As traditional brick-and-mortar retailers are expanding their sales channels to online channels, consumer adoption rate of these emerging channels is growing over time. In addition, consumers’ perception of the retailer’s service quality (including inventory service levels) is turning into an important determinant of their purchasing decision, as today’s consumers can easily obtain information on retailers’ service quality through online forums, reviews, retailer’s website, and/or social networks. This dynamically changing environment brings in new challenges for dual-channel retailers’ inventory planning process, and important questions arise: Do products at different phases of e-commerce adoption necessitate different service level strategies? Does the retailer’s optimal stocking policy change based on how consumers predict the retailer’s service levels prior to making their purchasing decision? When is it profitable to offer an online channel?

In this talk, we present a novel model to study the impact of dynamically evolving e-commerce adoption on a dual-channel retailer’s service levels by considering that consumers utilize their past experiences as well as publicly available information provided by other consumers in order to form expectations on the retailer’s service levels. For benchmarking purposes, we also study the setting in which consumers have perfect information on the retailer’s current service levels through, for example, utilizing product availability information provided on the retailer’s website.

We study structural properties of the retailer’s optimal solution and show that the retailer’s optimal service level in each time period follows a threshold policy. Using the optimal policy, we show that the retailer’s service levels vary over time due to the fact that consumers do not have perfect information on current availability and need to predict it. Further, both monotonic and non-monotonic patterns are possible for the retailer’s optimal service levels over time, depending on the speed of consumer learning. We also find out that the retailers that suffer most from inaccurate modeling of consumer behavior are those who operate with long online delivery times and within large service radii. Finally, our study indicates that introduction of an online channel is profitable only if the retailer is unable to capture the entire market through her store channel due to a large service radius.