A Tour of an Amazon Warehouse: The Value of Genchi Genbutsu and of Travelling Together
By Erick Rawlings, Alex Chown, John Raus, and Jack Tobin
In 2019 Amazon began offering tours of its fulfillment facilities across the country. A group of us from the Investment team decided to take Amazon up on its offer and toured their Fall River, MA fulfillment center. This blog is intended to share with you our findings and observations from that trip and is also structured in a way to highlight two broader beliefs in our investing approach: (i) there are benefits “to go and see,” akin to the idea of genchi genbutsu in Toyota’s Production System, and (ii) travelling together (in pairs or sometimes more) offers unique takeaways from varying perspectives.
Takeaways from the Group
First, we’ll tackle the shared observations from the tour and interesting statistics about the facility.
- The Fall River facility is what Amazon calls “non-sortable.” That means this fulfillment center deals with the large, bulky orders. Inventory is shipped to this location, where it is stored, packaged per new orders, and shipped to delivery centers and other fulfillment centers.
- All Amazon fulfillment centers are named after the nearest airport call letters and a number, and each has a mascot. The Fall River center is “BOS7” and its mascot is an anchor given Fall River’s nautical history. The first thing shipped from the BOS7 facility was, symbolically, an anchor.
- Sheer size
- The facility is 0.4 miles long
- The total square footage is the equivalent of about 25 football fields
- The facility handles 400,000 SKUs
- The highest volume item: diapers
- The minimum amount of time from order to out the door is about 2 hours
- The facility employs a total of 1,300 employees which will flex to near 2,000 for the holidays
- Operations data
- The facility runs nearly 24/7—it stops running 2 hours each day for maintenance. It is closed only Christmas day
- Each employee gets two 15-minute breaks and a 30-minute lunch per shift. There is a rotation of food trucks and restaurants available in the cafeteria
- There are metal detectors that employees must go through upon exit—and there is only one main entrance/exit (though lots of emergency exits throughout)
- Wages are $15/hr
- Employees get full health benefits effective on day one of employment
- Amazon offers employees a 95% education reimbursement with no strings attached/back-end commitment required
Takeaways from the Team Members
The above represents our common observations, the things each of us noted and discussed post-tour. What follows are the unique takeaways and observations from each team member. From them, we believe it becomes clear how beneficial it is to go and see, and why, by having multiple attendees, the conclusions and observations are that much richer and powerful.
Since negative news elicits stronger and more sustained reactions than positive news, it’s easy to be a pessimist while sitting at your desk. This is why genchi genbutsu is so valuable. Headlines such as the “trade war,” “earnings recession,” and “end of the bull market” dominated the news cycle on the day of our visit, yet watching those packages hum down the assembly line left me with a sense of exuberant optimism about US innovation and growth. This contrast reminded me of two similar experiences from earlier this year. The first being my visit to mainland China, where the overall pace of the economic activity I witnessed was much higher than the “doomsday” growth headlines I saw when I returned. The second was my recent trip to London, where I didn’t hear the phrase “Brexit” once over the course of ten meetings. Although these micro-observations pale in comparison to data-backed arguments, they help provide a sense of balance when compared against the media’s persistently negative tone, and the seemingly constant obsession over the “next market correction.”
For a more detailed overview of the tour experience, this article provides a great summary of a “sortable” tour. The most notable contrast between the author’s tour (“sortable” facility) and ours (“non-sortable” facility) was the reliance on robots to “pick” in the sortable facilities versus the use of manual labor in the non-sortable centers. In my opinion, this provides insight into the current complexities and cost of warehouse robotics. Since the products we saw have a much wider size variation than sortable items (e.g. those <16 inches), non-sortable facilities present a more logistically complex challenge, and thus require automation with more sophisticated software. As it stands today, human employment is presumably more economic than the capital expenditure required for full-scale automation—but this transition appears inevitable as the technology improves and/or the cost comes down.
As one of the largest employers in the US (650,000 as of 2018), I expect the rotation from manual labor towards robotics to keep the political pressures on Amazon as this transition occurs. The skeptic in me wonders if the heavy marketing of these tours (two members of the Athena team independently received targeted ads) is rooted in Amazon’s desire to influence the public perception in this debate—as propaganda regarding employee benefits (above-average wages, maternity/paternity leave, educational reimbursement, etc.) was sprinkled throughout our tour, as well as others who have written about the experience.
Lastly, as impressive as the BOS7 facility was, the company’s financial statements continue to show the superiority of Amazon Web Services (AWS) relative to the eCommerce division that the company is traditionally associated with. For example, the eCommerce business generated $140B of revenue in 2018, approximately six-times that of AWS’ $25B in revenue. Yet, they both contributed the same absolute dollar amount in operating profit ($7B)! This equates to a 5% operating margin for eCommerce, and 18% for AWS. Using a 30x multiple on the $7B income figure gives AWS a standalone value of $210B value, or roughly one quarter of Amazon’s $870B market cap. However, given AWS’ 50% year over year revenue growth rate, most investors are presumably using a higher multiple for this business segment. On a standalone basis, AWS is clearly a better business… but I imagine a tour of an AWS data center is slightly less fun.
