The emerging bottleneck in electronic commerce is that of converting vast amounts of customer behavior data into useful information. We view this as a problem of maximizing information liquidity – the rate at which organizations are able to transform the inherent information in a data set into an economically valuable action. We describe how to overcome this bottleneck, by presenting a model for maximizing information liquidity in electronic commerce. Our model is usable in a variety of situations. Specifically, when a large amount of transaction data already exists, the model is able to exploit this data to generate rules describing preferences that can be used to classify behaviors, and to subsequently map behaviors of noncustomers into known ones. Alternatively, where the predominant data available are about behaviors, the model can be used to cluster these behaviors and combine the resulting clusters with available transaction data to generate rules describing preferences. In both cases, the central question addressed is “when do I have enough information to make a meaningful offer?” Acting too early can result in inappropriate offers, while acting too late can result in missed opportunities. Good information and timing are therefore critical; the model in this paper is a first step in this direction.