Associations of herd milk production and management variables to a return-over-feed (ROF) herd profit index were examined among 95 dairy farms. The ROF index is derived from 2 important determinants of profit on dairy farms: milk income and feed cost. All producers were participants in the Dairy Herd Improvement (DHI) ROF program in Ontario, Canada during 2002. Nutrition, housing, health, and other management data were collected through a phone survey of herd managers. Herd milk production, milk component percentages, and somatic cell count data were obtained from the Ontario DHI database. The linear regression model accounting for significant variation in ROF with highest R2 (0.66) included standardized milk production, milk protein percentage, milk fat percentage, and use of monensin in lactating cow rations. A 1-kg increase in standardized milk production (kg/d per cow) or a 0.1 percentage unit increase in milk protein was associated with $0.35/d per cow or $0.26/d per cow increase, respectively, in the ROF of the dairy herd. However, a 0.1 percentage unit increase in milk fat was associated with a $0.10/d per cow decrease in ROF, probably because of a negative association of milk fat with milk yield. Use of monensin in lactating cow rations was associated with a $0.39/d per cow increase in ROF. In a separate model (R2 = 0.27) that examined management factors independent of production variables, herds using 3 times daily milking had a $1.25/d per cow higher ROF vs. herds using twice daily, whereas use of an Escherichia coli mastitis vaccine was associated with $0.59/d per cow higher ROF. Production-related variables accounted for more variation in the ROF index than management variables, and the latter, e.g., use of monensin, only marginally increased R2 of production-based regression models.