituents by multivariate analysis. The study involved tracking the processing of 30 lots of buffalo Mozzarella cheese in a dairy industry of northeast Brazil. The variables milk fat (MF), true protein (TPRO), casein (CAS), lactose, total solids (TS), solids-not-fat content (SNF), density (DS), cryoscopy, pH, titratable acidity (ACID), and somatic cell score (SCS) of raw milk were measured before processing. The variables pH, acidity, age of starter culture (ASC), volume of calcium chloride, volume of rennet, average pH of curd during stretching, coagulation time, time between cuts, fermentation time (F1T), stretching time for the whole curd, whole fermentation time, fat (FREC), protein, casein, lactose, total solids, and solids-not-fat recovery were measured during the processing and in the whey. MCY association to the variables was verified by PROC PRINCOMP procedure of SAS. The explained variability of PC1, PC2 and PC3 was 26.37%, 17.38% and 12.44% respectively, totaling 56.18%. A direct association between milk characteristics TS, SNF, CAS, TPRO, MF, ACID and FREC was observed, as well as an antagonistic association between them vs. MCY, pH vs. DS, and F1T vs. ASC. This means that low kg milk per kg cheese ratio is lower when using buffalo milk with higher concentrations of TS, SNF, CAS, TPRO and MF, and when there is greater loss of MF in whey. A direct association can be found between those representing the loss of non-fatty constituents in whey and SCS. Thus, the volume of cheese obtained may be lower when milk with higher SCS is used. The main components of Mozzarella indicated that the yield has more relevant associations with pH, DS, ASC, time elapsed between curd cuttings, and curd stretching time, indicating that these elements must be well controlled to achieve optimal efficiency in the manufacturing of this cheese.