There have been a few headlines recently about employee injuries at Amazon warehouses. While I caveat that my perspective was limited to a single one-hour guided tour, my takeaway differed from these headlines as I focused on the level of safety that Amazon enforces at its facility. When we began the tour, we were given bright-yellow vests. We soon discovered that all who enter the warehouse must wear these vests, and that workers must use Amazon-approved safety products. There were also numerous vending machine-like kiosks throughout the facility at which any worker could replace gloves, vests, earplugs, etc., free of charge.
We learned more about safety as the tour progressed. In some cases, these safety protocols were explicitly stated to us—namely that Amazon requires pedestrians to have three-points of contact (two feet and one hand on a rail) on all stairways. Others were more implicit, like the two additional tour guides that silently supervised us to make sure that we were standing in pedestrian-approved areas. The only time one spoke was when I stepped 6-inches into an unapproved space near a conveyor belt.
Amazon uses multi-ton machines to retrieve items, known as “picker machines.” Since these machines operate 22 hours each day and are the primary means that Amazon gathers merchandise, there are various safety protocols that operators must adhere to:
- Beeping a horn at every aisle’s intersection
- “Tying in” to units via a safety harness to prevent falls, as units can extend as high as a 7-story buildings
- Stacking merchandise in an Amazon-approved manner to prevent any drops (we saw first-hand one of many safety supervisors emerge when an operator didn’t stack items correctly)
Machines themselves also have embedded safety protocols. It can only follow a pre-programmed path via subterranean “guide wire,” ensuring that two pickers can operate on opposite sides of the same aisle with exactly 24 inches of clearance. Operators can only control speed and general direction and cannot reverse a machine, but must make a full loop around an aisle if they go past the merchandise or need to get to the opposite side (pickers can only access merchandise from the left side of a machine). Lastly, machines have a blue-light that constantly adjusts to denote 20 feet in front of the machine—regardless of how high the machine has extended—as a minimum distance requirement between vehicles.
From my perspective, all of these protocols demonstrate that Amazon distinctly emphasizes safety within the warehouse over cost and efficiency. Even with that emphasis, it’s quite amazing that Amazon can deliver an item to a Boston location with one-day delivery.
The way this single facility fits into the greater Amazon distribution network was impressive to me. The network is comprised of many similar fulfillment centers around the country, connected to our doorsteps by several layers of distribution channels. Packages are routed quickly through the system in a sophisticated fashion—our guide informed us that the time it takes for an order to be ready on a truck could be as little as two hours. Amazon has clearly focused a lot of energy on streamlining this process.
Within the fulfillment center itself, the systems for producing a deliverable package from a vast amount of inventory varied in level of efficiency. Since the facility processes “non-sortable” goods, it made sense that the irregularly sized items had to be selected by hand by a large team of warehouse associates. I was surprised to find that the system for storing and locating goods was randomized. Products are essentially placed at random locations throughout the fulfillment center, and a computer directs associates to collect items one by one. There was no fancy algorithm that tracks associate location to minimize travel distance, as I had expected. One would think that there could be some serious time and cost savings from optimizing the gathering of inventory.
I was a bit puzzled by the disparity in operational efficiency throughout the organization—Amazon’s distribution network is highly sophisticated and clearly optimized from the top down, while there seemed to be room for increased efficiency in their process for picking individual inventory items. I'd be curious to take another tour in a few years’ time to see which processes have changed and which have not.
I found the following contrast amazing: we think of the ubiquity of Amazon yet walking around that huge facility with the 400,000 skus and none of the products we saw said "Amazon" on them. This led me to consider, “What is the Amazon brand?” My initial conclusion is that it’s all about convenience. Or, as Amazon says at the entrance to the facility, "We start with the customer and work backwards. We work vigorously to earn and keep customer trust." And by trust, it's really about convenience.
Taking this contrast a step further, I then wondered how the cost of brand maintenance might differ between Amazon and the brands it ships. That is, what does a company like P&G who makes the diapers that BOS7 ships so much of have to spend on R&D as a proxy for brand maintenance to ensure relevance versus how much Amazon spends to deliver its convenience? To answer, I took a quick look at the gross margins and the return on research capital (which is the amount of gross profit earned in the current year for each dollar of R&D spent last year) for Amazon and P&G.
Here’s what I found (note that Amazon’s figures include Amazon Web Services and Whole Foods):
- The Gross Margin for P&G is 50%, and very stable year over year
- The Gross Margin for Amazon is 35% and more variable year over year
- The Return on Research Capital for P&G in 2018 is 3.7 ($33.5bn divided by approximately $9bn, comprised of $2bn in R&D and $7bn in marketing)
- The Return on Research Capital for Amazon is 1.8 ($60bn divided by $33bn, comprised of $23bn for R&D, for which I used “Technology & Content” as the R&D proxy, and $10bn for marketing)
What's the point? It appears the cost to maintain relevance for a retail distributor like Amazon is potentially higher than that of the relatively more niche product company like P&G.
A final thought: At the end of the tour the guide touted how amazing it was that Amazon started in a garage in Seattle just 25 years ago, as a bit of "look us now." I agree, it is amazing, just as it was amazing what Wal-Mart and Target did to retail 20 years before that, and what department stores did to ‘mom and pop’ stores before that. Which all leads me to think, what's next for product delivery? And in the end, is it all just finding more convenient ways to deliver you your Tide and Pampers, the products where the gross margins and returns to R&D appear to be higher?