We develop a model to understand and predict customers’ observed multichannel behavior in a customer support setting. Using individual-level data from a US-based health insurance firm, we model a customer’s query frequency and choice of using the telephone or web channel for resolving queries as a stochastic function of her latent “information stock”.
The information stock is a function of the customer’s “information needs” (which arise when customers file health insurance claims) and “information gains” (which customers obtain when they resolve their queries through the telephone and web support channels), and other factors such as seasonal effects (for instance, queries that arise at the time of annual contract renewal).
We find that average information gain from a telephone call is twice as much as that from visiting the web portal; customers prefer the telephone channel for health event-related information but prefer the web portal for structured seasonal information; and customers are polarized in their propensities of using the web channel and can be broadly classified into “web avoiders” and “web seekers.” Our model provides superior in-sample and out-of-sample fit than multiple benchmark models for aggregate and individual-level customer activity and has several managerial uses, such as capacity planning.
Joint work with Kinshuk Jerath (Columbia University) and Anuj Kumar (University of Florida)