Informs Annual Meeting 2017

SC51

INFORMS Houston – 2017

SC51

a current dedicated logistics scheme in an urban environment. We use an agent- based simulation to compare distinct scenarios distinguished by network structure, connectivity, inventory positioning strategy and delivery routing. The potential economic, environmental, and social impact of such hyperconnected logistics on last-mile operations is described. 2 - Synthetic Demand Data to Drive Warehouse Design Leon McGinnis, Georgia Institute of Technology, Isye Dept, 755 Ferst Drive, Atlanta, GA, 30332-0205, United States, leon.mcginnis@isye.gatech.edu, Di Liu Suppose you are designing a new automated order fulfillment center to serve many local stores with daily deliveries. How do you model the “demand” from these stores? This talk describes a bootstrap approach to creating synthetic demand data streams. The method enables structural modifications to the demand stream to reflect a wide range of possible changes in the demand pattern, allowing proposed system designs to be “stress tested” prior to installation. 3 - On-demand Warehousing Models On-demand warehousing matches independent operators who have extra space to ones who need warehouse space on-demand. We develop optimization models to determine which warehousing options (construction, long-term lease, or on- demand) is best to fulfill distributed demand requests over multiple periods. The models capture varying commitment and capacity granularity of the different options and are used to quantify the benefits of on-demand options in different customer environments. 4 - Modeling Parallel Process Flows in Intra-logistics Systems Debjit Roy, Indian Institute of Management, House Number 308, IIM.Ahmedabad, Vastrapur, Ahmedabad, 560078, India, debjit@iima.ac.in, Govind Kumawat We present an exact two-phase server based stochastic modeling approach for analyzing performance of parallel process flows found in several intra-logistics systems. In comparison to the existing sequential modeling approach, the proposed modeling approach gives substantially better throughput estimates for automated warehousing systems. 5 - Dynamic Slotting using Predictive Analytics Parvaneh Jahani, University of Louisville, 781 Theodore Burnett Court, Apt 2, Louisville, KY, 40217, United States, parvaneh.jahani@louisville.edu, Kevin Gue Dynamic slotting continuously adjusts the current state of the forward area with real-time decisions in conjunction with demand predictive analytics. Therefore, the layout of the fast picking area is updated over time with replenishment of the appropriate SKUs, as opposed to traditional methods that periodically reslot the forward area to reach a predefined target map. We explored the methods for demand pattern detection and demand forecasting as well as proposed MIP mathematical model for the dynamic forward-reserve problem for the first time. This model relaxes the major implicit assumptions of the static model and quantifies the effects of the static strategy versus the dynamic strategy. 361F Emerging Data Sources and Travel Demand Modeling I Invited: TSL, Intelligent Transportation Systems (ITS) Invited Session Chair: Alireza Khani, University of Minnesota, 136 Civil Engineering Building, 500 Pillsbury Drive S.E., MInneapolis, MN, 55455, United States, akhani@umn.edu 1 - Time-Dependent Origin Destination Matrix Calibration for Transit Network Modeling Jacqueline Nowak, University of Minnesota, Twin Cities, MN, 56267, United States, nowak123@umn.edu, Alireza Khani Outdated transit demand data presents a challenge for transit network modeling. This work aims to use Automated Passenger Count (APC) data to update an existing origin destination matrix to a time-dependent OD matrix for application in dynamic transit assignment. Results from the Twin Cities transit network are presented. SC53 Jennifer A.Pazour, Rensselaer Polytechnic Institute, 110 8th street, CI I.5217, Troy, NY, 12180, United States, pazouj@rpi.edu, Kaan Unnu

361D Digital Marketplaces and Platforms Sponsored: Information Systems Sponsored Session Chair: Jing Gong, Temple University, Philadelphia, PA, 19103, United States, gong@temple.edu 1 - Large Scale Cross Category Analysis of Consumer Review Content on Sales Conversion Leveraging Deep Learning Xiao Liu, New York University, Stern School of Business, 44 W. 4th Street, New York, NY, 10013, United States, xliu@stern.nyu.edu Using a dataset that tracks individual-level review reading and purchase behaviors, we quantify the causal impact of content information in the “read” reviews on sales. To extract content information, we apply deep learning natural language processing models and identify six dimensions of content. The comparative advantage of deep learning model is that it sifts category-specific content features across a wide range of product categories without human intervention or domain knowledge. We find that aesthetics and price content significantly affect conversions. Counterfactual simulation suggests that re- ordering reviews can have the same effect as a 1.6% price cut for boosting conversion. 2 - From Automobile to Autonomous: Does Self-driving Improve Traffic Conditions? Yingjie Zhang, 401 Shady Ave, Apt C-704, Pittsburgh, PA, 15206, United States, yingjie2@andrew.cmu.edu, Jinyang Zheng, Yong Tan Recently, Uber launched its autonomous car service. While debate has surrounded its impact, limited empirical work has been devoted to investigating the effects of this adoption on the society. This paper makes the first step to exploit this entry effect, using a DID approach applied on a unique accident dataset. We find a significant decrease in traffic accident volumes after the entry of Uber autonomous car service. A mechanism analysis suggests that this effect is mainly driven by the change in public reaction to the adoption. This pilot study contributes to the understanding of values in the adoption of autonomous car technology, as well as its social effects from the transportation perspective. 3 - Uber Might Buy Me a Mercedes Benz: An Empirical Investigation of the Sharing Economy and Durable Goods Purchase Jing Gong, Assistant Professor, Temple University, 1810 North 13th Street, 201C Speakman Hall, Philadelphia, PA, 19122, United States, gong@temple.edu, Brad N.Greenwood, Yiping Song In this work, we examine how the introduction of sharing platforms affects durable goods purchase. In particular, we use a unique dataset of new vehicle registrations in China and exploit a natural experiment, the variation in timing of Uber entry across different locations, to estimate the effect on vehicle purchase. Findings suggest that Uber entry is associated with a considerable increase (8%) in new vehicle ownership, suggesting that consumers are actively changing their stock of held resources in order to capture excess rents offered by these platforms. Further, results indicate that the effect of Uber entry varies considerably across gender, age, and vehicle types. 4 - Revenue Management in Crowdfunding Jiding Zhang, University of Pennsylvania, Philadelphia, PA, United States, jiding@wharton.upenn.edu, Senthil Veeraraghavan, Sergei Savin Crowdfunding, a mechanism in which funds are raised online using small donations from a large number of individual donors, has recently emerged as a popular approach to funding new ideas. In our paper, we model a setting where a creator of a crowdfunding project selects the amount of contribution it requests from donors and the duration of crowdfunding campaign with the goal of maximizing the raised amount. Our analysis provides project creators with detailed, practical, and intuitive guidelines on how to successfully manage the revenue generation process in a crowdfunding campaign. 361E Warehousing and Distribution Sponsored: TSL, Facility Logistics Sponsored Session Chair: Debjit Roy, Indian Institute of Management Ahmedabad, Ahmedabad, 560078, India, debzitt@gmail.com 1 - Hyperconnected Urban Deployment and Delivery of Large-items Nayeon Kim, Georgia Institute of Technology, Atlanta, GA, 30332- 0205, United States, nkim97@gatech.edu, Benoit Montreuil In this paper, we contrast a Physical Internet based hyperconnected deployment and last-mile delivery of large-items over an openly shared logistics network with SC52

86