Contextual segmentation: using context to improve behavior predictive models in e-commerce

In e-commerce, where the search costs are low and the competition is just a mouse click away, it is crucial to accurately predict customer purchasing behavior in order to offer more targeted and personalized products and services. Recent research has demonstrated that including the context in which a transaction occurs in customer behavior models improves their predictive performance, especially when studying individual customer behavior. However, several practical and managerial issues can arise, thus driving companies to focus on segments rather than on individuals. The main contribution of this work lies in presenting a conceptual framework to incorporating context when building predictive models of market segments, and in comparing different approaches, across a wide range of experimental conditions. Our experiments show that the most accurate approach is not the most efficient from a managerial perspective. Our findings provide insights of how companies can exploit context at best to support marketing decision-making.