2016 INFORMS Annual Meeting Program

WB48

INFORMS Nashville – 2016

2 - Coordination And Social Value Of Information In Networks Gowtham Tangirala, Columbia Business School, 3022 Broadway,

4 - Overcommitment In Cloud Services – Bin-packing With Chance Constraints Maxime Cohen, NYU Stern, New York, NY, 10012, United States, maxime.cohen@stern.nyu.edu, Phil Keller, Vahab Mirrokni, Morteza Zadimoghaddam A cloud provider needs to decide how many physical machines to purchase in order to accommodate the incoming virtual jobs efficiently. This is typically modeled as a bin-packing optimization problem. Overcommitting servers clearly improves the bin-packing objective, but induces a risk for the provider. In this work, we show that the bin-packing with chance constraints can be solved using a class of simple online algorithms that guarantee a constant factor from optimal. We explicitly model job size uncertainty to motivate new algorithms and evaluate them on realistic workloads. 210-MCC Business Applications in Social Media Analytics Invited: Social Media Analytics Invited Session Chair: Michel Ballings, University of Tennessee, 255 Stokely Management Center, Knoxville, TN, 37996, United States, michel.ballings@utk.edu 1 - Identifying New Product Ideas: Waiting For The Wisdom Of The Crowd Or Screening Them In Real-time Steven Hoornaert, Ghent University, Ghent, Belgium, steven.hoornaert@ugent.be, Michel Ballings, Edward C Malthouse, Dirk Van den Poel This article studies idea ranking in innovation communities using the contributor’s history of submitting ideas and comments, the Content of the idea suggestion, and the Crowd’s feedback on the idea. Results show that contributor and content variables improve ranking between 22.6% and 25.8% over exhaustive idea selection across classifiers. 2 - Evaluating The Importance Of Different Communication Types In Tie Strength Prediction On Social Media Matthias Bogaert, Ghent University, matthias.bogaert@ugent.be, Michel Ballings, Dirk Van den Poel The purpose of this paper is to evaluate which communication types on social media are most indicative of tie strength. To ensure that we have the best possible model we benchmark several classifiers. The results indicate that we can predict tie strength with very high accuracy. The top performing classification algorithm is adaboost with an AUC of 0.976. The top five communication predictors are the recency of commenting on links, posts, videos, the frequency of liking post comments and the recency of commenting on albums. To the best of our knowledge, this study is the first to provide such an extensive analysis of tie strength in social media. 3 - Evaluating Prediction Models For Targeting Product Reviewers Michel Ballings, Assistant Professor, University of Tennessee, 249 Stokely Management Center, Knoxville, TN, 37996, United States, michel.ballings@utk.edu, Rachel Van Deventer, Ryan Erwin, Miller Moore, Dirk Van den Poel As customers increasingly rely on product reviews while making their purchases, businesses must take action and make obtaining high review volume a priority. The purpose of this study is to develop a predictive model that identifies if an online reviewer is likely to write a review for a selected product. To develop our model, we extracted product reviewer data from Amazon.com. We find that the model accurately predicts if an individual will review the focal product. Businesses can target that population and obtain high review volume for their investment. While a large body of research has been published on product reviews we have focused on the individuals behind the reviews. 4 - Behavioral Engagement In Social Media: Measurement, Drivers And Impact On Purchase Behavior Welf H. Weiger, University of Goettingen, Platz der Goettinger Sieben 3, Goettingen, D-37073, Germany, welf.weiger@wiwi.uni- goettingen.de, Wendy W Moe, Hauke A Wetzel, Maik Hammerschmidt In this study, we focus on understanding and measuring behavioral consumer engagement in social media. Our research combines three sources of individual- level user data (i.e., matched survey, social media behavior and purchase behavior data) collected in the context of an online fashion retailer’s social media site. We develop a composite engagement measure and we identify its motivational drivers and consequences for purchase behavior. Our results reveal different drivers for the incidence (i.e., the “whether”) and the nature (i.e., the “how”) of engagement. As a counterintuitive finding, our results further show that complaining users buy more than complimenting users. WB48

