Two principles for describing catalyst deactivation are discussed, one based on the deactivation mechanism and the other based on the activity and catalyst age distribution. When the model is based upon activity decay, it is common to use a mean activity developed from the steady-state residence time distribution. We compare control-relevant properties of such an approach with those of a model based upon the deactivation mechanism. Using a continuous stirred tank reactor as an example, we show that the mechanistic approach and the population balance approach lead to identical models. However, common additional assumptions used for activity-based models lead to model properties that may deviate considerably from the correct one.