Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TC19

4 - Dynamic Salesforce Compensation Long Gao, University of California-Riverside, 900 University Avenue, 221 Anderson Hall, AGSM, Riverside, CA, 92521, United States We study a salesforce compensation problem in a dynamic setting, where the salespeople have evolving skills and can learn over time. We characterize the optimal contract and derive the managerial implications. 5 - Should Network Television Enable Consumer Binge Watching? Franco Berbeglia, PhD Candidate, Carnegie Mellon University, Tepper School of Business, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, Timothy Derdenger, Kannan Srinivasan, Joseph Xu With the broadcast television industry facing new forms of competition, networks have searched for new methods to distribute shows. Through the networks’ online interfaces shows are able to match the new competitors and offer all-at- once releases rather than follow the traditional linear strategy. This lowers consumers’ viewing costs, but also diminishes the viewer’s responsiveness to advertising. This paper studies the impact of this new channel on network shows with a signaling model. We find the adoption of this channel is highly profitable for low quality shows, reducing the necessity of costly signaling through advertising for the high quality shows, which results in a more efficient market. n TC19 North Bldg 128B Topics in Dynamic Pricing and Revenue Management Sponsored: Revenue Management & Pricing Sponsored Session Chair: Vineet Goyal, Columbia University, New York, NY, 10027, United States 1 - Aggregation Bias in Demand Estimation Zizhuo Wang, University of Minnesota, 1009 5th Street SE, Minneapolis, MN, 55414, United States, Chaolin Yang, Hongsong Yuan, Yaowu Zhang, Ruoqing Jiang Demand estimation is a core challenge in many industrial problems. In practice, companies often only have aggregate demand data in daily granularity, even though they may have changed prices within each day. A common approach in that case is to take a simple/weighted average of prices and use it as the price for each day (and the total demand as the demand for that day). However, such an approach may result in bias in demand estimation. In this work, we investigate under what conditions using such aggregate approach will result in an underestimation or an overestimation of price sensitivities. We also examine practical data from Alibaba to see the practical effects of demand estimation using aggregate data. 2 - Customized Individual Promotions: Model, Optimization, and Prediction Srikanth Jagabathula, NYU Stern School of Business, 44 W. 4th St, Kmc Rm 8-74, New York, NY, 10012, United States, Dmitry Mitrofanov, Gustavo J. Vulcano We propose a back-to-back procedure for running personalized promotions in retail operations contexts, from the construction of a nonparametric choice model where customer preferences are represented by directed acyclic graphs (DAGs) to the design of such promotions. We fit the models to historical purchases tagged by customer id, product availability and promotion data for a category of products. We provide new bounds for the likelihood of a DAG and show how to conduct the MNL estimation. We test our model to predict purchases at the customer individual level on real retail data. Finally, we illustrate how to use it to run personalized promotions. 3 - Impact of Network Structure on New Service Pricing Vahideh Manshadi, Yale University, 165 Whitney Ave, Rm 3473, New Haven, CT, 06511, United States, Saed Alizamir, Ningyuan Chen, Sang-Hyun Kim We analyze a firm’s optimal pricing of a new service offering when the firm faces consumers who interact among themselves in a network and exert positive externality on each other. In order to maximize the long-run revenue generated from network externality, the firm provides its service for free initially to promote usage growth, raising the price when a sufficient level of consumer interactions is established. We study the impact of network structure on the firm’s optimal pricing policy and revenue.

4 - Dynamic Influence Maximization in Social Networks Shatian Wang, Columbia University, New York, NY, United States, Van-Anh Truong, Zhen Xu We study algorithms for dynamically learning and spreading influence over social networks, by selecting the influencers in the network to incentivize. n TC20 North Bldg 129A Learning in Online Markets Sponsored: Revenue Management & Pricing Sponsored Session Chair: Negin Golrezaei, Massachusetts Institute of Technology, Cambridge, MA, 02142, United States Chair: Santiago Balseiro, Columbia University, Columbia University, New York, NY, 10027, United States 1 - Discontinuous Demand Functions: Estimation and Pricing N. Bora Keskin, Duke University, Fuqua School of Business, 100 Fuqua Drive, Durham, NC, 27708-0120, United States, Arnoud Victor den Boer Price-rankings in online marketplaces and price comparison websites create discontinuities in demand functions. Motivated by this, we consider a dynamic pricing-and-estimation problem with an unknown and discontinuous demand function. We show that ignoring such discontinuities results in substantial loss of revenues, and construct near-optimal policies that can handle demand discontinuities. 2 - Robust Repeated Auctions under Heterogeneous Buyer Behavior Shipra Agrawal, Columbia University, 500 West 120th Street, Mudd 423, New York, NY, 10027, United States, Konstantinos Daskalakis, Vahab Mirrokni, Balasubramanian Sivan We study revenue optimization in a repeated auction between a single seller and a single buyer. Traditionally, the design of repeated auctions requires strong modeling assumptions about the bidder behavior, such as it being myopic, infinite lookahead, or some specific form of learning behavior. Is it possible to design mechanisms which are simultaneously optimal against a multitude of possible buyer behaviors? We answer this question by designing a simple state-based mechanism that is simultaneously approximately optimal against a k-lookahead buyer for all k, a buyer who is a no-regret learner, and a buyer who is a policy- regret learner. 3 - Contextual Bandits with Cross-learning Jon Schneider, Google Research, New York, NY, United States, Santiago Balseiro, Negin Golrezaie, Mohammad Mahdian, Vahab Mirrokni In the contextual bandits problem, in each round t, a learner observes some context c, chooses some action a to perform, and receives some reward ra,t(c). We study the variant of this problem where the learner also learns the values of ra,t(c’) for all other contexts c’. This variant arises in several strategic settings, such as learning how to bid in non-truthful repeated auctions. The best algorithms for the classical contextual bandits problem achieve O(sqrt(CKT)) regret against all stationary policies, where C is the number of contexts, K the number of actions, and T the number of rounds. We demonstrate algorithms for our variant that remove the dependence on C and achieve improved regret bounds. 4 - Dynamic Pricing with Demand Learning and Reference Effects Arnoud den Boer, University of Amsterdam, Science Park 107, room F3.33, Amsterdam, Netherlands, N. Bora Keskin We consider a dynamic pricing problem with demand learning and reference effects. Customers are loss averse: they have a reference price that can vary over time, and the demand reduction when the selling price exceeds the reference price dominates the demand increase when the selling price falls behind the reference price by the same amount. Consequently, the demand function has a time-varying “kink” and is not differentiable everywhere. The seller neither knows the underlying demand function nor observes the reference prices. We consider several variants of this problem and design asymptotically optimal pricing policies.

314