For this and other AI automation content follow or add me on LinkedIn David Hall | LinkedIn
Amazon Automation 2021 and Beyond
Strangely, over the last 5 years, Amazon’s headcount (HC) growth rate has been about 20% higher than their sales growth rate. It’s is odd because we think of them as an automation, innovation, and machine learning company who heavily use robotics and benefiting from one of the highest revenue per employee ratios in retail. While these features are all true, there has been a consistent trend which is counterintuitive and offsets the improvements of innovation and robotics: as Amazon has tripled revenue from $100B to $300B, their shipping and fulfillment costs as a % of revenue have meaningfully increased from 23% to 28%. This is substantial, and represents +$50B more cost of shipping and fulfillment. What’s in this number?
Some of the drivers of the increasing % shipping and fulfillment:
- Increased amount of third party (fulfilled by amazon) sales — Following the Alibaba model, Amazon has increased the number of third party “fulfilled by Amazon” products (3P), which are now greater than 50% of total warehouse volume. Interestingly, Amazon does not report the revenue from sales 3P items (AMZN is only the platform), but they do recognize proceeds from the ~15% of revenue charge to 3P sellers to play on Amazon.com. Because of this, Increasing 3P volume always increase the fulfillment cost as a percent of total reported (first party) revenue. However, I have calculated that shipping and fulfillment is STILL increasing over this time period vs. the total “gross volume” including first part (1P) and third party (3P).
- Scale up of DC capacity to meet demand — holiday surges (normally a 5%-10% bump), COVID (30% volume increase), additional safety investment (amazon announced $4–6B), and new DCs to meet new customers globally.
- Meet customer needs and build a competitive moat — Amazon invested a huge amount in delivering 2-day and later 1-day shipment. These are now customer expectations, which makes a large competitive moat.
Net, capacity investments to enable growth have ended with an increase in total HC of 960K since 2015. Assuming that ~80% are warehouse workers, Amazon now has 895K warehouse employees which roughly ties to the assumption of 1500 employees per warehouse/distribution center.
The increase in investment ahead of revenue also came with a decrease in productivity (rev per worker) of 35%. However, rather than considering this a burden of additional expense, I see this as a longer-term investment cycle that will pay dividends as automation increases. These investments have enabled amazon to build a large moat with quick delivery and immense capacity, which will open the door for future savings from automation needed to reinvest in their continued growth.
Currently, the upside from automation in the warehouse is $18B of savings, which I believe will begin in toward the end of 2021 as amazon has reached the end of another capacity growth cycle.
The last capacity cycle
In 2015 and 2016, Amazon investment in HC significantly outpaced revenue (by 30%) as warehouse workers doubled from 130K to 260K. At the time, Amazon had been a relatively unprofitable retailer, never posting more than $500M profit per quarter (for reference, Walmart typically posts about $3B net income per quarter). However, this cycle of investment ended with 2 breakout years of profitability, posting $3B and $10B annual net income sequentially.
In part, this growth in profitability was enabled by automation and revenue growth finally surpassing HC growth. Capacity that was built out in the 2 year cycle was eventually put to work. Nowhere is this more evident than Q4 holiday sales, when volume spikes and HC typically needs to flex up more with seasonal workers and profit is significantly helped by the ability to ship that huge volume with existing capacity of people and warehouses. In holiday 2017 and 2018, Amazon was able to do that.
A large change in 2016 was the introduction of Amazon Robotics. Amazon paid $775M to buy an autonomous robotics company called Kiva, which developed hardware and software to increase efficiency of the warehouse significantly. Per Deutsche Bank, the “click to ship” cycle was reduced from 60 minutes to 15 minutes with the introduction of Kiva robots. They enabled warehouses to hold 50% more inventory, and save $22M in each warehouse deployed. The overall savings from this investment would have been about $2.5B, of which a significant portion fell to the bottom line as profit. These kiva systems have only improved since this time.
Note, this productivity increase happened while Amazon was still hiring people and growing warehouse capacity, but they did so at a reduced pace. While Kiva automation is not 100% the driver of profitability, it certainly had a big impact and increased competitiveness.
The next cycle
In 2018 and 2019, Amazon began increasing investment in HC and capacity (again ahead of revenue) as further international expansion and 1-day shipping were introduced. At the time, international sales were pretty unprofitable and other large retailers were beginning to grow in the US ecommerce space. These investments were significant, but enabled the long term strategy. With hindsight being 20/20, these already look like great investments as volume in India has increased, international business is now profitable, and market share has increased from 35% to 50% in US (only accelerated by 1-day shipping). Amazon’s moat was already significant before these investments further increased Prime users and site traffic which also benefits other parts of the business (ads, AWS).
