As I conduct research on the global warehouse automation market, I hear statements such as “warehouse automation demand from Amazon and other e-commerce behemoths is draining automation providers of available resources and capacity.” Furthermore, Amazon purchased Kiva to capture essentially all of Kiva’s output capacity for its own use. So when I signed up for a tour of Amazon’s fulfillment center in Fall River, MA, I expected to witness large amounts of advanced automation. This wasn’t the case. The reason? This Amazon facility fulfills large, heavy, atypical items. Have you seen the Christmas special about the island of misfit toys? Well this was the warehouse of misfit items.
Non-Conforming (Large) Item Fulfillment
In previous automated warehouse tours, I have been informed by tour guides about the range of items fulfilled from the operation (size, weight, shape, rigidity, etc), the location of given SKUs (fast movers, etc.) and then the discussion shifted to the process steps of the operation at hand. I didn’t give a second thought to the process for fulfilling items outside of the operation’s parameters. But highly automated warehouses are not well suited to efficient exception handling. They require a structured environment to deliver high performance. So how does Amazon fulfill items that don’t meet size, weight, shape, or other automation equipment guidelines? They fulfill these items from designated warehouses such as the facility in Fall River.
Amazon’s BOS7 Fulfillment Center
Amazon’s Fulfillment Center in Fall River (code BOS7) is 4/10 of a mile in length, with 1.3 million square feet of floor space. At the time of its opening in 2016, it was Amazon’s largest in the US. It is categorized as a large item fulfillment center, but it appears to support a range of items that do not conform to the load requirements for Amazon’s more standardized, higher throughput facilities. This facility runs for 20 hours a day, allowing 4 hours for equipment maintenance, upkeep, etc. It is an incredibly clean, well-lit facility. It’s worth noting that these are not just esthetic qualities. In fact, my colleague Steve Banker’s research showed these to be important determinants of a high performing warehouse.
Workers utilize a fleet of very narrow aisle lifts to store and retrieve items from the 40-foot racking in the facility. Although the lifts are primarily manually operated, there is automation used to support safety and accuracy. For example, sensors guide the lifts to the appropriate distance from the racks and other automation assures that each lift remains at a safe distance from the others while operating in a storage and retrieval aisle. Workers are tethered to the elevating platform to assure safety as they guide the retrieval of items from the racking.
My colleagues and I inquired about the slotting methods used to determine the storage locations for inventory. Surprisingly, there doesn’t appear to be any slotting optimization used at this facility. The storage locations are assigned somewhat randomly, based on slot availability and general dimensions of item, without consideration for proximity to complementary items, fast mover status, or other features associated with picking efficiencies. The tour guide stated that it simply wasn’t practical for many items stored in the facility, like the large bags of dog food stored in front of us, to be combined for shipping. He also noted that the same SKU could be stored in numerous locations across the facility. Overall, the facility seems to be organized for maximizing storage capacity rather than throughput. This leads me to believe that the potential efficiency gains from applying throughput optimization were insufficient to justify the costs of supporting the process.
Mezzanine and More
A portion of the facility contains a mezzanine structure on which pre-shipping conveyor operations and storage of long, thin, awkwardly shaped items are located. The long, thin items, like fishing rods are stored in Sonotubes of various sizes (Sonotubes are strong, cardboard tubes used for holding poured concrete in form while it sets). The facility staff is proud of this innovative solution to the storage of these awkward items. Finally, the conveyor system feeds a high-speed sliding shoe sortation system that directs items to chutes that lead to trailers at the dock being manually packed by Amazon staff.
I found the degree of automation in this facility to be limited compared to my expectations. However, I suspect that Amazon thoroughly evaluated it options for increasing efficiency and determined that manual operations combined with selective automation was the best path forward. I was also surprised at the lack of focus placed on picking throughput optimization. Once again, I suspect that Amazon evaluated alternatives and determined that optimizing storage density provided greater benefits than the efforts required for picking optimization. It appears that even optimization isn’t always optimal.