Informs Annual Meeting 2017

MA45

INFORMS Houston – 2017

2 - Can Good Jobs be Profitable in Low Cost Services? A Systemic Model and Estimation Hazhir Rahmandad, Associate Professor, MIT, 100 Main Steet, E62-442, Cambridge, MA, 02142, United States, hazhir@mit.edu Can profit maximizing firms offer good jobs in low cost services? In a dynamic model of service operations we capture feedbacks, managerial decisions, costs, and benefits related to viability of good jobs. Using data from Borders bookstores and utilizing extended Kalman filtering with latent variables we estimate this model. We find evidence for significant benefits of employee quality and building capabilities as well as notable managerial biases in allocating resources between customer service and capability building. The results point to potential viability of good jobs in low cost services and offers a method for assessing these costs and benefits. 3 - Estimating Customer Spillover Learning of Service Quality ”Spillover” learning is defined as customers’ learning about the quality of a service (or product) from their previous experiences with similar yet not identical services. In this paper, we propose a novel, parsimonious and general Bayesian hierarchical learning framework for estimating customers’ spillover learning. We apply our model to a one-year shipping/sales historical data provided by a world- leading third party logistics company. Our empirical results are consistent with information spillovers driving customer choices. We also develop policy simulation studies to show the importance of accounting for customer learning when a firm considers service quality improvement decisions. 4 - Investigating the Influencing Factors of Continuous Usage Behavior in Internet Broadcast Xiaoxi Luan, Tongji University, Shanghai, 200092, China, luanxiaoxi@163.com, Haifeng Zhao This study aims to investigate what, why and how a cohesive set of factors influence the continuous usage behaviors of Live show users.The model of ECM- IT is applied to the study to help build the Theoretical Model of Continuous Usage Behavior of Internet Broadcast Users. In the model, we set Perceived Usefulness, Perceived Ease of Use, Expectation Confirmation, Perceived Pleasure, and Perceived Trust as independent variables, Satisfaction and Continuous Use Intention as intermediate variables, Continuous Usage Behavior as the dependent variable, and Habit as the regulated variable. We introduce the method of SEM Andres I.Musalem, U. de Chile, Beauchef 851, Santiago, 8370456, Chile, amusalem@dii.uchile.cl, Yan Shang, Jing-Sheng, Jeannette Song Matthew Walsman, Assistant Professor, Rutgers Business School, 28 Cambridge Dr, Berkeley Heights, NJ, 07922, United States, mwalsman@business.rutgers.edu In this study, we use a behavioral experiment in a consulting context to investigate the impact of decisions that advisers and clients make and their influence on each other. Specifically, we test the impact of client initiated estimates on their consultant’s subsequent recommendations and the resulting client decisions and satisfaction. We test for this effect with both trained and novice clients. We find that when clients interfere with the process by providing their consultants with an initial estimate, expert consultants don’t ‘take the bait’, but instead are more likely to give a differentiated opinion, even when the client’s initial input was right. 360D JFIG Paper Competition II Sponsored: Junior Faculty JFIG Sponsored Session Chair: Jose Luis Walteros, University at Buffalo, SUNY, 413 Bell Hall, Bu, NY, 14260, United States, josewalt@buffalo.edu Co-Chair: Canan Gunes Corlu, Boston University, 808 Commonwealth Avenue, Boston, MA, 02215, United States, canan@bu.edu 1 - Joint Pricing and Inventory Management with Strategic Customers Yiwei Chen, Singapore University of Technology and Design, 8 Somapah Road, Singapore, 487372, Singapore, ywchen@mit.edu, Cong Shi We consider a model wherein the seller sells a single indivisible product to customers over an infinite horizon. At each time, the seller decides a set of purchase options offered to customers and the inventory replenishment quantity. Each purchase option specifies the price and the product delivery time. Customers stochastically arrive to the system according to a Poisson process. Customer product valuations are randomly drawn from a distribution and are heterogeneous. A customer’s arrival time and product valuation are his private information. Customers are forward-looking, i.e., they strategize their purchasing MA45 and the technology of AMOS21.0 to analyze the collected data. 5 - Do Advisers Give Bad Advise Experimental Evidence from Consulting