Then COVID hit. At the time, Amazon announced a $4B investment (Amazon pegs COVID-19 costs at an estimated $4 billion next quarter | TechCrunch) into COVID related capacity. This undefined investment has since increased to $7B, and represents a large amount of capital invested to keep workers safe (safety glass, PPE, safer warehouse processes) while expanding capacity to meet increased quarantine shipping demand (for Amazon products and AWS servers, holiday spike).
During this time, Amazon has continued hiring at an accelerated pace. As of their last quarterly earnings, amazon has 1,125,300 employees, making them the second largest corporate employer in the US.
With continued quarantine, Amazon will likely further dampen investor expectations of profitability due to the ongoing investment in capacity, safer warehouse processes, and potentially investing in vaccines for warehouse workers (which Amazon has asked to be considered front line essential workers). When these investments end, I would look at this $4–7B of additional annual spend as dry powder to reinvest in efficiencies and automation in 2022. I expect that the capacity investments (in HC, capital, and AWS servers) will begin to pay large dividends at the end of this investment cycle as revenue continues to grow against slowing employee growth.
Overall, employee count will continue to increase as Amazon is growing internationally, increasing distribution in US, and will enable reduced delivery time to a number of hours. This will dramatically increase their moat vs other retailers as anyone looking to catch up with Amazon will need to be willing to invest 30% of their revenue back into shipping for a significant amount of time, leading to billions of dollars of profit loss and cash/capital expense.
While it is the right thing to do, I think the 2020 investments will also benefits Amazon long term as it reset investor profit expectations and could enable large efficiencies post COVID.
Exiting this 3 year cycle, Amazon cost of shipping will be 30%+ as a percent of revenue (first party revenue). It is only 14% of total gross sales, which includes an estimate of 3P revenue. Once Amazon reaches that point, there will be a massive upside opportunity from driving warehouse efficiency, which they will likely do through further automation and robotics.
Automation for profit
Warehouse costs represent nearly $90B per year at Amazon today. Of this, I estimate about $20B is cost of warehouse employees (based on some fuzzy wage and employee estimates). With this increasing cost, one of Amazon’s largest profitability upsides is through increased warehouse efficiency. Today, there is a huge potential automation upside for Amazon given recent progress in AI and automation.
First, all repetitive picking, packing, and sorting operations can already be done by robots, just pretty poorly. Progress here continues every year, especially on item picking/packing which is a task typically reserved for human dexterity and opposable thumbs. I believe there will be continued automation in these key components of the ecommerce supply chain as well as other more complex tasks.
Examples of automation in Amazon’s warehouses that could drive meaningful efficiency.
- Robotic arms stacking higher racks — most warehouse efficiency is looked at per square foot, but should probably be looked at per cubic foot. There is lots of wasted space above the height that humans can reach. Amazon’s most automated warehouses are beginning to take advantage of this.
- Automated item picking — some companies are getting closer to finding this holy grail of warehouse automation. Robotic arm + camera item recognition could improve warehouse efficiency and reduce worker injury.
- Machine loaded trucks — inbound/outbound loading is a large part of the work in warehouses and relies on our human ability to play Tetris well. Longer term, this could be done mechanically, but here are some interesting concepts existing today.
- Autonomous workflows — longer term lights out warehouse concept where products are brought in in autonomous trucks (Amazon bought and autonomous car company called Zoox) and shipped out in drones (Amazon’s funny beehive shipping center patent).
Impact and Risk
If further automations are able to drive additional 20% efficiency, Amazon is set up for $18B savings as they will be able to hold additional inventory or absorb a few quarters of volume increase with the same number of warehouses/staff. If these efficiencies are delivered, shipping as a % of revenue would decrease to 25% and 12.5% as a % of total gross sales (include 3P).
Over time, I still believe Amazon will be on trend to become the largest employer but they will be doing it with higher revenue per HC, more creative jobs, and increased safety.
Obviously, this is not without risk as robotics and automation reduce the human capital needed per $ of revenue. Amazon is replacing a low efficiency (high HC per $ of revenue) industry in brick-mortar retail at nearly a 2–3x efficiency. These net job losses will be replaced over time as the workforce shifts, re-skills, and growth drives additional hiring. However, there is risk due to the speed of industry transformation over time, defined by our ability to re-skill workers or provide safety nets during frictional unemployment. We will need to work on this as a collective industry and society.
As a parting gift, a link to my favorite brand and Amazon analyst, Prof. Scott Galloway. His videos talk a lot about these automation and technology opportunities/risks…
For this and other AI automation content follow or add me on LinkedIn David Hall | LinkedIn