Essays on Assortment Planning and Inventory Management for Substitutable Products

This dissertation consists of three essays which study assortment planning and inventory management of substitutable products motivated by different practical problems. Chapter 2 considers the assortment planning problem for a retailer who faces customers that buy multiple differentiated products (n-pack) on a store visit. We develop two choice models: the probabilistic choice rule which captures the heterogeneous consumer population choice pattern and maximum choice rule which captures the homogeneous consumer population choice pattern. We find that, under probabilistic choice rule, the optimal assortment is such that it includes a certain number of the most and least popular products. In contrast, under maximum choice rule, the optimal assortment does not have a fixed structure except that it is guaranteed to include the most popular product. We develop an algorithm under maximum choice rule which is shown to have good performance. In addition, we derive the structure of optimal assortment under both choice rules when a retailer ignores key features of n-pack choice model including choice premium and basket shopping behavior. We also conduct a numerical study where we show that ignoring these key features can lead to significant profit loss for a retailer. Chapter 3 explores the assortment planning for a firm who faces a two-sided market. That is, the firm receives revenues from two distinct user groups: the customers, who pay for the products it sells and the advertisers who pay to advertise their brand to the customers. We obtain structural properties of the optimal assortment. We also consider the case where the firm is allowed to offer multiple products with the same attractiveness profile and price. In this case, we obtain conditions under which the optimal assortment is made out of distinct products. In addition, we show that ignoring the revenue from the customers or the advertisers, or focusing only on one segment when making product assortment decisions can lead to a significant revenue loss; specifically, we derive the theoretical bound on revenue loss in these situations. Chapter 4 studies the decision making of an inventory manager who needs to decide order quantities of multiple substitutable products in his store. As such, the decision maker typically checks the sales history of the products. When there is stock-out, the sales history provides inaccurate information because the lost sales are unobservable and the sales from substitution are indistinguishable from first-choice sales, which we refer to as the “doublecensoring effect”. To study the impact of substitution rate and information amount on decision maker’s performance, we design an experiment where subjects need to decide inventory levels for 2 substitutable products in consecutive 30 periods. The experimental data shows that subjects underestimate the demand for high demand product and overestimate the demand for low demand product. Moreover, the bias is worse when there is substitution in fully censored information treatment. Also, when subjects are provided with less information, they tend to order larger quantity in early periods in order to learn demand.