Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WA21

allocation rules and guidelines for designing sparse network structures such as generalized long chains. A real case study is conducted to confirm our findings as well as some of the flexibility principals conjectured in the literature. 2 - Management and Effects of In-store Promotional Displays Oguz Cetin, University of North Carolina, Chapel Hill, NC, United States, Adam J. Mersereau, Ali Kemal Parlakturk We examine a brick-and-mortar retailer’s choice of which product to include in a promotional display. The display provides a visibility advantage to both the featured product and its category, but it also has consequences for customer traffic and substitution. While there has been considerable academic interest in the assortment planning problem and in the shelf-space allocation problem, little attention has been paid to the problem of where to place products in the store. We develop analytical insights using a problem formulation based on a nested multinomial logit model. Our work provides guidance for how retailers can use and value promotional displays effectively. 3 - Dynamic Pricing under a Static Calendar Jinglong Zhao, Massachusetts Institute of Technology, 77 Massachusetts Avenue, E17 IDSS 495G, Cambridge, MA, 02139, United States, Will Ma, David Simchi-Levi Our work is based on the classical dynamic pricing problem. From our collaborations with a large Consumer Packaged Goods company, we have found that while they appreciate the advantages of dynamic pricing, it is operationally beneficial for them to plan out a deterministic price calendar in advance. It is possible to deviate from this calendar as demand is observed, but there is a significant overhead in doing so, and thus deviation should be reserved for situations where the realized demand was drastically higher or lower than expected. Motivated by this, we formulate the dynamic pricing problem under static calendar constraints, and study how classical dynamic pricing intuition may break down. 4 - Contingent Stimulus in Crowdfunding Longyuan Du, Rotman School of Management, University of Toronto, 105 St George St, Toronto, ON, M5S 3E6, Canada, Ming Hu, Jiahua Wu We study a model where backers arrive sequentially at a crowdfunding project. We characterize the dynamics of the pledging process. To boost success, we focus on a commonly practiced stimulus policy of running promotions, e.g., on social media, to attract backers. We show that the optimal promotion intensity depends on the pledge progress and is not monotone. We analyze the deterministic heuristics and show that their performances are compromised because of the all or nothing nature of crowdfunding. We propose a modified resolving heuristic that greatly improves the performance and has provable optimality gaps. Testing with the Kickstarter data, we demonstrate the benefit of the stimulus policies. n WA21 North Bldg 129B Practice- Optimization I Contributed Session Chair: Maziar Sanjabi, University of Southern California, Los Angeles, CA, 90007, United States 1 - Job Scheduling with Simultaneous Assignment of Machines and Multi-skilled Workers: A Mathematical Model Cinna Seifi, Research Assistant, Clausthal University of Technology, Julius-Albert-STrasse, Clausthal-Zellerfeld, 38678, Germany, Jürgen Zimmermann The primary task of mining companies is the extraction of raw materials. Based on a planning program, a certain quantity of materials is expected to be extracted within a given time horizon. The applied mining method is characterized by nine process steps each of which has to be executed by an appropriate machine and worker. The processing time of a process step depends on the assigned devices and workforces. In this paper, we formulate a mathematical program for a simultaneous assignment of devices and personnel to a selection of jobs. The aim is to minimize the difference between the predetermined quantity and the amount of extracted material, cumulatively over all the process steps, for a work shift. 2 - Effective Rules of Thumb in Analytics Modeling Yanqi Xu, Director of Applied Technology, Princess Cruises, Valencia, CA, 91355, United States Various industries build optimization models to provide meaningful solutions to practical problems so as to increase revenue, reduce costs, manage risks, streamline operations, etc. In this talk, we will share our model building experience and lessons learned across different industries. Topics (with real world examples) include: number one concern of model building, how to solve the problem at the right level of details, what to include/exclude, signs that the issues of a model cannot fixed by incremental improvement, performance tradeoffs, and modeling with sparse data.

n WA19 North Bldg 128B Innovative Pricing Strategies Sponsored: Revenue Management & Pricing Sponsored Session Chair: Pnina Feldman, Univeristy of Pennsylvania, Philadelphia, PA, United States 1 - Pricing Capacity over Time and Recourse Strategies: Facilitate Reselling, Offer Refunds, or Overbook? Pnina Feldman, Boston University, 595 Commonwealth Ave, Boston, MA, 02215, United States, Gerard P. Cachon Perishable capacity is often sold before it is used which creates the opportunity to include in the pricing mechanism a recourse strategy, such as resale, refunds and overbooking. We investigate alternative recourse strategies and the implication for firm’s profit and social welfare. 2 - On the Power of Bounded Memory Peak End Demand Models Tamar Cohen, MIT, Somerville, MA, 02144, United States, Kiran Panchamgam, Georgia Perakis We consider the problem of promotion planning for a single product, with bounded memory peak end demand. We show that when using this demand model, the optimal promotion policy will require at most two price levels. We then show that with a finite planning time horizon and a limit on the number of deviations from full price, we can find the structure of the optimal pricing policy. We characterize when the optimal policy is cyclic, as well as when a cyclic policy is not optimal. Finally, we show that even in the cases that we misclassify the demand model, and the true demand model does not fit the bounded memory peak end model, the estimation error is small. We provide an analytical bound on this estimation error 3 - Intertemporal Price Discrimination via Randomized Pricing Jiahua Wu, Imperial College Business School, Office 382, Tanaka Building, Imperial College London, London, SW7 2AZ, United Kingdom, Hongqiao Chen, Ming Hu The undesirable, but inevitable, consequence of dynamic pricing is that consumers are trained to time their purchases strategically. We consider a novel randomized pricing approach, where the firm varies prices in a random fashion, and study its effectiveness in addressing consumers’ strategic behavior. We show that the optimal price distribution follows a simple structure. We also benchmark the optimal randomized pricing policy against optimal static pricing and cyclic pricing policies, and characterize conditions when the former dominates. Lastly, we extend randomized pricing in two different directions, namely, cyclic randomized pricing and Markovian pricing. 4 - Revenue Optimization under a Preannounced Dynamic Price Function Dana Pizarro, University of Chile, Beauchef 851, Santiago, 8320000, Chile, Jose Correa, Gustavo J. Vulcano We consider a dynamic game between a seller and a single buyer. The seller operates over an infinite time horizon. At the beginning of the horizon the seller commits to a price function p(t) . The buyer arrives according to a general random process, has a private valuation for the item, and decides when to buy in order to maximize his expected utility. The seller’s problem is to determine the price function p(t) to maximize her expected revenue. We analyze this dynamic game and provide numerics to show the achievable revenue performance. n WA20 North Bldg 129A Revenue Management in Retail and Online Platforms Sponsored: Revenue Management & Pricing Sponsored Session Chair: Yanzhe Lei, University of Michigan, Ann Arbor, MI, 48104, United States Co-Chair: Stefanus Jasin, University of Michigan, Ann Arbor, MI, 48105, United States 1 - Online Demand Fulfillment under Limited Flexibility Zhen Xu, Columbia University, New York City, NY, United States, Rachel Q. Zhang, Jiheng Zhang We extend the work by Asadpour et al. (2016) on online demand fulfillment for systems with the same number of resources and request types to general systems and show that a positive Generalized Capacity Gap (GCG) introduced by Shi et al. (2016) is both necessary and sufficient for a system to achieve bounded performance. Both theoretical and numerical evidence point to the GCG as the most important indicator of system performance, which leads to simple inventory

417