The need to control the ultimate particle size distribution (PSD) of a crystalline material subject to tight double-sided specifications is a problem that may be dealt with by a number of strategies. The use of a mill to provide the necessary output PSD when the input material exhibits considerable variability in its size characteristics suggests the use of a feedback control scheme, such as an on-line particle size analyzer to control the rate at which unmilled material is fed. For certain systems, this may be an undesirable tactic due to installation expense, delays associated with retrofitting the existing process train, or the need for extensive calibration. We present here a multivariate feedforward approach to mill control, based upon three measurements that represent the history of the unmilled material, and two control parameters of the mill itself. The need to use a multivariate approach is discussed from the perspective of variability in the unmilled material and the need to provide a robust solution to a “short and fat” problem – many variables combined with a relatively small number of observations.