Amid the pandemic, online retailers face a massive and rapidly expanding problem: their costs have been spiraling as they struggle to fill millions of orders from relatively few warehouses. Some are responding with a warehouse-building spree: Amazon reportedly plans to open 1,000 of them in cities and suburbs across the United States.
But rather than build more warehouses, online retailers should focus on reducing costs by routing orders more cheaply, according to Chicago Booth PhD candidate Yanyang Zhao, Alibaba Group’s Xinshang Wang, and Booth’s Linwei Xin. Their research could boost profitability for such companies.
US online retail sales of physical goods in 2019 totaled $365.2 billion, write the researchers, and the cost of filling orders is taking a big bite. Amazon’s shipping and warehouse costs surged to almost 28 percent of net sales in 2019 from less than 16 percent in 2009, they note.
Some of the growing cost reflects the often-complicated calculations faced by online retailers, including Amazon. Should they take more time deciding how to fill orders most efficiently at the cost of slower deliveries? Should they try to factor in demand forecasts? Should they just ship goods out as orders come in, even if they have to eat the higher costs of splitting orders into two or more shipments from multiple warehouses?
E-tailing giants such as Amazon and China’s Alibaba use a two-layer distribution system that involves forward distribution centers and regional distribution centers. FDCs are closer to customers but tend to have lower inventories and carry fewer items. This makes it “notoriously challenging to make effective real-time fulﬁllment decisions” when an FDC can fill only part of an order right away and the retailer has to determine how to split up the shipments, the researchers write. Amazon’s reported building spree is of smaller warehouses, or FDCs.
Zhao, Wang, and Xin constructed a model assuming multi-item orders, no demand forecasts, and a two-layer RDC-FDC distribution system. Their model also allows orders to be split into two shipments at most, which aligns industry practice. They then conducted a series of numerical experiments using the model to analyze solutions for different situations. They find that companies should go with the least expensive fulfillment option in the moment without considering the impact on future orders, a strategy they call a “myopic policy.”
Consider a hypothetical customer in Maryland who orders two best-selling items used for remote learning amid the pandemic: a Blue Yeti microphone and a Logitech C920S webcam. In a perfect world, the distribution center closest to the customer would carry both items, and the cheapest option would be for the retailer to bundle and ship the products together.
However, it’s possible that the nearby distribution center carries the microphone but not the webcam. Meanwhile, a distribution center in California carries both, but bundling and shipping the items cross-country would be expensive. If it’s cheaper for the retailer to split the order so that the microphone ships from the center in Maryland and the webcam travels from California, the retailer should do that, the research suggests.
For many e-tail scenarios, the model achieved consistent performance results comparable with more-complex algorithms that rely on demand forecasts. The findings may offer a key for keeping large retailers in the black, and helping smaller ones to stay open, Zhao says. That could be especially meaningful for retailers struggling to cope with unreliable demand forecasts.
“How many businesses will survive as just traditional stores?” he asks. “The margins are thin, so it may sadly be efficiency or close your doors.”