Informs Annual Meeting 2017

TA10

INFORMS Houston – 2017

TA11

2 - Route Preferences in Bike-sharing Systems Pu He, Columbia University, Uris Hall, Cub 4H, New York, NY, 10027, United States, phe19@gsb.columbia.edu, Fanyin Zheng, Elena Belavina, Karan Girotra Over 300 cities have implemented large scale bike-share systems, including London, Paris and New York city. Using transaction-level trip and bike-availability data from London’s bike sharing system, we construct a structural model of commuters’ preferences over different routes that captures the effects of two components of route-length: the biking and the walking distances, along with the service-levels experienced at both ends. We illustrate the use of our estimates for system design, e.g. to choose between adding stations to increase station density in existing coverage areas or to expand the coverage area, to compare a docking station model with a dock-less model, etc. 3 - Managing Flexibility and Relationships in an Online Marketplace: Experiment and Prescriptions for Pricing and Matching Ken Moon, The Wharton School, Philadelphia, PA, 19104, United States, kenmoon@wharton.upenn.edu, Elena Belavina, Karan Girotra Exploiting the recent introduction of relationship-specific pricing by the largest online intermediary connecting freelancers to jobs, we present empirical evidence relating the operational performance of the platform’s pricing and matching regimes to a simple market characteristic: whether clients derive greater relative value (A) by pursuing repeat relationships with trusted freelancers or (B) by exploiting the platform as an intermediary to flexibly hire different freelancers. We discuss how platforms can follow intuitive rules to tailor their pricing and matching policies based on how participants value relationships and flexibility, respectively. 4 - A Structural Estimation Approach to Study Agent Attrition Seyedmorteza Emadi, University of North Carolina at Chapel Hill, 330B Personnel Scheduling Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Rachid Hassani, Polytechnique Montreal/Gerad, rachid.hassani@gerad.ca Co-Chair: Dalia Attia, Gerad, 1, Montreal, QC, Canada, dalia.attia@gerad.ca 1 - Employee Scheduling with Inter Departments Transfers Dalia Attia, Polytechnique Montreal & GERAD, Montreal, QC, Canada, dalia.attia@gerad.ca, Guy Desaulniers, Francois Soumis Employee scheduling with inter-department transfers integer program is intractable for large instances. We propose a three-phase heuristic, each phase solves a small integer program. The first phase is a pre-processing phase which identifies probable inter-departments transfers needs. The second phase creates for each department apart, employee schedules using previously gathered information. The third phase, globally fulfills any remaining demand. Several variations for each phase are implemented and compared showing promissing results along with quick execution time. 2 - Real-time Personnel Re-scheduling After a Minor Disruption Rachid Hassani, Polytechnique Montreal & GERAD, Montreal, QC, Canada, rachid.hassani@gerad.ca, Issmail El Hallaoui, Guy Desaulniers This talk is about a real-time optimization method to adapt a pre-set schedule after a small disruption that can result from delay or absence of employees. The method provides schedule’s rectification choices to the decision-maker taking into account the immediate and a deterministic future costs. This method considers five types of decision and is based on the dual values of the linear relaxation of the personnel scheduling problem . We also propose a procedure exploiting a multivariate adaptive regression splines method for updating the dual values after each disruption when several ones occur in the same week. 221 Tremont Cir, Chapel Hill, NC, 27516, United States, seyed_emadi@kenan-flagler.unc.edu, Bradley R.Staats We use a structural estimation approach to investigate involuntary and voluntary attrition decisions of agents in a call center. We estimate different parameters of agents including their sensitivity to salary and their degree of forward looking behavior. TA10

332A Empirical Studies in Service Operations Sponsored: Manufacturing & Service Oper Mgmt, Service Operations Sponsored Session Chair: Carri Chan, Columbia Business School, New York, NY, 10027, United States, cwchan@columbia.edu Co-Chair: Fanyin Zheng, Columbia University, New York, NY, 10027, United States, fz2225@gsb.columbia.edu 1 - Why do Automakers Initiate Recalls? Robert Louis Bray, 830 Hinman Ave., 2s, Evanston, IL, 60202, United States, robertlbray@gmail.com, Ahmet Colak Are automakers averse to consumer complaints or to the government recalls they attract? We study this question with 8,439 manufacturer recalls, 5,685 government recalls, and 976,062 defect reports submitted to the government. We model the agents’ joint recall decisions as an asymmetric dynamic discrete choice game. Our estimates suggest (i) there is a little overlap in the sets of products the agents recall and in the types of complaints they respond to, (ii) the cost of a recall does not depend on who initiates it, and (iii) auto manufacturers recall faulty products to avoid receiving defect reports but not to preempt anticipated government recalls. 2 - Estimating Dynamic ICU Admission Decision Fanyin Zheng, Columbia University, Columbia Business School, 412 Uris Hall, New York, NY, 10027, United States, fz2225@gsb.columbia.edu, Carri Chan We study physician’s ICU admission decisions using structural estimation methods. 3 - Understanding the Impact of Remote Patient Monitoring in the Hospital ICU: An Empirical Study Lesley Meng, The Wharton School, University of Pennsylvania, 3730 Walnut Street, Suite 500 JMHH, Philadelphia, PA, 19104, United States, lmeng@wharton.upenn.edu, Christian Terwiesch Innovation in healthcare technology has led to new healthcare delivery methods such as the addition of remote patient monitoring in the hospital intensive care unit. We utilize data on patient vitals to study the effect of remote monitoring on provider responsiveness to adverse patient health states in the ICU. 4 - Forecasting Product Life Cycle Curves: Practical Approach and Empirical Analysis Jan A. Van Mieghem, Northwestern University, Kellogg School of Mgmt (MEDS.Dept), Leverone Hall 2001 Sheridan Rd, Evanston, IL, 60208-2009, United States, vanmieghem@northwestern.edu, Kejia Hu, Jason Acimovic, Francisco Erize, Douglas Thomas We present an approach to forecast customer orders of ready-to-launch new products that are similar to past products. The approach fits product life cycle (PLC) curves to historical customer order data, clusters the curves of similar products, and uses the representative curve of the new product’s cluster to generate its forecast. We propose three families of curves to fit and forecast the PLC: Bass diffusion curves, polynomial curves and simple piecewise-linear curves (triangles and trapezoids). We evaluate our approach using a large data set of customer orders for 4,037,826 units of 170 Dell computer, representing several billion dollars in revenue.

272