Basket Composition and Choice among Direct Channels: A Latent State Model of Shopping Costs

Shoppers of multi-channel retailers often place orders using different channels on different shopping occasions. The differential use of channels is related to both basket composition and channel characteristics, such as the ability of the channel to provide additional information that resolves uncertainty about the purchase. In this paper, we examine the impact of basket composition on the choice among direct channels. We develop a two-stage, shopping cost model with two, latent states. Given a shopping basket, the shopper first decides if she needs additional information about items in the basket. If she is uncertain about the items in the basket meeting her needs, she uses an information rich channel, such as the retailer's website or call center, and risk reduction costs become salient in addition to the other shopping costs. If she does not require additional information, she places her order by choosing among all available channels, and she may incur a welfare loss from making a purchase that does not optimally meet her needs. We operationalize welfare loss with Shannon information and various metrics based on purchase history. Our empirical setting is a data set from a catalog retailer that offers multiple direct channels. Our estimates show that basket composition impacts channel choice. Large baskets shift to the Internet channel, suggesting that the Internet channel has lower ordering costs. High-risk baskets shift to call centers and this suggests that the call center has lower risk reduction costs. Collectively these estimates provide evidence for the notion of channel specialization—some channels are better at addressing certain shopping costs compared to others. Our estimates also show that electronic self-service channels have high initial access costs and a significant learning curve compared to the call center suggesting that these channels might be better suited to heavy users. We use the estimated model to quantify the value of channels, to identify categories that need risk reduction, and to segment and target shoppers for Internet ordering based on basket size and the potential to accumulate experience.