2015 Informs Annual Meeting

TC45

INFORMS Philadelphia – 2015

TC43 43-Room 103A, CC Joint Session RMP/MSOM: Choice Models: Estimation and Optimization Sponsor: Revenue Management and Pricing Sponsored Session Chair: Sumit Kunnumkal, Indian School of Business, Gachibowli, Hyderabad, India, Sumit_Kunnumkal@isb.edu 1 - Formulation, Motivation, and Estimation for the D-Level Nested Logit Model Guang Li, University of Southern California, Bridge Hall 401, None, Los Angeles, CA, 90089-0809, United States of America, guangli@usc.edu, Huseyin Topaloglu, Paat Rusmevichientong Using a tree of depth d, we provide a novel formulation for the d-level nested logit model. Our model is consistent with the random utility maximization principle and equivalent to the elimination by aspects model. Using new concavity results on the log-likelihood function, we develop an effective parameter estimation algorithm. Numerical results show that the prediction accuracy of the d-level nested logit model can be substantially improved by increasing the number of levels d in the tree. 2 - Assortment Optimization Over Time James Davis, Cornell University, 290 Rhodes Hall, Ithaca, NY, United States of America, jamesmariodavis@gmail.com, Huseyin Topaloglu, David Williamson Inspired by online retail we introduce a new type type of assortment optimization problem: assortment optimization over time. In this problem the retailer must choose which products to display but must also choose an ordering for the products. This is a relevant problem when items are displayed as a list; this is common when returning results from a search query, for example. We provide a framework to analyze this problem, provide an approximation algorithm, and some hardness results. 3 - Tractable Bounds for Assortment Planning with Product Costs Sumit Kunnumkal, Indian School of Business, Gachibowli, Hyderabad, India, Sumit_Kunnumkal@isb.edu, Victor Martínez-de-Albéniz Assortment planning under a logit demand model is a difficult problem when there are product specific costs associated with including products into the assortment. In this paper, we describe a tractable method to obtain an upper bound on the optimal expected profit. We provide performance guarantees on the upper bound obtained. We describe how the method can be extended to incorporate additional constraints on the assortment or multiple customer segments. 4 - Clustering Consumers Based on Their Preferences Ashwin Venkataraman, New York University, Preference-based clustering is an important and challenging problem. We propose a non-parametric method to cluster consumers based on their preferences for a set of items. Our method combines the versatility of model-free clustering (such as k-means) with the flexibility and rigor of model-based clustering (based on EM algorithm). Our approach is fast, can handle missing data, identify general correlation patterns in consumer preferences, and has provable guarantees under reasonable assumptions. TC44 44-Room 103B, CC Pricing in Online Markets Sponsor: Revenue Management and Pricing Sponsored Session Chair: Kostas Bimpikis, Stanford GSB, 655 Knight Way, Stanford, CA, 94305, United States of America, kostasb@stanford.edu 1 - Modeling Growth for Services: Evidence from the App Economy Ken Moon, PhD Candidate, Stanford GSB, 655 Knight Way, Stanford, CA, 94305, United States of America, kenmoon@stanford.edu, Haim Mendelson We present a model of service operations to grow and sustain customers by the operational design and performance of services, rather than marketing alone. Applying our framework to data from services in the app economy, we show (i) that customers’ engagement contributes as powerfully to growth as virality; and (ii) evidence of an experience curve (from customer interactions) for service operations. We present a model of incentive-compatible pricing for this setting. 715 Broadway, New York, NY, United States of America, ashwin.venkataraman@gmail.com, Srikanth Jagabathula, Lakshminarayana Subramanian

2 - Mobile Technology in Retail: The Value of Location-based Information Marcel Goic, Assistant Professor Or Marketing, University of Chile, Republica #701, Santiago, 8370438, Chile, mgoic@dii.uchile.cl, Jose Guajardo We analyze the value of location-based information in mobile retailing and the conditions under which incorporating geolocation information increase effectiveness metrics for retailers. 3 - Dynamic Pricing in Ride-Sharing Platforms Siddhartha Banerjee, Postdoc, Stanford University, 475 Via Ortega, Stanford, CA, 94305, United States of America, sidb@stanford.edu, Carlos Riquelme, Ramesh Johari We develop a model for ride-share platforms, which combines a queueing model for the platform dynamics with strategic models for passenger and driver behavior. Using this, we study various aspects of this system - the value of dynamic pricing versus static pricing; the robustness of these policies; the effect of heterogenous ride-request rates and traffic between different locations. Joint work with Ramesh Johari, Carlos Riquelme and the Data Science team at Lyft. 4 - Pricing with Limited Knowledge of Demand Maxime Cohen, MIT, 70 Pacific Street, Apt. 737B, Cambridge, MA, 02139, United States of America, maxcohen@mit.edu, Georgia Perakis, Robert Pindyck How should a firm price a new product with limited information on demand? We propose a simple pricing rule that can be used if the firm’s marginal cost is constant: the firm estimates the maximum price it can charge and then sets price as if demand were linear. We develop bounds that show that if the true demand is one of many commonly used demand functions, the firm will do “very well” - its profit will be close to what it would earn if it knew the true demand. TC45 45-Room 103C, CC Behavioral Issues in RM Sponsor: Revenue Management and Pricing Sponsored Session Chair: Anton Ovchinnikov, Queen’s University, 143 Union Str West, Kingston, Canada, anton.ovchinnikov@queensu.ca 1 - Should Consumers be Strategic? Arian Aflaki, Doctoral Student, Duke University, 100 Fuqua Drive, Box 90120, Durham, NC, 27708, United States of America, arian.aflaki@duke.edu, Robert Swinney, Pnina Feldman We consider whether strategic consumer behavior benefits consumers when they purchase from a rational, revenue-maximizing firm that sets prices over multiple periods. We show that strategic behavior does not benefit all consumers. Then, by studying a wide range of pricing and inventory strategies in a unified setting, we find that different strategies may induce different levels of interest in strategic behavior. 2 - Intertemporal Pricing under Minimax Regret Ying Liu, Stern School of Business, New York University, 44 West 4th Street, KMC 8-154, New York, NY, 10012, United States of America, yliu2@stern.nyu.edu, Rene Caldentey, Ilan Lobel We consider a monopolist selling a product to a population of consumers who are heterogeneous in valuations and arrival times. We study the policies that attain minimum regret when selling to either myopic or strategic customers. We characterize the set of optimal policies and demonstrate their structural properties. 3 - Behavioral Anomalies in Consumer Wait-or-Buy Decisions and the Implications for Markdown Management Nikolay Osadchiy, Emory University, 1300 Clifton Rd NE, Atlanta, GA, 30322, United States of America, nikolay.osadchiy@emory.edu, Anton Ovchinnikov, Manel Baucells A decision to buy at a tag price or wait for a possible markdown involves a trade- off between the value, delay, risk and markdown magnitude. We build an axiomatic framework that accounts for three well-known behavioral anomalies along these dimensions and produces a parsimonious generalization of discounted expected utility. We consider a pricing/purchasing game and show that accounting for the behavioral anomalies results in substantially larger markdowns and leads to noticeable revenue gains.

333