Thesis Title: Hyperconnected fulfillment and inventory allocation and deployment models

Advisors: 

Dr. Benoit Montreuil, School of Industrial and Systems Engineering, Georgia Tech

Dr. Walid Klibi, Supply Chain Center of Excellence, KEDGE Business School

Committee members:

Dr. Alejandro Toriello, School of Industrial and Systems Engineering, Georgia Tech

Dr. Alan Erera, School of Industrial and Systems Engineering, Georgia Tech

Dr. Eric Ballot, Centre de Gestion Scientifique, MINES-ParisTech - PSL

Date and Time: 3:00 - 4:00 pm EST (Tuesday) April 20, 2021

Meeting URL: https://bluejeans.com/790503442

Meeting ID: 790 503 442 (BlueJeans) 

Abstract:

Customer expectations for a faster delivery are increasing, with the prevalence of e-commerce and home delivery. In fact, many customers are expecting for same-day or x-hour deliveries now and offering such responsiveness becomes more and more critical for e-retailers to survive in a fierce market competition. However, many companies are simply lacking financial, physical, and/or operational resources to increase their responsiveness. Focusing on solving the challenges in the perspective of fulfillment and inventory allocation, we find a recently emerging concept, Physical Internet (PI), can potentially enable responsive yet affordable fulfillment for companies of any size by transforming asset-driven logistics operations to service-driven operations. This thesis investigates the potential of the hyperconnected fulfillment and provides an academic foundation for the relevant inventory operations to effectively satisfy the growing customer expectations for responsive deliveries.

In Chapter 2, a hyperconnected fulfillment and delivery system is designed in the context of the last-mile operations in urban areas. We present a comprehensive system and decision architecture of the hyperconnected system. Then, we design the scenarios which show a gradual transformation from dedicated to hyperconnected system in each thread of delivery and fulfillment. We conduct a scenario analysis using a simulation platform built upon the system and decision architecture where autonomous agents are optimizing their decisions and interact with the environment. The experimental results clearly demonstrate the potential benefit of the proposed system by concurrently improving often opposing performance criteria: economic efficiency, service capability and sustainability.

Chapter 3 tackles an inventory allocation problem among multiple sales outlets. Specifically, we analyze a case where a dropshipper allocates availability to multiple e-retailers via availability promising e-contracts (APCs). Under the APC, the e-retailers do not observe actual availability and this information asymmetry leads them to pose a promised availability threshold (PAT). PAT is a retailer’s internal threshold set for each product, when remaining promised availability falls below which no more customer orders are accepted. The dropshipper's APC problem with PAT is modeled as 2-stage stochastic program with two stochastic parameters: demand and PAT. We design and evaluate three contract policies differentiated by the allowance level for overpromising: guaranteed fill-rate, controlled fill-rate, and penalty-driven fill-ratepolicies. The numerical results show the penalty-driven fill-ratepolicy is the dominating strategy for dropshippers especially under a lean availability.

Chapter 4 tackles a network inventory deployment problem for responsive fulfillment. The physical availability of inventories near the delivery locations becomes necessary for very responsive deliveries, which requires a broad and dense fulfillment network. The open asset utilization and service-driven fulfillment operations of the PI can enable affordable access to a decentralized open fulfillment network. Here, we evaluate the benefit of such a decentralized fulfillment network for a responsive fulfillment and develop an appropriate inventory deployment model, which possesses a partially pooled demand and inventory structure induced by responsiveness requirements, as a variant of Newsvendor. We derive a pragmatic heuristic inventory solution, W-solution. The numerical experiments over both theoretical and empirical demand distributions demonstrate the advantage of decentralized network and w-solution over centralized network and allocation-based inventory model, respectively.