Informs Annual Meeting 2017

TE12

INFORMS Houston – 2017

TE10

1 - Zoning for Last Mile Delivery: A Data-driven Approach Sheng Liu, University of California, Berkeley, 2249 Bonar Street, Unit C, Berkeley, CA, 94702, United States, lius10@berkeley.edu, Long He, Zuo-Jun Max Shen In this study, we propose a data-driven approach for strategic zoning planning in last mile delivery services. Using historical data to calibrate driver behaviors, we construct a practical routing distance prediction model. We then develop a mixed integer linear program as a sample average approximation model and show significant delay reductions from the benchmark model in numerical studies. 2 - Efficiency-fairness Trade-offs: the Case of the Federated Locker System Guodong Lyu, National University of Singapore, 1 Business Link, BIZ2 Building, Singapore, 117592, Singapore, guodong.lyu@u.nus.edu, Chung-Piaw Teo We study the efficiency-fairness trade-offs in a class of last mile delivery problems, using automated parcel lockers. We show that the concerns on reduction in efficiency can be mitigated with the proper choice of the scale of the system. To see this, we develop a locker choice model and calibrate the parameters to predict the volume of parcels that the system can attract based on location decisions. Contrary to conventional wisdom, our model does not place lockers near areas with peak usage because of endogeneity issues in existing demand data. 3 - Managing a Portfolio of Self-scheduling Workers Kaitlin Daniels, Olin Business School, Washington University in St. Louis, Saint Louis, MO, United States, k.daniels@wustl.edu Self-scheduling workers value their ability to decide for themselves how much they work. However, granting this flexibility costs the coordinating platform its ability to directly control its capacity. We study a system of heterogeneous workers who decide how much to work in response to incentives offered by a platform. In this setting, the platform may elect to place restrictions on a worker’s decision. We study how the resulting portfolio of working-behaviors changes as workers make costly demands of the firm (e.g. expense reimbursement, overtime pay, minimum wages), like those made by Uber drivers in recent lawsuits. 4 - Shared Mobility for Last-mile Delivery: Design, Operational Prescriptions and Environmental Impact Wei Qi, McGill University, Montreal, QC, Canada, qiwei.0216@gmail.com, Lefei Li, Sheng Liu, Zuo-Jun Max Shen We evaluate the prospect where shared mobility of passenger cars prevails throughout urban areas for home delivery services. We develop logistics planning models that characterize drivers’ responses to wages, optimal open-loop routes, interactions with the ride-sharing market and service zone design. Then we prescribe several scenarios where this business model is economically and environmentally favorable. 332B Models for Evolving Public Transportation Systems Sponsored: Transportation Science & Logistics Sponsored Session Chair: Nicholas E. Lownes, University of Connecticut, 261 Glenbrook Road, Unit 3037, Storrs, CT, 06269-3037, United States, nlownes@engr.uconn.edu 1 - Research on Integrated Development Plan for the Urban Rail Transit Station with the Concept of Linkage Development of Station and City Xinmei Chen, Southwest Jiaotong University, Chengdu, China, siyu.tao@okstate.edu Xinmei Chen, National United Engineering Laboratory of Integrated and Intelligent Transportation, Chengdu, China, siyu.tao@okstate.edu, Qiyuan Peng, Siyu Tao, Yida Li Expanding the city function blindly around the transit stations led to the contradiction between the traffic jams and land use. For linkage development of the urban rail station and city, multi-goal programming model was used to determine the final integrated development plan of the station and land, considering the investment risk, economic benefit, traffic impact of real estate for the layout of the connection channel between the station and city. TE12

330B New Trends in Pricing and Revenue Optimization Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Gustavo Jose Vulcano, Universidad Torucato di Tella, New York University, New York, NY, 10012, United States, gvulcano@stern.nyu.edu 1 - Large-scale Bundle Size Pricing: A Theoretical Analysis Tarek Abdallah, New York University, Stern School of Business, New York, NY, United States, tabdalla@stern.nyu.edu, Arash Asadpour, Joshua Reed Bundle size pricing (BSP) is a multi-dimensional selling mechanism where the firm prices the size of the bundle rather than the different possible combinations of bundles. In BSP, the firm offers the customer a menu of different sizes and prices. The customer then chooses the size that maximizes his surplus and customizes his bundle given his chosen size. In this talk, we provide a simple and tractable theoretical framework to analyze the large-scale BSP problem. Solving the BSP problem is in general hard, however we show that for large numbers of products, the BSP problem transforms from a hard multi-dimensional problem to a simple multi-unit pricing problem. 2 - Randomized Pricing, Fulfillment, and Product Ranking for Ecommerce Retailers We consider an etailer selling multiple products to customers from multiple regions. Products are stored at multiple warehouses and the decisions are price, fulfillment (i.e., from which warehouse to ship the request), and products display ranking (i.e., how we should order the products—ordering affects product visibility, which in turn affects purchase probability). The deterministic version of the problem can be solved as an MILP, however it may have a large number of binary variables. We propose a different approach: We first relax the problem (turning the MILP into an LP) and then we construct a randomization strategy that guarantees the optimal performance under the relaxed LP in expectation. 3 - Demand Learning and Earning with Online Bundle Recommendations Anna M. Papush, Massachusetts Institute of Technology-ORC, 411 Norfolk Street, Apt 2C, Somerville, MA, 02143, United States, apapush@mit.edu, Pavithra Harsha, Georgia Perakis, Divya Singhvi Product recommendation systems have grown to become a necessity in the vast majority of online retail businesses. However, most existing literature focuses on efficient design under the assumption that consumer demand is known. In this work we propose and discuss an approach for online pricing and recommendation of product bundles that makes strategic offers in order to learn consumer demand efficiently over time while maximizing profits and balancing inventory. We test our algorithm’s performance and provide theoretical guarantees against optimal offline approaches with known demand. 4 - Customized Individual Promotions: Model, Optimization, and Prediction Dmitry Mitrofanov, NYU. Stern, 44 West 4th Street, 8th Floor, New York, NY, 10012, United States, dm3537@stern.nyu.edu, Srikanth Jagabathula, Gustavo Jose Vulcano We consider the problem of predicting individual responses to a product promotion using historical transaction data, tagged by customer ID. In our proposal, we model each individual through a partial-order that embeds promotion information. Then, we calibrate a multinomial logit model over the partial orders and quantify their prediction power on out-of-sample transactions. Then, we use this information to optimize personalized promotions, obtaining a remarkable performance. Stefanus Jasin, University of Michigan, 2836 Barclay Way, Ann Arbor, MI, 48105, United States, sjasin@umich.edu, Yanzhe Lei, Joline Uichanco, Andrew Vakhutinsky

TE11

332A Operations in Delivery Services Sponsored: Manufacturing & Service Oper Mgmt,

Service Operations Sponsored Session Chair: Long He, National University of Singapore, Singapore, 119245, Singapore, longhe@nus.edu.sg Co-Chair: Wei Qi, McGill University, Montreal, QC, 1, Canada, qiwei@berkeley.edu

413