Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TC13

3 - A Primal-dual Approach to Analyzing Assemble-to-Order Systems Levi DeValve, Duke University, 716 Turmeric Lane, Durham, NC, 27713, United States, Sasa Pekec, Yehua Wei Assemble-to-order (ATO) problems with general structure and integrality constraints are well known to be difficult to solve. We provide new approximation guarantees by approaching the ATO problem from a primal-dual perspective. We design an LP rounding algorithm for the one-period problem that achieves both asymptotic optimality as demand grows large, and a constant factor approximation of 1.8 for any problem instance. We apply our LP rounding analysis to design an asymptotically optimal integral policy in a related dynamic model. 4 - On the Performance of Tailored Base-surge Policies: Theory and Application at Walmart.com Linwei Xin, University of Chicago, 5807 S. Woodlawn Avenue, Chicago, IL, 60637, United States, Long He, Jagtej S. Bewli, John Bowman, Huijun Feng, Zhiwei Qin We consider the following dual-sourcing inventory problem: one supplier is reliable but has a longer lead time; the other one is not always reliable but has a shorter lead time. It is motivated by a real-world problem at Walmart.com and the lead time differences of many import items could be as large as 12 weeks. We prove that a Tailored-Base Surge (TBS) policy is asymptotically optimal as the lead time difference grows. We also test the performance of TBS by using data from Walmart.com. Our result shows that Tailored-Base Surge outperforms other heuristics. n TC12 North Bldg 126A Data Driven Research on the Interface Between OM and Marketing Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Xuying Zhao, University of Notre Dame, Notre Dame, IN, 46556, United States 1 - A Reinforcement Learning Approach for Hotel Revenue Management Jun Zhang, Fudan University, School of Management, Fudan University, Shanghai, 200433, China, Ji Chen, Yifan Xu, Peiwen Yu We develop a data-driven approach for hotel revenue management. In this approach, the dynamic capacity allocation problem for multiple class customers is solved in two steps: First, a recommended average price is computed, and then the capacity allocation decision is made based on the average price. The recommended average price is computed with a reinforcement learning algorithm. The capacity allocation decision is made based on a linear programming model taking into account a hotel’s preference for different classes of customers. 2 - Can Leanness Predict Financial Distress? Feng Mai, Stevens Institute of Technology, 1 Castle Point Terrace, Hoboken, NJ, United States, Nagihan Comez-Dolgan, Xuying Zhao We investigate whether operational slack can predict financial distress. We use quarterly panel data of US manufacturing firms for the period from 1980-2016. Using Merton’s distance to default model, we identify a sample of financial distress firm. We then apply LASSO regression to assess the relative importance of operational slack variables as predictors for financial distress. Our results show that operational slack can complement financial ratios commonly used in the existing credit risk literature. 3 - Product Quality, Consumer Returns, and the Moderating Role of Salesperson Xiaojing Dong, Santa Clara University, Marketing Department, Lucas Hall, Santa Clara, CA, 95053, United States, Necati Ertekin The literature has rigorously explored the relationship between product quality and consumer returns through analytical frameworks under the assumption that improving quality reduces consumer returns. We empirically (1) investigate this relationship with respect to two product quality dimensions, namely objective quality and perceived quality, that have been the focus of operations management and marketing, respectively, and (2) examine the moderating role of salesperson.

4 - Data-Driven Pricing and Assortment Decisions for Soybean Seeds Durai Sundaramoorthi, Washington University in Saint Louis, 10352 Conway Road, Saint Louis, MO, 63131, United States, Lingxiu Dong, Iva Petrova Rashkova Offering vertical and horizontal assortments is a common and a strategic practice of firms to expand their share of the market. We introduce a hierarchical- ensemble of machine learning and optimization for product assortment in the agricultural context. We optimize the assortment of soybean varieties to grow in the mid-west of the USA. The hierarchical-ensemble framework created in this research paves way to optimize the assortment of seeds of other crops like corn, rice, and wheat. We present the data utilized, the data-driven methodology utilized, and results obtained in our presentation.

n TC13 North Bldg 126B Topics in Learning Sponsored: Revenue Management & Pricing Sponsored Session

Chair: Mohamed Mostagir, University of Michigan, Ann Arbor, MI 1 - Bayesian Social Learning with Heterogeneous Preferences: Effects of Diversity Ali Makhdoumi, MIT, 77 Massachusetts Ave. 32D-640, Cambridge, MA, 02139, United States Two opposite forces are present in social learning with heterogeneity. First, heterogeneity will make social learning more difficult as the actions of others become less informative. Second, each individual uses more of her information in making decisions and ``herds’’ become less likely. In this talk, we study the interplay between these two forces and characterize the effect of heterogeneity on social learning and its speed. 2 - Information and Learning in Contests: An Experimental Study Mohamed Mostagir, University of Michigan, 701 Tappan Ave, Ann Arbor, MI, 48109, United States, Yan Chen, Iman Yeckehzaare Contests are a common mechanism for extracting effort from participants. Their use is widespread in a variety of settings like workplace promotions or crowdsourcing. One of the pivotal aspects of contest design is the contest’s information structure: what information should the contest designer provide to participants and when should this information be revealed? The answers to these questions have important implications to how players behave and the overall outcome of the contest. We show how the behavior of players significantly deviates from theory predictions and provide recommendations for how these information mechanisms should be designed in light of these experimental findings. 3 - Impatience and Learning In Services Senthil Veeraraghavan, University of Pennsylvania, Wharton School OPIM Department, 545 3730 Walnut Street, Philadelphia, PA, 19104, United States, Hanqin Zhang, Xiao Li Customers often abandon waiting in queues when they get impatient. Prior literature on Markovian queues shows that it is not rational to quit “midway”: it is rational for customers to quit either immediately on arrival (balk) or wait till the completion of their service. We show how abandonment behavior may occur when customers wait and learn in Markovian queues. We model the learning and abandonment behavior of customer who arrives to a system without full knowledge about the service rate but learns more about service rate. Our work reveals interesting features in waiting behavior, showing that customers can be (rationally) more patient in slower shorter queues, than in faster longer queues. 4 - Technology Adoption and Information Acquisitionin a Partnership Sasa Pekec, Duke University, Durham, NC, United States We study technology adoption decisions under uncertainty in a partnership. This requires coordination among partners with non-identical objectives. We show that classical results for a single decision-maker (SDM policy) do not extend to and do not provide guidance for optimal decision-making in a partnership, We also show that it could be optimal for the partnership to prematurely adopt/reject the technology as compared to the SDM policy, and that anticipating premature decisions in a later period could trigger unraveling which leads to a series of premature decisions in earlier periods.

311