The complexity of Amazon’s retail operation is staggering. Amazon sells ~600 million items, which flow through 175 fulfillment centers (FCs) and ship to pretty much anywhere on Earth. We enabled this with one part customer obsession and two parts operational excellence.
When you place an order on Amazon, you initiate a complex set of automated decisions and processes. When you dive into this complexity, the accuracy, efficiency, and scale of Amazon’s operational excellence are amazing.
For example, when your order includes two or more items, the first step is a complex calculation to determine which fulfillment centers (FCs) to use for picking, packing, and shipping. This depends on the location and in-stock status of each item ordered across 175 FCs, the distance from your home to each FC, and how soon the item might be received from a manufacturer or seller at each FC.
This algorithm seeks to balance the promised delivery speed with the cost of delivery options, as well as the economics and customer experience of bundling all items into a single shipment or splitting them into two or more shipments.
Then, the algorithm has to consider Amazon’s vast options for transporting items by van, truck, or plane, as well as its own last-mile delivery or 3Ps like USPS and UPS. All of these calculations rely on Amazon’s inventory record accuracy, transportation systems accuracy, and third-party data such as supplier item-level data for inbound shipments.
In 2000, I took on the role of Divisional Merchandising Manager for the Video category. I managed the team responsible for our supplier relationships (movie studios), the terms and conditions for buying from them, and the management of optimal inventory levels for tens of thousands of movies and TV series.
This role thrust me into needing to understand the complex fulfillment and supply chain systems, including attempting to grasp the complexity and variables of the algorithm I described above, which was named “decide_warehouse.”
I recall many lengthy sessions with our then VP of Supply Chain, Jim Miller, where he described and explained these systems and algorithms. I found the complexity of these problems fascinating and loved the challenge of optimizing the buying-side process to improve our performance.
In 2000, we were better at e-commerce logistics than any other company. That, however, was a low bar. If one could compare the key input metrics such as order-to-deliver time, cost per shipment, mean squared forecast accuracy, inventory record accuracy, and percent of order “Switcheroos” (our quirky name for when we shipped the wrong item to a customer), the numbers would be drastically better today than 25 years ago even though the scale and complexity of Amazon’s operation have increased by multiple orders of magnitude.
(cont. in comments)
