Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MA20

2 - Carpool Services for Ride-sharing Platforms Renyu Zhang, New York University Shanghai, 1555 Century Avenue, Shanghai, 200122, China, Xuan Wang We study the carpool services of a ride-sharing platform. The carpool services allow passengers heading towards the same direction to share a ride at a discount fare. We study the operational issues of a ride-sharing platform in the presence of carpool services. We show that carpool services enable the platform to achieve a larger market coverage and lower prices. Using the operational data from San Francisco, we find that carpool services could also reduce the price variabilities riders face. As two operational leverages to match supply and demand, carpool services and surge pricing are complements when demand-supply ratio is large or Ozan Candogan, University of Chicago, Booth School of Business, Chicago, IL, 27708, United States, Baris Ata, Alexandre Belloni Agents in a social network consume a product that exhibits positive local network externalities. A seller has access to data on past consumption decisions/prices for a subset of observable agents, and can target these agents with discounts. The observable agents potentially interact with additional latent agents, who can purchase the same product from a different channel. Observable agents influence each other both directly and indirectly through the influence they exert on the latent part. The seller does not know the underlying network structure. We provide algorithms that allows the seller to estimate the influence structure from the available data, and improve her pricing decisions. 4 - Real-time Spatial Dynamic Pricing for Balancing Supply and Demand in a Network Qi Chen, London Business School, London, United Kingdom, Yanzhe Lei, Stefanus Jasin Motivated by recent expansion of mobile ride-hailing apps in the taxi industry in big cities, we study a real-time spatial dynamic pricing problem where a firm who uses many units of reusable resources (e.g., taxis) in a network to serve price- sensitive customers who arrive over a finite selling season (e.g., one day) in a stochastic and nonstationary fashion. For any origin-destination pair, the quoted price equals a nominal price times an origin-specific price multiplier. The firm can dynamically change quoted prices by adaptively adjusting the price multipliers over time. We develop a Network Balancing Control that has asymptotically optimal performance and discuss some extensions. n MA22 North Bldg 130 Revenue Management under Consumer Choice Behavior Sponsored: Revenue Management & Pricing Sponsored Session Chair: Zizhuo Wang, University of Minnesota, Minneapolis, MN, 55414, United States 1 - Consumer Choice and Market Expansion: Modeling, Optimization and Implementation Ruxian Wang, Johns Hopkins University, Carey Business School, 100 International Dr, Baltimore, MD, 21202, United States The market size, measured by the number of people who are interested in the products, may be highly affected by the management strategy. We refer to this effect as market expansion. In this paper, we incorporate market expansion effect into consumer choice models and to investigate the problems on assortment, pricing and estimation. 2 - Irrational Behavior Modeling and Decision Making Yi-Chun Chen, UCLA Anderson School of Management, Los Angeles, CA, United States Customer preferences are often assumed to follow weak rationality, which assumes that adding a product to an assortment will not increase the choice probability of a product already in that assortment. In this paper, we study a new choice model that relaxes this assumption and can model a wider range of customer behavior, such as anchoring effects between products. We develop efficient procedures for model learning and subsequent decision making. Using synthetic and real data, we show that the model can better predict customer behavior and lead to higher revenue. 3 - Space Constrained Assortment Optimization under the Paired Combinatorial Logit Model Jacob Feldman, Olin Business School, United States We study the space constrained assortment optimization problem under the paired combinatorial logit choice model. The goal in this problem is to choose a set of products to make available for purchase with the intention of maximizing the expected revenue from each arriving customer. Each offered product occupies a specific amount of space and there is a limit on the space consumed by all of the offered products. The purchasing decision of each customer is governed by the paired combinatorial logit choice model. We provide the first efficient constant factor approximation for this problem. small, but are substitutes when demand-supply ratio is moderate. 3 - Latent Agents in Networks: Estimation and Pricing

n MA20 North Bldg 129A Marketplaces Sponsored: Revenue Management & Pricing Sponsored Session Chair: Xuanming Su, University of Pennsylvania, Philadelphia, PA, 19104, United States 1 - Ride-hailing Networks with Strategic Drivers: The Impact of Platform Control Capabilities on Performance Zhe Liu, Columbia Business School, New York, NY, USA, Philipp Afeche, Costis Maglaras We study the performance impact of two operational controls, demand-side admission control and supply-side repositioning control, in ride-hailing networks with strategic drivers. We characterize the system equilibria for various control regimes, show how the performance gains from these controls depend on the capacity and the network demand imbalance, and provide new results on how admission control affects drivers’ repositioning decisions. 2 - Referral and Learning on Social Network: Implications for Inventory Guangwen Kong, University of Minnesota, 111 Church Street SE, Minneapolis, MN, 55414, United States, Yuanchen Su, Ankur Mani We consider a firm selling differentiated products to customers whose preferences are correlated in a social network. We investigate how the network structure and customer learning influence the demand distribution and therefore have an impact on the firm’s inventory decision and the optimal design of referral programs 3 - Managing Market Thickness in Online B2B Markets Wenchang Zhang, University of Maryland, College Park, MD, 20742, United States, Kostas Bimpikis, Wedad Jasmine Elmaghraby, Kenneth Moon Platforms can obtain sizable returns by operationally managing their market thickness (i.e., the availability of supply-side inventory). Using data from a natural experiment on a major B2B auction platform specializing in the $424 billion secondary market for liquidating retail merchandise, we find that thickening the platform’s market by closing auctions on certain weekdays increases its revenue by 7.3%, due to the bidders’ participation frictions. Using structural modeling, we then study two complementary design levers to control the market thickness: (i) its listing policy, which determines the ending times of auctions, and (ii) a recommendation system. 4 - Centralized versus Decentralized Routing in Ride-hailing Networks Xuanming Su, University of Pennsylvania, The Wharton School, 3730 Walnut Street, Philadelphia, PA, 19104, United States We build a spatial equilibrium model to study traffic flows in ride-hailing networks. We consider two regimes: drivers follow a central routing plan (i.e., centralized) or choose their own directions (i.e., decentralized). Our results measure the value of centralized routing. n MA21 North Bldg 129B Emerging Topics in Revenue Management Sponsored: Revenue Management & Pricing Sponsored Session Chair: George Chen, UT Dallas, Dallas, TX 1 - Personalized Assortment Optimization with High Dimensional Customer’s Data Sentao Miao, University of Michigan-Ann Arbor, Ann Arbor, MI, 48104, United States, Xiuli Chao Motivated by online retailing with mass amount of customer’s data, we study the online personalized assortment selection problem. Because the customer’s data, such as browsing history, often has extremely high dimension, there are two challenges: the first is high time complexity; the second is performance of the algorithm (to maximize the total revenue). In this paper, we explicitly address these two challenges by combining an upper-confidence-bound (UCB) type algorithm with the so-called random projection method. We prove that our algorithm has low computational time and provably near-optimal performance. Numerical experiments show that our algorithm has great empirical performance.

126