Media exposures can have effects that do not always lead to the immediate purchase of a brand. Some media are effective at initiating search and trial, while others are more effective at promoting purchase once search and trial have taken place. The idea of intermediate communication effects have long been posited in textbooks and academic literature, but their practical existence has not been shown in quantitative models. This paper proposes a hierarchical Bayesian model for identifying media response segments in cross-sectional data that differ in their likelihood of purchase, and shows that effects are obscured in aggregate analyses that attempt to directly relate media exposure to purchase. Data from a national brand-tracking study are used to illustrate our model, where we find large intermediate media effects.