Stochastic models for omnichannel retail order fulfillment

Abstract

With a massive increase in the number of internet users and rapid urbanization, online retail is on a path of unprecedented growth. The online retail sales in India - with over 600 million internet users - is expected to grow from US$ 55 billion to US$ 125 billion by 2024. Online customers not only demand a seamless omnichannel experience in placing an order from a variety of listed product assortments, but also expect sharp delivery windows for variable order sizes even during the busiest holiday seasons. These demanding performance expectations make the role of order fulfillment operations at the distribution warehouses invincible. Omnichannel retailers are rising to these challenges by automating warehouses, gaining visibility on real-time inventories, and developing the right order fulfillment strategy. The decisions for an omnichannel retailer include: 1) Selection of the storage and material handling system technology for the distribution center, 2) Operational policies for the distribution center such as right order-batching strategy and organization of pick stations, and 3) Operational policies for the retail store such as the inventory replenishment and online demand allocation decisions. In many warehouses, robotized shuttle-based storage and retrieval technologies have replaced the traditional manual picker-to-stock systems. The impact of these technologies on the performance of integrated storage-order picking systems, particularly with multi-line orders (more than one SKU) and item commonalities, is largely unknown. In this dissertation, we propose three studies on omnichannel retail order fulfillment. In the first essay, we propose stochastic models and solution methodology for the performance comparison of different batching strategies for multi-line orders in an automated e-commerce distribution center with the stock-to-picker system. The second essay further considers the order waiting times and proposes a semi-open queuing network for an integrated storageorder picking system. The storage system includes a robotized shuttle-based storage and retrieval system upstream and a pick station downstream. In the third essay, we study the inventory replenishment and online demand allocation decisions for an omnichannel retailer with the ship-from-store order fulfillment strategy, using the Markov decision process (MDP).