I went on a tour of an Amazon “Fulfillment Center” today. That’s a million+ square foot building where Amazon assembles and ships orders.
The Amazon Fulfillment Center is a massive operation, both in size and complexity. Trucks bearing goods from manufacturers are unloaded, the contents categorized and placed in inventory, the location and status of each item is tracked, customer orders are picked, assembled and shipped constantly in a carefully constructed ballet of smoothly coordinated functions. There are such centers all over the country, and each new one that is constructed builds on what has been learned before. Systems, methods, machinery, and policies or constantly reevaluated and modified to both improve the operation and correct shortcomings. As you go through the Amazon facility, you see hundreds of clever and unexpected tweaks, subsystems, and subroutines that have been developed over time to increase the efficiency and workability of the system. For instance, since the beginning of time, warehouses have been organized by placing objects in categories so they can be located easily. Each cubbyhole (a small segment of shelf space separated by a divider) on the stock shelves of a book warehouse, for instance, would have copies of a single book by category, author name, or file number. A warehouse stocking different products, such as Amazon, would have sections devoted to electronics, household products, clothing, etc, similar to a Sam’s Club. All nice and need and traditional.
But Amazon does it completely differently. If you walk among the inventory shelves in an Amazon fulfillment center, you are shocked to find that the shelves are not organized by any visible system. A cubbyhole might contain a book, along with a box of baby food, a portable radio, a pair of gloves and a stuffed toy. The place looks like the world’s biggest attic. How can they possible find anything in the chaos? The answer is simple: scanners, bar codes and computers. Amazon realized the the main reason for inventory location by item type was to enable finding the item when needed, but why not let technology do the work? Each item has a bar code and each cubby hole has one as well, so once the item is stocked, the computer knows exactly where it is regardless of what is around it. When an order comes in, the computer finds the picker closest and sends him the item description and its location on the shelves. The item is picked, scanned, and placed in a yellow plastic tub that also has a bar code. Once the tub is full, it is placed on a conveyor belt to an order assembly area, then to packaging, then to shipping. Along the way, the package is automatically weighed and compared against the expected weight of the items ordered to catch errors in order fulfillment. When items are restocked, the stocker places them on any available open space, then records the location by scanning the bar code on the shelf cubby and the item. The computer will know where to find it when it is ordered.
This system is more efficient than carefully placing items by category, and has other advantages. Multiple identical items wind up scattered all over, but this makes it less likely there will be a bottleneck for more popular items sought by multiple pickers simultaneously. The random stocking makes a more efficient use of shelf space, with less of it sitting empty awaiting replenishment of a specific item. The computer will even prevent you from placing items that are identical, but of different colors in the same or adjacent cubbyholes, to avoid possible confusion when the picker needs them. Of course, the work is repetitive and can be stressful due to pressure to keep up the demanded pace. If you don’t meet the work standards, you could be in trouble.
The computer knows that as well.