Informs Annual Meeting 2017

TA13

INFORMS Houston – 2017

TA12

TA13

332B Field Experiments in Service and Retail Operations Sponsored: Manufacturing & Service Oper Mgmt, Service Operations Sponsored Session Chair: Qiuping Yu, Kelley School of Business, Indiana University, Bloomington, IN, 47405-1701, United States, qiupyu@indiana.edu 1 - The Reference Effect of Delay Announcements: A Field Experiment Qiuping Yu, Kelley School of Business, Indiana University, 1309 E. Tenth Street, Bloomington, IN, 47405-1701, United States, qiupyu@indiana.edu, Gad Allon, Achal Bassamboo We explore whether customers are loss averse in time and how the amount of delay information available may impact such reference-dependent behavior by conducting a field experiment at a call center. Our results show that customers exhibit loss averse regardless of the availability or accuracy of the delay information. While delay announcements may not alter the fact that customers are loss averse, they do seem to impact the reference points customers use when the announcements are accurate. However, when those announcements are not accurate, customers may completely disregard them. 2 - Low-acuity Patients Delay High-acuity Patients in the ED Danqi Luo, Stanford University, Stanford, CA, United States, dluo@stanford.edu, Sara Kwasnick, Mohsen Bayati, Erica Plambeck Causal analysis of data from 4 different hospitals shows that an additional low- acuity patient in the ED significantly increases the wait time to start of treatment for a high-acuity patient. Through a pseudo-randomized experiment at 1 hospital, we estimate that arrival of a low-acuity patient increases the aggregated wait time for high-acuity patients by 55 minutes on average. We identify 2 mechanisms by which low-acuity patients delay high-acuity patients, analyze a novel model of a priority queue with transition delays, and recommend ways to mitigate delays for high-acuity patients. 3 - Setting Retail Staffing Levels: Methodology and Implementation Serguei Netessine, University of Pennsylvania, Philadelphia, PA, United States, netessin@wharton.upenn.edu, Marshall L.Fisher, Santiago Gallino We describe a three-step process that a retailer can use in setting retail store sales staff level. First, use historical data on revenue and planned and actual staffing levels by store to estimate how revenue varies with staffing level at each store. We disentangle the endogeneity between revenue and staffing levels by focusing on randomly occurring deviations between planned and actual labor. Second, using historical analysis as a guide, validate these results by changing the staffing levels in a few test stores. Finally, implement the results chain-wide and measure the impact. We describe the successful deployment of this process with a large specialty retailer. 4 - Stable Schedules for Retail Associates: Field Experiment at the Gap Saravanan Kesavan, University of North Carolina-Chapel Hill, Kenan-Flagler Business School, Cb 3490 Mccoll Building, Chapel Hill, NC, 27599-3490, United States, skesavan@unc.edu, Susan Lambert, Joan Williams Employer scheduling practices are an important source of employment instability that can limit earnings, impede worker performance, and create stress and work- life interferences that undermine worker health and well-being. Widespread acknowledgement of the impact of these challenges on thousands of working class households has triggered several policy debates regarding the need for more stable schedules across a variety of industries. In this study, we ran a field experiment in 30 stores of the Gap to improve stability in workers’ schedules without affecting the store performance.

332C Game Theory Contributed Session Chair: Xiaojun (Gene) Shan, UHCL, Houston, TX, United States, shan@uhcl.edu 1 - Cost Allocation Mechanisms in a Peer-to-peer Network We consider a P2P network where the service provider broadcasts the content across the network and the users collaborate to seed the content to a subset of users in the network. The objective of the service provider is to determine the minimum cost network solution and to allocate this joint-cost fairly among the users. The minimum cost network solution can be determined by solving a minimum cost Steiner Tree problem. We propose four cost allocation mechanisms: a dual linear programming based mechanism, an approximation mechanism to the Shapley Value, a partition-based mechanism, and an approximation mechanism to the nucleolus. 2 - Safe and Nested Subgame Solving for Imperfect-Information Games Tuomas W. Sandholm, Professor, Carnegie Mellon University, Gates Center for Computer Science, Carnegie Mellon University, Pittsburgh, PA, 15213, United States, sandholm@cs.cmu.edu, Noam Brown Unlike perfect-information games, imperfect-information games cannot be solved by solving subgames independently: decisions must consider the strategy for the whole game. Yet one can improve strategies by solving subgames. We introduce subgame-solving techniques that outperform prior methods in theory and practice. We also adapt them to respond to opponent actions that are outside the action abstraction; this significantly outperforms the prior state of the art, action translation. Finally, we show that subgame solving can be repeated as the game progresses down the tree, leading to lower exploitability. These techniques are key components of Libratus, the first AI to beat top humans in no-limit poker. 3 - Tiered Cloud Storage Pricing via Two-stage, Latency-aware Bidding In this paper, we wish to maximize the overall profit of the Cloud Service Providers (CSPs) that utilize a tiered storage architecture (with hot and cold storage tiers) with file placement and access request scheduling flexibilities. To this end, we propose a scheme where the CSP offers a two-stage auction process for (a) requesting storage capacity, and (b) requesting accesses with latency requirements. Our two-stage bidding scheme provides a hybrid storage and access optimization framework with the objective of maximizing the CSP’s total net profit over four dimensions: file acceptance decision, placement of accepted files, file access decision and access request scheduling policy. 4 - Empirical Study on Effects of Defense Resource Allocations Xiaojun (Gene) Shan, UHCL, Houston, TX, United States, shan@uhcl.edu, Jun Zhuang In this paper, we use the method of statistical analysis based on data from Global Terrorism Database, which is available at National Consortium for the Study of Terrorism and Responses to Terrorism (START) and the Urban Area Security Initiative (UASI) grant allocations from FY 2004 to FY 2012 to study the effects of defense resource allocations. We conclude that ingeneral, effects of defense resources onreducing success probability of attack are mixed due to scarcity of terrorism data. Basak Altan, Ozyegin University, Istanbul, Turkey, basak.altan@ozyegin.edu.tr, Okan Orsan Ozener Vaneet Aggarwal, Purdue University, 315 N Grant St., West Lafayette, IN, 47906, United States, vaneet@purdue.edu, Yang Zhang, Arnob Ghosh, Lan Tian

273