The robustness of efficiency estimates depends on theoretical consistency of models from which those estimates are developed; functional forms of the variables must be globally consistent with theoretical properties regarding feed utilization for maintenance and gain in growing and finishing cattle. Model parameter estimates and their dimensions must be unique or estimates of feed utilization and gain will not reflect reality. A linear equation commonly used to estimate daily DMI by the th individual animal (ADFI), based on mean weight and gain during a feeding period, was evaluated to determine if that model was correctly specified and if the vector predicted ADFI differed from the vector observed ADFI. Three independently gathered data sets were evaluated using a multiple linear regression model; variability described by that model failed to capture observed variability in the data (lack of fit, < 0.10), and predicted ADFI differed from observed ( < 0.05); for 1 of the 3 data sets, residuals were not normally distributed ( < 0.001). Functional forms of the variables in the first model evaluated, characterizing ADFI required for maintenance ( × BW) and gain ( × ADG), were consistent with neither published empirical nor theoretical relationships among ADFI, BW, and ADG. Parameter estimates determined for that linear model were not BLUE. Better fits among final BW, initial BW, and ADFI were found for a first-order relationship, in which final BW was a function of initial BW and ADFI, as indicated by > 0.90. The linear model and, to a lesser degree, the first nonlinear model lacked theoretical and global consistency. A second nonlinear model, which described retained energy as a function of ME intake, best fit the data, and functional forms of variables describing ME intake at maintenance and the efficiency of ME utilization for gain were consistent with theoretical estimates found in the literature. Changes in feed intake and live BW in linear and nonlinear models failed to adequately describe efficiencies of metabolic processes, which are better characterized by changes in retained energy as a function of ME intake in nonlinear models.