Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SD20

exploding offer when the alternative offer market is favorable to the responder. When the alternative offer market is unfavorable to the responder, the proposer can profit from making an exploding offer with a smaller size in a harsher market. Finally, when the proposer is only allowed to manipulate offer size, the optimal offer size first decreases then increases in the given offer duration. 3 - Dynamic Segmentation to Explore Markets with Unknown and Time-varying Customer Heterogeneity N. Bora Keskin, Duke University, Fuqua School of Business, 100 Fuqua Drive, Durham, NC, 27708-0120, United States, Meng Li Consumers are often heterogeneous in their preferences for product quality, and firms usually face uncertainty about consumer preferences when they sell vertically differentiated products to such heterogeneous consumers. We study this problem in a setting where a firm can dynamically optimize its prices. We construct and analyze near-optimal dynamic pricing policies in this context. 4 - Online Learning and Optimization of (some) Cyclic Pricing Policies for Revenue Management with Patient Customers Huanan Zhang, Penn State University, 903B West Aaron Dr., State College, PA, 16803, United States, Stefanus Jasin We consider the joint learning and optimization problem of cyclic pricing policies in the presence of patient customers. We first introduce a learning algorithm that can converge to an optimal decreasing cyclic policy with a logarithmic regret, by only using the total sales information. Then, we introduce a larger family of policies, called threshold-regulated policies, which contains both the decreasing cyclic policies and the nested decreasing cyclic policies. For this broader set of policies, we introduce our second learning algorithm that can converge to an optimal threshold-regulated policy with logarithmic regret. n SD22 North Bldg 130 Computational Methods Sponsored: Revenue Management & Pricing Sponsored Session Chair: Dan Zhang Co-Chair: Saied Samiedaluie, University of Alberta, Alberta School of Business, Edmonton, AB 1 - Exact First-choice Product Line Optimization The first-choice product line design problem is to select a set of candidate products to offer to a set of customers that choose according to a first-choice rule. We present a new MIO formulation of the problem, prove that this formulation is stronger than alternative models and develop a large-scale solution approach based on Benders decomposition. Using real data, we consider a product line problem involving over 3000 candidate products and over 300 respondents and show that our approach can solve this problem to provable optimality in 10 minutes (compared to 1 week in prior work). 2 - Optimizing the Anheuser Busch Trailer Selection and Matching Problem AB delivers its beer to vendors via third party trucks. For each third party truck that arrives to the warehouse, AB must match this truck with a preloaded trailer of beer. The gross weight of the trailer and truck cannot exceed 80K lbs, which is set forth by federal and state law enforcement. The revenue for any matching is proportional to the actual weight of trailer loaded on the truck. We study the problems of choosing weights/inventory levels for the trailers and the problem of developing a policy to optimally match these trailers to arriving trucks. We show that revenue management approximate DP techniques can be employed to solve all three of the previously mentioned problems. 3 - An Approximate Dynamic Programming Approach to Queueing Admission Control Problems Saied Samiedaluie, University of Alberta, 3-40C Business Building, Alberta School of Business, Edmonton, AB, T6G 2R6, Canada, Dan Zhang We study a classical queueing control problem with multiple classes of customers. The queue is a loss system; i.e., arriving customers are rejected if all servers are busy. When a server is available, the decision is whether to admit an arriving customer and collect a lump-sum revenue. We model this problem as a continuous-time infinite-horizon dynamic program and solve it using approximate linear programming (ALP). We study several alternative approximation architectures and numerically investigate their policy performance. Our approach is potentially useful for a wide variety of queueing control problems. Velibor Misic, UCLA Anderson School of Management, Los Angeles, CA, 90095, United States, Dimitris Bertsimas Jacob Feldman, Olin Business School, United States, Panos Kouvelis, Xingxing Chen, Seung-Hwan Jung

n SD20 North Bldg 129A Revenue Management for Online Ad Markets Sponsored: Revenue Management & Pricing Sponsored Session Chair: Negin Golrezaei, Massachusetts Institute of Technology, 30 Memorial Dr, Cambridge, MA, 02142, United States Co-Chair: Antoine Desir, Google Inc, New York, NY, United States 1 - Using Bundle Sales to Speculate on Price Elasticities for an Online Retailer Will Ma, Massachusetts Institute of Technology, 71 School Street, Cambridge, MA, 02139, United States, David Simchi-Levi This work is motivated by the observation that bundle sales numbers generally contain richer information about demand, and specifically, contain information about price elasticities that individual sales numbers do not. We propose a basic rule for using bundle sales to speculate on price elasticities. We test the validity of this simple rule on a data from a large online retailer and show that our speculations are correct with statistical significance. 2 - Planning Online Advertising using Lorenz Curves John G. Turner, University of California - Irvine, The Paul Merage School of Business, Room SB2 338, Irvine, CA, 92697-3125, United States, Miguel Lejeune Lorenz curves are commonly-used to depict dispersion; e.g., income inequality. Motivated by online advertising campaigns that desire impressions spread over targeted audience segments and time, we formulate a problem that minimizes Gini Coefficients (area under the Lorenz curve), and develop a specialized decomposition technique to solve instances quickly. 3 - Auction Design for ROI-constrained Buyers Ilan Lobel, New York University, New York, NY, United States, Negin Golrezaei, Renato Paes Leme We combine theory and empirics to (i) show that some buyers in online advertising markets are financially constrained and (ii) demonstrate how to design auctions that take consider such financial constraints. Using a field experiment on Google’s advertising exchange, we find that a significant set of buyers lowers their bids when reserve prices go up. We show that this behavior can be explained if we assume buyers have constraints on their minimum return on investment (ROI). We proceed to design auctions for ROI-constrained buyers. We show that optimal auctions for symmetric ROI-constrained buyers are either second-price auctions with reduced reserve prices or subsidized second-price auctions. n SD21 North Bldg 129B Learning and Optimization in Revenue Management Problems Sponsored: Revenue Management & Pricing Sponsored Session Chair: Huanan Zhang, Penn State University, 310 Leonhard Building, University Park, PA, 16801, United States 1 - Nonparametric Learning with Covariates Ningyuan Chen, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, Guillermo Gallego Decision makers usually need to optimize an initially unknown objective function with observed covariates. When the form of the objective function is unknown, postulating a parametric model of the function may lead to misspecification. We consider a nonparametric policy, which achieves a regret of order $O(\log(T)^2T^{(2+d)/(4+d)})$, where $T$ is the length of the horizon and $d$ is the dimension of the covariate. The algorithm is shown to be near-optimal. The role played by $d$ highlights the complex interaction between the nonparametric formulation and the covariate dimension; It also suggests the decision maker incorporate contextual information selectively when possible. 2 - Size Matters, so does Duration: The Interplay between Offer Size and Offer Deadline Zhenyu Hu, National University of Singapore, 15 Kent Ridge Drive, Mochtar Riady Building, BIZ1 8-69, Singapore, 119245, Singapore, Wenjie Tang We study a Stackelberg game involving a proposer and responder. The proposer acts first by making an offer to the responder with a deadline, and the responder, following a continuous finite-horizon search for alternative offers, has to respond to the offer by the deadline. We find that the proposer should offer a non-

98