When Dave Alperson got his first job at an Amazon warehouse in 1997, as a temporary hourly employee, it involved walking around the warehouse with a list of where to find products—mostly books—that customers had ordered.
Twenty years later, as a regional director of operations for Amazon in Indiana, he oversees 18 warehouses that barely resemble where he started. Amazon now sells millions of products; each of its 149 warehouses ship tens of thousands of them each day; and those warehouses now look like live-action games of Chutes and Ladders—whizzing with a meticulously coordinated system of conveyor belts, slides, and machines that do everything from attach labels to boxes to check weight for quality control.“As soon as people realized, you can order something and get it tomorrow, that turned the industry upside down.”
In the process of building this elaborate system, Amazon has completely redefined warehouse efficiency and customer convenience. Through its Prime membership, it has promised tens of millions of customers free two-day shipping on more than 100 million products, and, last year, it shipped 5 billion items to them. “That was the major innovation,” says Daniel Theobald, who cofounded a warehouse robotics company called Vecna in 1998 and counts major retailers and logistics companies as clients. “As soon as people realized, you can order something and get it tomorrow, that turned the industry upside down.”
The core of this disruptive efficiency, though, is not Amazon’s automated shelf-moving warehouse robots, which is the innovation that gets the most attention. And it isn’t, on its surface, something that you would associate with a well-oiled machine. It’s not even a breakthrough technology. In fact, some version of it was already in place when Alperson worked in Amazon’s early warehouses.
What makes Amazon’s warehouses work is the way they organize inventory: with complete randomness.
How ecommerce made randomness efficient
At a traditionally organized warehouse, when a shipment of, say, toothpaste arrives, an employee looks up where the toothpaste shelf is located, and then moves the box to that shelf.
When a box of toothpaste arrives at an Amazon warehouse, though, the process works differently. An employee removes each individual tube and stows it wherever he finds open space. Placement is completely random. Items aren’t organized by where they’re being shipped; they aren’t—aside from very big items—organized by size; and they aren’t organized by the type of customer who is likely to order them. A shipment of 50 tubes of toothpaste may ultimately be distributed to and stored in 50 different places.
On a visit to an Amazon warehouse in New Jersey last year, I saw a box of Irish breakfast tea, next to a board game called “Quick Cups,” next to a Hamilton Beach Juicer.The random system has been in place since early on in Amazon’s 24-year history, and to a casual observer, the result appears chaotic.
This random system has been in place since early on in Amazon’s 24-year history, and to a casual observer, the result appears chaotic. The reason it makes sense to group these random products together has everything to do with technology: the speed and frequency with which customers order online, and the tools that Amazon has developed to keep track of every item in its vast warehouses.
First, random storage makes finding the toothpaste faster in an era of on-demand efficiency. If there were a dedicated “toothpaste shelf” and someone ordered toothpaste, a “picker”—how Amazon refers to employees who gather items—would need to travel there, whether he were 10 feet or 100 yards away from that location. But if the warehouse stores toothpaste in 50 different locations, there’s a much better chance that there’s a tube close to some picker. There’s also a greater chance that the second item the customer ordered is also nearby.
“With the millions of items that we ship, every opportunity to improve a process by a second is relevant,” Alperson says.
Randomness is also preferable when it comes to managing the wide range of items customers now order online—most practically by saving space. Amazon warehouses carry a huge variety of items that can be ordered at any moment, but they do not carry a huge number of each item. “They may only have one box of Cheerios,” says Tom Galluzzo, the founder of Iam Robotics, which makes warehouse robots. “If you were to have a space for every product, you would need a gigantic warehouse.” Amazon’s largest warehouse is already 1 million square feet, which is about 17 NFL football fields in size. Reserving empty space on the “toothpaste shelf” while waiting for the next shipment of toothpaste would mean its warehouses would need to be even bigger. It’s more efficient to use any free shelf space available.
That Amazon and other ecommerce companies sell directly to customers (as opposed to retailers) is also a factor in making randomly stowing items efficient. They ship a single tube of toothpaste to a customer, not a box of 50 tubes to a store, so there’s no reason to keep all 50 tubes together in the warehouse. They’ll be unbundled before they’re shipped out, in any case.
Amazon didn’t invent this strategy, but the company has employed it at a scale that has never been seen before.
The company’s inventory management system, which it built itself, makes this random storage philosophy possible. At Amazon’s warehouses today, before a worker places something on a shelf, she uses a handheld computer to scan a barcode on both the product and the shelf. This allows the computer to keep track of where every item is located.
Adding robots to random
In 2012, Amazon acquired the company that makes its robots, called Kiva Systems, for $775 million, and since 2014 it has deployed more than 100,000 of the machines in 25 of its 149 warehouses worldwide. Though these robots are often hailed as the key to Amazon’s efficiency, they wouldn’t work as well without Amazon’s simple system of random storage.Amazon workers in robot-equipped warehouses don’t need to bring items to the shelves for stocking—the robots bring the shelves to them.
Inventory at warehouses with robots is stored using the same strategy as in Amazon’s non-robotized warehouses. The difference is that Amazon workers in robot-equipped warehouses don’t need to bring items to the shelves for stocking—the robots bring the shelves to them. Rather than humanoid robots, they’re small box-like machines that pick up shelves and move them along a grid, shuffling other shelves within the grid to make way for whichever shelf the computer system says contains available space. As with non-robotized warehouses, the workers then scan the item, place it on one of the shelves—wherever there is free space—and scan the shelf so the computer knows where the item is located.
Kiva’s robots have taken Amazon’s random organization strategy and made it even faster. Deutsche Bank estimated in 2016 that the “click to ship” cycle—the time it takes to pick a product from the stacks, pack it, and ship it—was around 60 to 75 minutes when employees manually handled the process. With the aid of robots, the same job could be done in 15 minutes. Warehouses equipped with Kiva robots can also hold 50% more inventory than those without them.
Theobald, whose own warehouse robotics solution works with both the traditional and random strategies, says that when he started working in the industry 20 years ago, about 20% of warehouses used the random strategy. Now, he says it’s around 50%. Random stow makes sense when there’s limited space, or when the timing or quantity of inventory is unpredictable.Twenty years ago, about 20% of warehouses used the random strategy. Now, it’s around 50%.
“Retail is fickle, especially when you are shipping to individual customers like you and me (rather than, say, stores),” says John Bartholdi, a professor of industrial and systems engineering at Georgia Tech. “Both Amazon and Walmart have huge populations of product and move lots of it, a few pieces at a time. There is huge product churn as they reconfigure their offerings constantly. It is impossible to plan and manage space in such a dynamic environment.” With products offered in stores changing quickly, it makes even less sense to save dedicated space for one product or another.
After Amazon moved the bar for immediate gratification by offering free two-day delivery to Prime customers, many retailers have meanwhile adjusted partly by spreading a product not only throughout a warehouse, but throughout different warehouse locations, so it has a better chance of being closer to the customers who order it, and thus less expensive to ship quickly. Many companies aim to reduce labor costs by automating more of their warehouses so they can spend more on faster shipping.
Amazon has also focused on improving its delivery process over the “last mile,” from warehouse to doorstep. The company is building its own cargo airline and has been experimenting with drones around the world, both efforts that have attracted the attention of tech enthusiasts. But while the key to a speedy delivery is making it as streamlined as possible, it all starts at the warehouse, where Amazon’s strategy rests on a simple decision to leave stuff wherever there’s space.