4th Floor West, New York, NY, 10027, United States, gtangirala18@gsb.columbia.edu, Alireza Tahbaz-Salehi

This paper studies the social and equilibrium value of information in network games. We provide a complete characterization of conditions under which equilibrium is efficient under incomplete information and study the impact of varying the commonality of information across agents and the network structure, on equilibrium welfare. In particular, we find that when social conformity is desirable (undesirable), the more interconnected the network is, the lesser (greater) its equilibrium welfare. 3 - The Effect Of Information On Traffic Congestion Ali Makhdoumi, Massachusetts Institute of Technology, makhdoum@mit.edu We study the implications of additional information about routes provided to certain users (e.g., via GPS-based route guidance systems) in a traffic network. We formulate the question in the form of Informational Braess’ Paradox (IBP), which extends the classic Braess’ Paradox in traffic equilibria, and asks whether users receiving additional information can become worse off. We provide a necessary and sufficient condition for the occurrence of this paradox in terms of network characteristics. 4 - Optimal Promotion Period Of Products With Network Externality Ningyuan Chen, HKUST, Hong Kong, Hong Kong, nc2462@columbia.edu, Saed Alizamir, Vahideh Manshadi Many products exhibit network externality: a customer who has purchased the product makes his/her neighbors or friends more likely to buy the same product. This includes eco-friendly products such as electronic cars and solar panels. The government subsidizes customers to promote such products. We find that it is optimal for the government to stop the subsidy when the total externality of the owners reaches a threshold, which depends on the spectrum of the externality matrix. The optimal stopping time is not monotone in the strength of the externality between customers. We investigate how the structure of the network affects the stopping time and the optimal reward of the government. Chair: Maxime Cohen, Google NYC, 110 Bleecker Street Apt 6F, New York, NY, 10012, United States, maxccohen@gmail.com 1 - Simple Pricing Schemes For Consumers With Evolving Values Balasubramanian Sivan, Research Scientist, Google Research, New York, NY, United States, balusivan@google.com, Shuchi Chawla, Nikhil R. Devanur, Anna Karlin We consider a pricing problem where a buyer is interested in purchasing/using a good, such as an app or music or software, repeatedly over time. The consumer discovers his value for the good only as he uses it, and the value evolves with each use as a martingale. We provide a simple pricing scheme and show that its revenue is a constant fraction of the buyer’s expected cumulative value. 2 - Strategic And Proactive Pricing Optimization In The Airline Industry Michael Benborhoum, British Airways, New York, NY, United States, michael.benborhoum@ba.com, Maxime Cohen Pricing in the airline industry has become increasingly competitive, with a strong emphasis on reactive fare matching, arguably to the detriment of more strategic and proactive decision frameworks. Setting the right price in a strategic and proactive fashion raises at least three questions: (i) when is the right time and what is the right level for a proactive fare change; (ii) what is the right fare ladder structure leading to optimal sell-ups and fare rule segmentation; and (iii) how non-pricing factors should affect pricing decisions. In this talk, we propose an original approach to the strategic and proactive pricing problem in the airline industry. 3 - Dynamic Pricing With Heterogeneous Patience Levels Ilan Lobel, NYU Stern, ilobel@stern.nyu.edu We consider the problem of dynamic pricing in the presence of patient consumers. We call a consumer patient if he is willing to wait a certain number of periods for a lower price, but will purchase as soon as the price is equal to or below her valuation. We allow for arbitrary joint distributions of patience levels and valuations. We propose a dynamic-programming-based polynomial-time algorithm for finding optimal pricing policies. Our findings suggest that pricing for patient consumers is a more challenging problem than pricing for strategic consumers, in the sense that the dynamic program requires a larger state-space. WB47 209C-MCC Applications of RM and Pricing Sponsored: Revenue Management & Pricing Sponsored Session

411