In this study, the classical first-order kinetic model was combined with a properly built statistical model based on a factorial experimental design, in order to accurately predict the rougher flotation efficiency for various flotation conditions. A three-level, three-factor experimental design was used to develop a statistical model to predict each of the kinetic model parameters as a function of the air flow rate, the feed grade and the froth thickness (or the pulp level). The statistical evaluation of the experimental results indicated that the ultimate recovery, the rate constant and time correction factor are not constant but each of these kinetic model parameters can be defined as a function of the variables considered. Furthermore, the rate of change in the kinetic parameters due to the feed grade fluctuation and their effects on the metallurgical performance can accurately be predicted by using the models developed. Thus, in order to reduce the detrimental effect of the feed grade fluctuations on the metallurgical performance, the operating variables of the flotation can be manipulated to obtain the desired concentrate grade. In addition, predictions with an error of less than 3.3% indicated that the versatility and viability of the classical first-order kinetic model could be improved by using the models developed. (C) 2003 Elsevier Ltd. All rights reserved.