<jats:p> Different objectives such as category demand expansion or market share stealing warrant the use of different marketing instruments. To help brand managers make informed decisions, it is essential that marketing mix models appropriately measure their effects. Random Utility Models that have been applied to this problem might not be adequate because they do not allow the effects of marketing instruments of one brand to spillover to preference for competing alternatives. Additionally, they have the Invariant Proportion of Substitution (IPS) property, which in some situations imposes counter-intuitive restrictions on individual choice behavior. Recognizing that effects of marketing instruments can spill across brands in a category, we propose an alternative choice model that relaxes the IPS property: the cross attributes flexible substitution logit model. We apply the model in two very different empirical settings, i.e., consumer choices of brands of refrigerated yogurt, and prescription-writing choices of physicians in the hyperlipidemia category. In both settings the proposed model provides consistent evidence that certain marketing instruments produce sales gains primarily from growing the category pie, while others produce gains from stealing share. By contrast, the random coefficient logit and generalized nested logit models both predict that gains from all marketing instruments would have similar sources. </jats:p>