Netvendor: Newsvendor Selling to a Social Network

Abstract

We study the inventory problem faced by a newsvendor, referred to as a "netvendor," when selling a product of network effects to customers located on a social network. The customers decide on product adoption subject to product availability, from which they derive values not only from the product itself but also from the adoptions by their network neighbors. Both the customers and the netvendor only have distributional information regarding the underlying network structure. Each individual customer also has local information on the exact number of their connected neighbors. The random demand for the product is driven by the fashion trend that determines the customers' intrinsic product valuation. The netvendor needs to commit to an inventory level before observing the realization of this intrinsic product value. We provide complete and closed-form characterizations of the equilibrium of the customer adoption game, as well as the optimal inventory level for the netvendor. In the context of a scale-free network characterized by its balancedness and density, we show that the netvendor benefits (both in terms of the optimal profit and sales volume) from selling to a more unequal network with stronger unbalancedness in connectivity. We find that for a netvendor, increasing network density will lead to not only higher expected profit and order quantity but also rising expected waste (i.e., leftover inventory) at optimum, all in a convex manner. The expected waste, however, can increase or decline with network balancedness depending on the comparison of topological and economic indicators. Analogous comparisons also determine whether the effects of network density and network balancedness are substitutable or complementary in various monotonicity results of performance measures. Moreover, if a netvendor such as Shein is equipped with quick response capability to operate in the make-to-order mode as opposed to the base model of the make-to-stock mode, we show that the value of quick response capability convexly increases with the network density. The value of quick response can either rise or fall with increasing network balancedness, depending on the relative strengths of topological and economic indicators, even though the netvendor without quick response always benefits from increasing network unbalancedness.