INFORMS 2021 Program Book

INFORMS Anaheim 2021

TE23

2 - Managing Projects in Virtual Settings: Information Exchange Networks And Project Performance Sukrit Pal, Doctoral Candidate, Michigan State University, East Lansing, MI, United States, Anand Nair Projects are increasingly being managed in virtual settings where project team members collaborate online. Collaborative information exchanges among team members result in the creation of complex communication network that can have non-trivial influence on the outcome due to the asynchronous nature of these exchanges and the lack of visual and non-verbal cues. We examine the impact of communication network characteristics on the number of issues closed within an open source software (OSS) development project. Additionally, the study examines the role of project managers’ active participation in these communication networks on project outcome. We analyzed a panel dataset comprising of 1842 OSS development projects spanning 104 weeks from the time of project initiation was carefully compiled for this research. 3 - Double Utilization of Consumer Appreciation: Is the Ecosystem Linkage the New Leverage? Esma Koca, Imperial College London, London, United Kingdom, Robert Peach, Hang Ren In this paper, we model the release of a new product category when two asymmetric firms engage in price and quality competition. The asymmetry is because one of the firm can strategically link a new good with a previous product category (ecosystem leverage). We examine whether the ecosystem leverage enables the ecosystem firm to keep its rival at bay or discourage firms to innovate in the new product category, as some industry experts argue. We conclude that the effect of the ecosystem leverage on its rival or the innovation is not necessarily adverse, but under some conditions, even facilitates the welfare of the rival. TE25 CC Room 205B In Person: Practical Optimization: What it Takes to Make Optimization Succeed in Real Life General Session Chair: Richard Oberdieck, Gurobi Optimization, Hvidovre, 2650, Denmark 1 - Deploying a Hybrid MILP Solution for Highly Complex Semiconductor Scheduling Problems Semya Elaoud, Flexciton Limited, London, United Kingdom, Dionysios Xenos, Ioannis Konstantelos Job scheduling in semiconductor factories is an NP-hard problem. It is a non- identical parallel machines job shop problem with secondary resources. Practical applications typically resort to the use of approximate techniques. Many factors render the use of MILP optimisation challenging: problem complexity, high uncertainty, multiple objectives. We present a novel solution strategy that combines MILP optimisation with heuristic techniques to schedule thousands of wafers. Flexciton has been deployed in semiconductor fabs and shown to outperform existing approaches. Using case studies we showcase how we can accommodate complex problem features and provide high quality schedules 2 - Tools and Processes for Rapid Prototyping of Optimization Applications Richard Oberdieck, Gurobi Optimization, Nordlundsvej 27, Hvidovre, 2650, Denmark Ideally, projects involving optimization start out with the business problem for which a decision strategy is needed. Based on this, a mathematical model is designed, implemented and validated against a set of test data, before it is encoded in an optimization application which solves the business problem.Unfortunately, most projects do not follow this path, for example due to complicated business logic, stakeholder management and data quality issues. Therefore, it is crucial to be able to iterate through each step quickly in order to quickly identify and resolve any blocking issues.In this talk, we will share some of the tools and processes used inside Gurobi that we have found to work well in these situations. In addition, we will provide recommendations to OR experts, developers and project managers on common pitfalls we see and several strategies on how to mitigate them.

TE23 CC Room 204C In Person: Managing Behaviors in Healthcare/Behavior in Waiting Lines General Session Chair: Brett Hathaway, Johns Hopkins University, MD, 21784, United States 1 - Intra-day Dynamic Rescheduling under Patient No-shows Aditya Shetty, University of Rochester, Rochester, NY, 14620-4436, United States, Henri Groenevelt, Vera Tilson Existing work on appointment scheduling assumes that appointment times cannot be updated once they have been assigned. In this paper, we describe an intra-day dynamic rescheduling model that takes into account the observed no- shows and service times as the day goes on, to makes adjustments to subsequent appointment times. We find that, unlike optimal static schedules, optimal dynamic schedules do not follow some seemingly intuitive characteristics making it computationally intensive to solve the problem using brute force. We propose a more efficient approach to compute the optimal scheduling policy and find the conditions under which switching from static to dynamic scheduling is most beneficial. 2 - The Impact of Procedural Justice on Patient Flow in Hospitals Galit Bracha Yom-Tov, Technion Israel Institute of Technology, Technion City, Haifa, 32000, Israel, Anat Rafaeli, Matias Kohn, Michal Medan We investigate the impact of procedural justice in routing patients between ED and inpatient wards on patient LOS. Using diff-in-diff analysis we show a huge reduction (of more than 20%) in patient’s LOS after implementing equalized routing. We investigate the mechanisms that drive this reduction. 3 - Personalized Priority Policies in Call Centers Using Past Customer Interaction Information Brett Hathaway, Johns Hopkins University, Baltimore, MD, 21784, United States, Seyed Emadi, Vinayak V. Deshpande We show how call centers can improve customer service using personalized priority policies, where managers use customer contact history to predict individual-level caller abandonment and redialing behavior and prioritize them based on these predictions to improve operational performance. We provide a framework for how companies can use individual-level customer history data to capture the idiosyncratic preferences and beliefs that impact caller abandonment and redialing behavior, and quantify the improvements to operational performance of these policies by applying our framework using caller history data from a real-world call center. We achieve this by formulating a structural model that uses a Bayesian learning framework to capture how callers’ past waiting times and abandonment/redialing decisions affect their current abandonment and redialing behavior. TE24 CC Room 205A In Person: Emerging Topics in Technology and Innovation Management General Session Chair: Esma Koca, Imperial College London, London, SW10 9JH, United Kingdom Co-Chair: Esma Koca, Imperial College London, London, SW10 9JH, United Kingdom 1 - Search in the Dark: The Normal Case Manel Baucells, University of Virginia, Darden School Of Bus. Charlottesville, VA, 22903-1760, United States, Sasa Zorc Search problems where the reward is equal to the highest sampled value are ubiquitous in real life. We tackle the important unresolved case of sampling from a normal distribution with unknown mean and unknown variance. We find that single threshold stopping rules—-prevalent to search theory—-are no longer optimal. Instead, the optimal stopping region consists of up to two bounded intervals. We may stop if the current value is near the previous sampling mean (a signal of low variance and reduced search profitability), or if it improves on the previous best (future samples must now exceed a higher hurdle to be of value), but not by much (very high or very low values signal high variance, and recommend more sampling). After two draws, it is optimal to discontinue search if the two draws are sufficiently similar.

144