In this paper we consider the role of variety attributes in an agricultural household model of variety planting decisions. In an application to banana production in Uganda we derive a system of derived demands for a set of available banana varieties. Our empirical model uses a hudle/count data framework to examine simaltaneously the likelihood a household has experience with a given variety, and the amount of the variety that is planted. We find that production, consumption, and pest resistance attributes significantly influence planting decisions. These findings have implications for emerging crop biotechnologies.