Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WD15

2 - Mental Accounting in Consumer Inventory Problem with Uncertain Benefit Haonan He, University of Science and Technology of China, Hefei, 61820, China Newsvendor problems describe a situation in which the vendor needs to predict the demand by a buyer when a fixed unit profit is predetermined. But sometimes, the vendor can effectively affect the demand, that is, when he is also the buyer. This paper first considers how a consumer’s mental accounting behavior affects whether consumer reaches a theoretically prespecified optimal inventory level when the unit profit is not fixed and how to exacerbate and mitigate potential judgment inaccuracies. The result shows that changes to the benefits affect inventory more strongly and lead to more deviation than to the cost. Besides, inventory could increase even when the volatility of demand increases. 3 - Double Auctions for Truthful Information Sharing in Sales and Operations Planning Frank Hage, Technical University of Munich, Production and Supply Chain Management, Munich, 80333, Germany, Martin Grunow We study a sales and operations planning problem. Operations persons are responsible for production sites and have private information on production costs. Sales persons are responsible for customer markets and have private information on customer demand. Due to function-oriented bonus schemes, both parties have conflicting interests: While operations persons aim to minimize production and inventory costs, sales persons aim to maximize turnover. Both parties have an incentive to misrepresent their private information, which leads to inefficient capacity allocation and lower profitability. To address this problem, we develop an iterative double auction for capacity allocation. n WD14 North Bldg 126C Joint Session MSOM/Practice Curated: Healthcare Operations Sponsored: Manufacturing & Service Oper Mgmt/Service Operations Sponsored Session Chair: Sukriye Nilay Argon, University of North Carolina, Chapel Hill, NC, 27599, United States Co-Chair: Serhan Ziya, University of North Carolina, Chapel Hill, NC, 27599, United States 1 - Who is Next: An Empirical Study of Patient Prioritization in an Emergency Department Zhankun Sun, City University of Hong Kong, Kowloon, Hong Kong, Jeff Hong, Li Wenhao In the emergency department (ED), priority scores are assigned to patients at triage based on their acuity levels. However, using operational data from more than 150,000 patient visits, we find that doctors may deviate from this priority sequence, and within each priority class, patients may not be served in a first- come-first-serve manner. Our analysis shows that when selecting the next patient to treat, doctors prioritize patients who are more likely to be discharged after treatment at ED when many ED beds are occupied by boarding patients, in an effort to avoid further access block to the ED. 2 - How to Manage Doctor Appointments with a Shared Medical Appointment Option Nazli Sonmez, London Business School, Regent’s Park, London, NW1SN, United Kingdom, Kamalini Ramdas, Sarang Deo Shared medical appointments are an alternative to traditional one-on-one appointments for routine care of chronic diseases that offer an innovative, interactive approach to healthcare delivery. They are not widely used even though there is a high potential. To adopt this new healthcare delivery method, the service providers must make an upfront decision on how to allocate service capacity. We develop a model that will incorporate what we learn about how patients make trade-offs while choosing an appointment, when offered two different appointment types using the data from a healthcare provider. This model will provide insight into how many appointments from each type need to be scheduled. 3 - Equilibrium Behavior of Patients Served by a Dual Practice Physician Lerzan E. Ormeci, Koc University, Dept of Industrial Engineering, Rumeli Feneri Yolu, Istanbul, 34450, Turkey, Dimitrios Andritsos, Yiannis Dimitrakopoulos We consider a physician who has a dual practice in public and private sectors. We model this system as two M/M/1 queues with a coupled server. The server alternates between the queues with exponential rates. The system has three levels of decisions: First, government provides the main rules for having a dual practice, which may limit the time she spends in her public and private offices and regulate the private service fee. Second, the physician decides on how much time she will

spend in each practice and the private fee to maximize her revenue. Finally, patients decide on the office to visit to maximize their utility which is a function of their waiting time, value of service and private fee. 4 - Modelling the Use of Patient Activation Measure (PAM) in Chronic Care Management Evrim D. Gunes, Koc University, Rumeli Feneri Yolu, Sariyer, Istanbul, 34450, Turkey, Lerzan E. Ormeci, Odysseas Kanavetas, Christos Vasilakis We develop a Markov Decision Process framework to manage care for individual patients with multiple chronic conditions through a complex care hub. Complex care provision influences the evolution of Patient Activation Measure (PAM), an indicator for healthy behavior, which affects the evolution of health state of patients. We define a general model where the transition probabilities and the rewards are time dependent parameters. Then, we explore optimal and heuristic policies which maximise the welfare for static parameters. Through numerical experiments we explore the performance of alternative policies that focus on managing more complex patients or improving activation of all patients. n WD15 North Bldg 127A New Topics in Demand Learning and Assortment Planning Sponsored: Manufacturing & Service Oper Mgmt/Service Operations Sponsored Session Chair: Victor Araman, American University of Beirut, Beirut, Lebanon Co-Chair: Rene A. Caldentey, The University of Chicago, Chicago, IL, 60637, United States 1 - Crowdvoting New Product Introduction Victor Araman, American University of Beirut, Lebanon, Rene A. Caldentey Launching new products into the marketplace is a complex and risky endeavor that companies must continuously undertake. In this paper, we consider a seller who has the ability to first test the market and gather demand information before deciding whether or not to launch a new product. In particular, we consider the case in which the seller sets up an online voting system and offers multiple versions of the product - differentiated through their quality levels and prices - for potential customers to vote on. We investigate the optimal design of such a crowdvoting system in order for it to provide an effective demand forecast and allow the seller to identify which versions if any to commercialize. 2 - Reputation Formation in Social Networks Mohamed Mostagir, University of Michigan, Ann Arbor, MI, 48109, United States, Asuman Ozdaglar, James Siderius Reputation is one of the main drivers that sustains relationships between consumers and firms. This work generalizes the reputation literature —which typically assumes sequential arrivals of agents— to a network setting. We show how a firm can manipulate the learning process of agents in a network so that they never learn the true state of the world (for example, that the firm intentionally provides low quality service or that a news outlet provides false news). We then characterize those societies that are immune to manipulation and those that are not, and suggest potential remedies for the latter. 3 - Dynamic Assortment Planning under Nested Logit Models Xi Chen, New York University, 44 W. 4th St, NYU KMC Room 8- 50, New York, NY, 10012, United States, Yining Wang, Yuan Zhou We study a stylized dynamic assortment planning problem, where for each arriving customer, the seller offers an assortment of substitutable products and customer makes the purchase among offered products according to a nested logit model. The goal of the seller is to maximize the expected revenue, or equivalently, to minimize the worst-case expected regret. We propose a lower and upper confidence bound algorithm with an aggregated estimation and establish the corresponding regret bound. One advantage of our policy is that our regret does not depend on the number of products. 4 - Learning Customer Preferences from Personalized Assortments Yifan Feng, University of Chicago, 5807 South Woodlawn Avenue, Chicago, IL, 60637, United States, Rene A. Caldentey, Christopher Ryan A company wishes to commercialize the best version of a product out of a menu of available alternatives. The company does not know customers’ preferences over the set of alternatives and relies on a voting system that allows potential buyers to vote for their preferred version. Under a general ranking-based choice model framework, we study how to dynamically customize each individual voter’s choice set, so as to identify the top-ranked alternative with a fixed probabilistic confidence level, while using a minimal number of votes.

499