Informs Annual Meeting 2017

WC47

INFORMS Houston – 2017

3 - Busting the Bracket Toward an in Depth Predictive Analysis of College Mens Basketball Tournaments Jie Tao, Assistant Professor in Information Systems, Fairfield University, 1073 N.Benson Rd, 100 Silliman St APT1A, Fairfield, CT, 06824, United States, jtao@fairfield.edu, Michael Mitchell, Kanlun Wang The NCAA Division I basketball tournament attracts significant attention each year. In this study, we aim at developing models to predict outcomes of tournament games. Compared to previous studies, our results rely solely on the matchups rather than any pre-established rating systems. The contribution of this study is tri-fold: we develop predictive models for the tournament - yielding satisfactory results; we also select several predictors for developing a new rating system; lastly, we use unsupervised learning to investigate how erroneous predictions would reveal stochastic factors in predictive results. Our findings can be applied to other related scenarios such as elections and polls. 4 - A Data Driven Analysis on Automotive Dealership Management Haidar Almohri, PhD Candidate, Wayne State University, 2833 Catalpa Circle, Ann Arbor, MI, 48108, United States, almohrih@yahoo.com, Ratna Babu Chinnam, Mark Colosimo Automotive dealership management is challenging in that they have to run their business for producing profit, while keeping the OEM happy with selling new vehicles. In this work, a deep data-driven study and analysis for dealership management is performed to find the key driving factors for different departments in a dealership. The model-based clustering and recommendation methodology developed seeks to maximize the dealer profit while satisfying sales targets from the OEM. 361B Behavioral Operations Contributed Session Chair: Julie Niederhoff, Syracuse University, Syracuse, NY, United States, jniederh@syr.edu 1 - Screening Out Biased Managerial Decisions to Interfere with Automatic Store Replenishment Antti Tenhiala, IE Business School, Calle Maria de Molina 12, piso 5, Madrid, 28006, Spain, antti.tenhiala@ie.edu, Shivom Aggarwal We study managerial decisions to deviate from automatic store replenishment (ASR) proposals using multi-store data from a European supermarket chain. After controlling for the drivers of this behavior already identified in the existing research (including package sizing, shelf space considerations, product margins, forecast errors, demand seasonality, transit lags, item durability, and others), we find support for the prevalence of the basic decision-making biases of anchoring and demand chasing. We further find that these biases strongly predict unsuccessful replenishment decisions and suggest a way to incorporate a warning system in ASR software to discourage biased decision making. 2 - Experimental Analysis of the Newsvendor Problem with Minimum Order Quantity Contracts Özge Tüncel, Student, Singapore University of Technology and Design, 8 Somapah Rd, Singapore, 487372, Singapore, tuncel_ozge@mymail.sutd.edu.sg, Niyazi Taneri, Sameer Hasija Experiments between human suppliers and automated retailers revealed that the cognitive burden associated with MOQ contract is lower and the profits are higher than with either buyback or revenue-sharing contracts. Moreover, when given a choice, subjects tend to pick the MOQ contract more often, and make quicker and better decisions. 3 - An Experimental Study of Fair Transitivity in a Three-tier Supply Chain Junlin Chen, Associate Professor, Central University of Finance and Economics, 39 South College Road, Haidian District, Beijing, 100081, China, chenjunlin@cufe.edu.cn We study a three-tier supply chain consisting of a supplier, a manufacturer, and a retailer. We conducted experiments to investigate the sequential bargaining process between tiers, in high and low demand markets. We measure how the underlying three-tier structure influences the profit distribution across the supply chain. We find that the structural issue blocks access to people’s social preferences. Most suppliers behave few fairness concerns, and always set a high wholesale price close to the rationally theoretical value. Most manufactures follow a high wholesale price when the received offer is not kind. Our data also show that a high demand market will not promote people to be kind. WC49

4 - Addressing Project Complexity: the Role of Contractual Functions Nan Gao, Tianjin University, TIanjin, China, gona1219@126.com, Yongqiang Chen Project complexity leads to transaction risks in interorganizational exchanges and can negatively affect performance if appropriate governance mechanisms are absent. Though the contract prevails in construction projects to address transaction risks, its role in coping with complexity-related risk has not yet been fully investigated. Empirical results in this paper show that contractual coordination can deal with risks induced by technical, organizational and environmental complexity, while the adaptation function can address environmental complexity-related risk. However, contractual control is ineffective when either technical or environmental complexity is high. 5 - Collective Choice Impact on Organizational Performance: Why Do Megaprojects (Seem to) Fail? Yongcheng Fu, The University of Manchester, Manchester, United Kingdom, yongcheng.fu@postgrad.mbs.ac.uk, Nuno Gil This study investigates the impact of collective choice on organizational performance. Our empirical motivation is a long-standing puzzle about the root causes of the performance slippages which are endemic to megaprojects. By piecing together organizational structure and performance data for the life-cycle of three megaprojects, we show that: i) cost escalation is positively correlated with the disputes endemic to the growth of a pluralistic actor-network throughout planning; and ii) cost hikes are negatively correlated with the growth of a buyer- supplier network throughout implementation. A recurrent neural network was further developed to predict cost. 6 - Designing under Pluralism: the Value of Slack to Reconcile Private Interests in Collective Choices Yongcheng Fu, The University of Manchester, Manchester, United Kingdom, yongcheng.fu@postgrad.mbs.ac.uk, Nuno Gil This study develops a theoretical perspective on how slack resources affect value creation in ‘pluralistic’ product developments—settings where the project promoter shares decision-making authority over design choice with multiple autonomous actors with differing subgoals. We argue that, under some circumstances, the slack carried by the promoter enables to reconcile differing private interests, but the actual use of slack rarely benefits all participants alike. We illustrate this tension by drawing on data for a new railway network, High- Speed 2 in the UK. We further derive formal boundary conditions for creating and distributing collective value using a game theoretic representation. 360F Business Applications Contributed Session Chair: Haidar Almohri, Wayne State University, Ann Arbor, MI, United States, almohrih@yahoo.com 1 - Practical Applications of Predictive Models Suchitra Veera, Strategy and Analytics Consultant, Snayu Inc, 4512 Dartmoore Lane, Colleyville, TX, 76034, United States, suchitra.veera@gmail.com The objective of this session is to present the topic of the practical applications of predictive model and to help the audience understand the use of predictive models in business and management. The presentation will discuss the use of predictive models in Finance, Human Resources, Marketing and Finance. It will allow attendees to take away ideas about the use of predictive models in business and how this is facilitated through the increasing availability of data through many different sources. 2 - Identifying Sources of Assignable Error via Process Pattern Mining Bhupesh Shetty, PhD Candidate, University of Iowa, 707 Oakcrest Street, Apt E, Iowa City, IA, 52246, United States, bhupesh-shetty@uiowa.edu, Nick Street, Jeffrey W. Ohlmann We apply pattern mining techniques to event logs generated by a manufacturing process in order to identify the root causes of product defects. We use an association-based method algorithm to generate the frequent patterns and then use correlation measures to identify the patterns of interest. For a varying levels of target pattern complexity, we compare the effectiveness of correlation measures for varying sample sizes with and without knowledge of the underlying process map. WC47

520