Characteristics models have been found to be useful in many areas ofeconomics. However, their empirical implementation tends to rely heavilyon functional form assumptions. In this paper we develop a revealedpreference-based nonparametric approach to characteristics models. Wederive the minimal necessary and sufficient empirical conditions underwhich data on the market behaviour of individual, heterogeneous, pricetakingconsumers are nonparametrically consistent with the consumercharacteristics model. Where these conditions hold, we show how informationmay be recovered on individual consumer’s marginal valuationsof product attributes. In some cases marginal valuations are point identi-fied and in other cases we can only recover bounds. Where the conditionsfail we highlight the role which the introduction of unobserved productattributes can play in rationalising the data. We implement these ideasusing consumer panel data on the Danish milk market.