times. A customer incurs delay disutility from postponing to make the purchasing decision and waiting for the product delivery. A customer’s delay disutility rate is perfectly and positively correlated with his valuation. The seller has zero replenishment lead time. The seller incurs fixed ordering cost and inventory holding cost. The seller seeks a joint pricing, delivery, and inventory policy that maximizes her long-run average expected profit. We propose a simple policy under which the dominant equilibrium is that customers behave myopically. We show that this policy is asymptotically optimal in the regime wherein customer arrival rate and the seller’s fixed ordering cost proportionally grow large. We adopt a mechanism design approach to prove these results. 2 - Posterior Sampling for Markov Decision Processes: Worst-case Regret Bounds Shipra Agrawal, Columbia University, Industrial Engineering and OR, 423 S. W. Mudd Building, New York, NY, 1002 7, United States, ashipra@gmail.com, Randy Jia We consider discrete-time communicating Markov Decision Processes (MDP) with finite state space and action space. We present a posterior sampling (aka Thompson sampling) based algorithm for adaptive learning and decision making in such MDPs when the underlying transition matrix is unknown. Our main result is a high probability worst-case regret upper bound of Ö(D √ SAT) for this algorithm, for any communicating MDP with finite diameter D, S states, A actions, and T ≥ S^5A. Here, regret compares the total reward achieved by the algorithm to the total expected reward of an optimal infinite-horizon undiscounted average reward policy, over a time horizon T. This result improves over the best previously known upper bound of Ö(DS √ AT ) achieved by any learning algorithm in this setting, and closes the crucial gap of S from the known lower bound of Ω ( DSAT ) for this problem. Our analysis involves deriving some novel results about the anti-concentration of Dirichlet posteriors, which may be of independent interest. 3 - Ambiguous Risk Constraints with Moment and Unimodality Information Bowen Li, University of Michigan, Ann Arbor, MI, United States, libowen@umich.edu, Ruiwei Jiang, Johanna Mathieu Optimization problems face random constraint violations when uncertainty arises in constraint parameters. Effective ways of controlling such violations include risk con- straints, e.g., chance constraints and conditional Value-at-Risk (CVaR) constraints. This paper studies these two types of risk constraints when the probability distribu- tion of the uncertain parameters is ambiguous. In particular, we assume that the distributional information consists of the first two moments of the uncertainty and a generalized notion of unimodality. We find that the ambiguous risk constraints in this setting can be recast as a set of second-order cone (SOC) constraints. In order to facil- itate the algorithmic implementation, we also derive efficient ways of finding violated SOC constraints. Finally, we demonstrate the theoretical results via a computational case study on power system operations. 360F eBusiness Sponsored: EBusiness Sponsored Session Chair: Jong Seok Lee, PhD, University of Memphis, Memphis, TN, 38152, United States, jslee4@memphis.edu 1 - Modeling Retail and Online Channel Allocation Decisions Roshanak Mohammadivojdan, University of Florida, 303 Weil Hall, P.O. Box 116595, Gainesville, FL, 32611, United States, rmohammadivojdan@ufl.edu, Joseph Geunes Consider a retailer who sells items via both a retail store and an online channel. Space limitations in each channel require judicious selection of items and the allocation of these items to the two channels. We model these assortment and channel selection decisions via a stylized mathematical model that accounts for economic inventory replenishment decisions in each channel. The resulting model corresponds to a specially structured two-knapsack problem. Analysis of this problem permits gaining insights on key decision drivers and characteristics of optimal decisions. 2 - The Disruptive Effects of Digital Platform’s Entry on Non-high-tech Industry: A Text Mining and Econometric Approach Sungjin Yoo, University of Memphis, Memphis, TN, United States, syoo3@memphis.edu In this study, we examined the effect of digital platform entry on the landscape of competition in non-high-tech industry. Using the accommodation industry as our empirical setting, we identify how the incumbents’ competitive action portfolio change in response to the entry and how such action differs between the high- end companies and the low-end companies. By analyzing a longitudinal data set of competitive actions of incumbents in the industry, our finding shows that incumbents not only implement more competitive actions but also adapt diverse strategies which are deviant from the industry norm. Furthermore, high-end companies attempt to respond more actively than low-end companies. MA47

148