Joint Inventory Allocation and Assortment Personalization with Performance Guarantees
In this talk, we give approximation algorithms for a joint inventory allocation and assortment personalization problem motivated by an online retail setting. In our problem, we have a limited amount of storage capacity that needs to be allocated among multiple products to serve customers that arrive over a selling horizon. At the beginning of the selling horizon, we decide how many units of each product to stock. Over the selling horizon, customers arrive at the platform one by one to make a purchase. Based on the remaining inventories of the products and the information available on the arriving customer, we offer a personalized assortment of products to each customer. The customer either makes a choice within the offered assortment or leaves without a purchase. Our goal is to decide how many units of each product to stock at the beginning of the selling horizon and to find a policy to figure out which personalized assortment to offer to each arriving customer to maximize the total expected revenue over the selling horizon. Our problem is motivated by same-day-delivery applications in online retail, where the retailer needs to allocate the limited storage capacity in an urban warehouse among different variants in a product category, while having the capability of offering personalized assortments to customers to make better use of remaining inventories. Allocating the storage capacity among the products requires tackling a combinatorial problem, whereas finding an assortment personalization policy requires approximating a dynamic program with a high-dimensional state variable. When the choices of the customers are governed by the multinomial logit model, we give a constant-factor approximation algorithm for this joint inventory allocation and assortment personalization problem. Under a general choice model, we give an algorithm that is asymptotically optimal as the storage capacity gets large. In the latter result, the demand can be scaled in an arbitrary fashion along with the storage capacity. This is joint work with Yicheng Bai, Omar El Housni and Paat Rusmevichientong.
Huseyin Topaloglu is the Howard and Eleanor Morgan Professor in the School of Operations Research and Information Engineering at Cornell Tech. He holds a Ph.D. in Operations Research and Financial Engineering from Princeton. His recent research focuses on constructing tractable solution methods for large-scale network revenue management problems and building approximation strategies for retail assortment planning. Huseyin Topaloglu is currently serving as an area editor for Analytics in Operations area at Manufacturing and Service Operations Management.