2016 INFORMS Annual Meeting Program

TD42

INFORMS Nashville – 2016

TD42 207D-MCC RM in Online Markets Sponsored: Revenue Management & Pricing Sponsored Session Chair: Dragos Florin Ciocan, INSEAD, Fontainebleau, France, florin.ciocan@insead.edu 1 - Learning Demand Curves In B2b Pricing: A Case Application Of Optimal Learning Ilya Ryzhov, Robert H. Smith School of Business, University of Maryland, iryzhov@rhsmith.umd.edu, Huashuai Qu, Michael Fu We consider a sequence of B2B transactions involving a wide variety of buyers, products, and other characteristics, where the seller only observes whether buyers accept or reject the offered prices. The seller must adapt to this uncertain environment and learn quickly from new deals as they take place. We propose a new framework for statistical and optimal learning in this problem, based on approximate Bayesian inference, that has the ability to measure and update the seller’s uncertainty about the demand curve based on new deals. A case study demonstrates the practical potential of this approach. 2 - Pricing Of Conditional Upgrades In The Presence Of Strategic Consumers Yao Cui, Cornell University, Ithaca, NY, United States, yao.cui@cornell.edu, Izak Duenyas, Ozge Sahin We study a conditional upgrade strategy that is recently used by the travel industry. A consumer can accept an upgrade offer in advance and pay the upgrade fee at check-in if higher-quality products are still available. Consumer make forward-looking decisions regarding which product type to book. We characterize the firm’s optimal upgrade pricing strategy and identify multiple benefits of conditional upgrades. We also evaluate the revenue performance of conditional upgrades by comparing to other policies. 3 - Revenue Management With Repeated Interactions Dragos Florin Ciocan, INSEAD, florin.ciocan@insead.edu, Andre Du Pin Calmon, Gonzalo Romero We consider an RM problem where a seller of heterogeneous goods interacts with a dynamically evolving population of buyers over multiple periods. The basic trade-off we explore is between myopically optimizing the seller’s revenues over one period versus optimizing buyer surplus in the hope of increasing buyer participation in future periods. We exhibit a simple policy that is asymptotically optimal. 4 - Joint Pricing And Inventory Management With Strategic Customers Yiwei Chen, Singapore University of Technology and Design, stevenyiweichen@gmail.com, Cong Shi We consider a joint pricing and inventory management problem wherein a seller sells a single product over an infinite horizon via dynamically determining anonymous posted prices and inventory replenishment quantities. Customers have a deterministic arrival rate but heterogeneous product valuations. Customers are forward-looking, who can strategize their times of purchases. A customer incurs waiting and monitoring cost if he delays his time of purchase. The seller seeks a joint pricing and inventory policy that maximizes her long-run average profit. We show that the optimal policy is cyclic. Under the optimal policy, strategic customer equilibrium behaviors are proven to be myopic. TD43 208A-MCC Portfolio Decision Analysis Sponsored: Decision Analysis Sponsored Session Chair: Janne Kettunen, The George Washington University, The George Washington University, Washington, DC, 00000, United States, jkettune@gwu.edu 1 - Systematic Bias, Selection Bias, And Post-decision Disappointment Eeva Vilkkumaa, Aalto University, Helsinki, Finland, eeva.vilkkumaa@aalto.fi, Juuso Liesiö Based on empirical studies, the realized values of highest-ranked decision alternatives tend to be lower than estimated. This phenomenon has been explained by a systematic bias in the alternatives’ value estimates, or selection bias. We develop models for estimating the relative magnitudes of these biases using data on the estimated values of all alternatives but the realized values of selected alternatives only. Results obtained from real data on 5,610 transportation infrastructure projects suggest that out of the total cost overrun of $2.77 billion, 25% is due to systematic bias and 75% due to selection bias.

2 - Scheduling Public Procurement Contracting Janne Kettunen, Assistant Professor, The George Washington University, Washington, DC, United States, jkettune@gwu.edu, Young Hoon Kwak We show that the schedule according to which procurements are contracted can impact the number of proposals and thereby the cost of procured services or products. To help the owner of the procurements to schedule the contracts, we develop an optimization framework and apply it to the Florida Department of Transportation’s procurement contracts scheduling problem. Our results show that the optimal schedule yields about 2% ($15 million) cost savings annually. 3 - Multi-period, Multi-objective Portfolio Optimization Ernest H Forman, George Washington University, forman@gwu.edu Multi-Period, Multi-Objective and Multi-Perspective portfolio optimization requires synthesis of hard data as well as judgment. The Analytic Hierarch Process and extensions are increasingly being used to facilitate the portfolio optimization process in a variety of applications, ranging from project portfolio optimization to capital budgeting.

TD44 208B-MCC Applications of Multiattribute Preferences Sponsored: Decision Analysis Sponsored Session

Chair: Jay Simon, American University, 4400 Massachusetts Ave NW, Washington, DC, 20016, United States, jaysimon@american.edu 1 - Preference Programming For Spatial Multiattribute Decision Analysis Mikko Harju, Aalto University, Espoo, Finland, mikko.harju@aalto.fi, Kai Virtanen, Juuso Ilari Liesio The additive spatial value function models preferences among decision alternatives with spatially varying multiattribute consequences. Use of this value function can be challenging as it requires assessing a weighting function across the (uncountably infinite) set of spatial locations. To overcome this challenge, we develop (i) methods for capturing incomplete preference information about relative importance of locations, and (ii) models for identifying the resulting dominance relations among the alternatives. The use of these models is demonstrated with a military planning application. Finally, we provide some new insights about the axiomatic basis of spatial value functions. 2 - Multiattribute Preference Models For Computational Creativity Debarun Bhattacharjya, IBM T. J. Watson Research Center, Yorktown Heights, NY, United States, debarunb@us.ibm.com, Lav Varshney There is vigorous debate around definitions of creativity, yet there is general consensus that creativity inherently involves a subjective value judgment by an evaluator. In this talk, I will present evaluation of creative artifacts and computational creativity systems through a multiattribute preference modeling lens. Various implications are illustrated with the help of examples from and inspired by the creativity literature. 3 - Multilinear Utility Functions For Multiattribute Portfolio Decision Analysis Juuso Ilari Liesio, Aalto University, Helsinki, Finland, juuso.liesio@aalto.fi, Eeva Vilkkumaa In project portfolio selection applications, portfolio utility is often modeled as the sum of the projects’ multi-attribute utilities. We establish the preference assumption underlying this linear portfolio utility function. Furthermore, we show how relaxing some of these assumption leads to a more general class of multilinear portfolio utility functions, which can capture risk preferences on the portfolio level. We also develop techniques to elicit these utility functions, and optimization models to identify the project portfolio which maximizes the expected utility subject to resource and other portfolio feasibility constraints. 4 - Multiattribute Procurement Auctions With Unknown Buyer Preferences Jay Simon, American University, jaysimon@american.edu In procurement auctions for new large-scale products or services, the length of time between the request for bids and the selection of the winning bid can be extremely long. During this time, the specific needs of the buyer may change. Additionally, the new product or service being procured may involve technology that is not fully understood. Thus, the buyer may not know what her eventual preferences will be when the bid selection decision is made. However, the buyer must tell the bidders how their bids will be scored at the start of the process. This work explores optimal strategies for the buyer in the case where preferences at the time of bid selection are uncertain when the scoring rule is chosen.

348