Informs Annual Meeting 2017

MD03A

INFORMS Houston – 2017

MD03A Grand Ballroom A

SIR, cost sharing can also be viewed as benefit sharing, under which we introduce a natural definition of sequential fairness—-the total incremental benefit from a new customer is shared among existing customers in proportion to the incremental inconvenience suffered. We demonstrate the effectiveness of these notions by applying them to a ridesharing system, where unexpected detours taken to pick up subsequent passengers inconvenience the existing passengers. 4 - A Tale of Timescales: Transient Network Dynamics, Demand Hot Spots, and Surge Pricing We consider an online platform that operates a ride-hailing network with price and delay sensitive passengers and strategic drivers that supply processing capacity. We study the interplay between demand shocks and surge pricing that is meant to a) moderate demand, and b) incentivize supply to reposition to demand hot spots, exploring the impact of travel delays on drivers’ responsiveness to the surge signals. MD03C Grand Ballroom C Empirical Studies on Pricing Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Antonio Moreno-Garcia, Kellogg School of Management, Evanston, IL, 60201, United States, a-morenogarcia@kellogg.northwestern.edu 1 - The Hidden Costs of Dynamic Pricing Empirical Evidence from Online Retailing in Emerging Markets Richard Zhiji Xu, PhD Student, Kellogg School of Management, Evanston, IL, United States, zhiji-xu@kellogg.northwestern.edu, Antonio Moreno-Garcia, Chaithanya Bandi We investigate how dynamic pricing can lead to higher operational costs through more product returns in the online retail industry. Using detailed sales data of more than 2 million transactions with return information from the Indian online retail market, we document two types of strategic customer behavior, opportunistic returns and strategic choice of payment method, which have not been considered by previous research. 2 - Managing Market Thickness in Online B2b Markets Wenchang Zhang, Robert H. Smith School of Business, 7621 Mowatt Lane, PHD Carrel, College Park, MD, 20740, United States, wzhang@rhsmith.umd.edu, Kostas Bimpikis, Wedad Jasmine Elmaghraby, Kenneth Moon Excess inventory amounts to $424 billion a year, and much of this inventory is sold through online auctions. Based on a natural experiment on a major B2B auction platform, we find that increasing the market thickness by concentrating the auction ending times to certain days of the week has a significant positive effect on the platform’s revenues. This finding illustrates the role of listing policies in matching supply to demand and provides evidence of trading frictions. We develop a structural model, which characterizes the bidders’ equilibrium strategies. The counterfactual analysis provides a systematic way for the platform to design its listing policy and other market design levers. 3 - Customer Perceptions of Dynamic Pricing in on Demand Services Nil Karacaoglu, Kellogg School of Management, 531 Grove Street, Evanston, IL, 60201, United States, n- karacaoglu@kellogg.northwestern.edu, Antonio Moreno-Garcia On-demand services adopt dynamic pricing practices to match supply and demand. We investigate how customers perceive the trade-offs between increased prices, waiting time, and unavailability. We examine ways of reducing negative effects of dynamic pricing on customers’ fairness perceptions and satisfaction by controlling price and the supply representation. 4 - Dynamic Pricing of the Ride Sharing Market in a Spatial Search Model Weiming Zhu, IESE Business School, Barcelona, Spain, zhuwm923@gmail.com, Jingting Fan, Wenlan Luo, Liu Ming Ride-sharing platforms employ flexible pricing during peak hours to match supply with demand. We build a spatial search model to study the geographic dynamics among drivers. Utilizing data from a leading ride-sharing platform, we assess the impact of different pricing schemes on drivers’ capacity distribution, platform profit and consumer surplus. Zhe Liu, Columbia University, New York, NY, United States, zliu18@gsb.columbia.edu, Philipp Afeche, Costis Maglaras

Markov Decision Processes, Alpha Go, and Monte Carlo Tree Search: Back to the Future Invited: Tutorial Invited Session Chair: Jiming Peng, University of Houston, Houston, TX, 77204, United States, jopeng@Central.uh.edu Co-Chair: Rajan Batta, University at Buffalo (SUNY), 410 Bell Hall, Buffalo, NY, 14260, United States, batta@buffalo.edu 1 - Markov Decision Processes, AlphaGo, and Monte Carlo Tree Search: Back tothe Future In 2016, a computer Go-playing program called AlphaGo stunned the (human) world by winning a match (4 games to 1) against the reigning human world champion, a feat more impressive than previous victories by computer programs in chess (Deep Blue) and the TV game show Jeopardy (Watson). The main engine behind AlphaGo combines machine learning approaches in the form of deep neural networks with a technique called Monte Carlo tree search, whose roots can be traced back to an adaptive multistage sampling simulation-based algorithm for Markov decision processes (MDPs) published in Operations Research back in 2005 (and introduced even earlier in 2002). This tutorial describes AlphaGo and the simulation-based MDP algorithm, as well as providing contextual and historical background material for both, and uses simple examples to illustrate the main ideas behind Monte Carlo tree search. MD03B Grand Ballroom B Operations and Economics of Ride-Hailing Networks Sponsored: Revenue Management & Pricing Sponsored Session Chair: Costis Maglaras, Columbia University, New York, NY, 10027, United States, c.maglaras@columbia.edu Co-Chair: Philipp Afeche, University of Toronto, Toronto, ON, M5S 3E6, Canada, afeche@rotman.utoronto.ca 1 - Designing Incentives to Scale Marketplaces Ashish Kabra, INSEAD, Boulevard de Constance, Fontainebleau, 77305, France, ashish.kabra@insead.edu, Elena Belavina, Karan Girotra Marketplace operators run aggressive incentive schemes to achieve scale, that is key to the efficacy, survival and eventual domination of a marketplace. This study quantifies and compares the effect of incentives given to the “buyer” side and “seller” side using data from a leading ride-hailing market. We build a structural model to accurately capture the driver and passenger response to incentives, and the nature of incentives. Driver effort on the platform is unobserved, for which we devise a novel local matching model based imputation method. We find that in short-term (current week) passenger incentives are more effective while the opposite is true in the long-term (next 3 months). 2 - We are on the Way - Analysis of On-demand Ride-hailing System Michael Fu, University of Maryland-College Park, Robert H.Smith School of Business, Van Munching Hall, College Park, MD, 20742, United States, mfu@rhsmith.umd.edu Zizhuo Wang, University of Minnesota, 1009 5th Street SE, Minneapolis, MN, 55414, United States, zwang@umn.edu, Guiyun Feng, Guangwen Kong In this paper, we build a stylized model to analyze the efficiency (average waiting time of passengers) of on-demand ride-hailing systems and compare it to the traditional street-hailing systems. We find that the on-demand matching mechanism could result in higher or lower efficiency than the traditional street- hailing mechanism, depending on the parameters of the system. We further propose adding response caps to the on-demand hailing mechanism and develop a heuristic method to calculate a near-optimal cap. 3 - The Costs and Benefits of Sharing: Sequential Individual Rationality and Fairness Ragavendran Gopalakrishnan, Research Scientist, Conduent Labs India, Etamin Block, 4th Floor, Wing - A, Prestige Technology Park.- II, Bangalore, 560103, India, ragavendran.gopalakrishnan@conduent.com, Koyel Mukherjee, Theja Tulabandhula We introduce the notion of sequential individual rationality in dynamic shared service systems, that requires that the disutility of existing customers is nonincreasing as the system state changes due to new customer arrivals. Given

236