In Amazon’s empty fulfilment centre, department general manager Evan Shobe (Evan Shobe) sits in front of nine computer screens. This command centre is called the “Quarterback Table” (QB) internally. The fulfilment centre covers an area of approximately 15 football fields, and the complex operations are monitored in this command centre.
The dense blue dots represent the robots transporting products; the yellow patterns resembling the toilet signs represent the humans who load and unload the robots. The intricate green lines represent the process by which the conveyor belt quickly transfers the order to various stations on the production line, and finally to the delivery truck.
Located in the suburbs of Kent, Washington, BFI4 is Amazon’s flagship fulfilment centre. In addition to delivering products, it is also used to regularly receive company executives. The new CEO Andy Jassy has been here recently. We want to understand the workflow of the fulfilment centre after the buyer places an order.
This is Amazon’s first facility that can handle more than 1 million items a day. Ten years ago, the company’s state-of-the-art warehouse handled only one-third of the current volume. The continuous leaps in technology have allowed Amazon to take a few steps ahead of its friends such as Wal-Mart and Target (a US retail company). These competitors are now following many of Amazon’s practices over the years.
Table of Contents
Algorithm-the magic weapon for efficiency leaps
The core of this facility is not a physical robot, but the algorithm behind it, a set of computer instructions designed to solve specific problems. The software can solve many problems in the warehouse operation process: the number of goods handled by a single facility, the placement of goods, the night shift manpower required during peak holidays, and it is most suitable for truck delivery. Xiao Bo, the general manager of BFI4, said:
“Algorithms have made many correct decisions for us.”
The high degree of automation allows the supervisors of each fulfilment centre to manage dozens of employees, and this factory-style operation has also become the industry standard. Data from the US Bureau of Labor Statistics show that in 2012, a logistics warehouse manager supervised about 10 front-line workers. In 2020, after Amazon became the largest employer in the industry, this number almost doubled.
Such high efficiency naturally attracted a lot of attention, and competitors began to follow suit.
Although automation has improved efficiency, it has also attracted many criticisms. Critics are dissatisfied with the working conditions of Amazon’s hourly workers. Amazon’s algorithm tells workers what to do in the warehouse, sets productivity goals for them, and flags employees who do not meet the standards. An employee said in an interview that he feels like a gear embedded in a giant machine. The company “takes down the gear” and drives away, and only needs to send an email.
Amazon also admits that this algorithm is not perfect. Amazon claims that most of the automated processes in its facilities allow human supervision or intervention. Managers said that with strong software support, they can do more work, and the company is continuing to improve its business processes.
Maju Kuruvilla, the engineering director who left last year, said that Amazon noticed a few years ago that managers are sometimes just “faces hiding behind computers.” Kuruvira said:
“The managers of the performance centre rarely have contact with the employees. When the employee care is poor to a certain extent, it is when the workers want to form a union.”
A big chess game played by Bezos
When Jeff Bezos was just a bookseller, he started thinking about replacing labour with software. In a famous story in Amazon’s early days, book reviews and recommendations that were previously done by editors were replaced by code. The same work used to require manpower, but now it only needs to be coded by mining shopping patterns.
Similar procedures were later used to manage many aspects of Amazon’s daily operations, such as ordering, placing inventory, monitoring online markets, and so on.
Bezos is playing a long-term big game. He bet that the algorithm can handle certain tasks more efficiently or more stably than manually.
In the middle of the last century, Amazon regarded automation as the focus of the large-scale expansion of its packaging, transportation and delivery departments. Amazon logistics experts will reflect on their resumes the experience of
“reducing the cost of single product transportation by a few cents.”
In 2012, Amazon acquired Kiva Systems, an American manufacturer of automated robots. Amazon hopes to use Kiva to replace manual sorting of goods, which requires a large-scale redesign of the fulfilment centre.
BFI4, which went into operation in 2016, was one of the first facilities built specifically for small robots. 3,500 employees and 110 managers work in this assembly-line factory under the “monitoring” of an accurate tracking system.
The workers push the mobile conveyor belt behind the truck that transports the inventory to feed the product into a system. The system will automatically scan the incoming items, and then put them on the Amazon website for sale, and then pay the supplier after the goods are sold. After the products are purchased by the users, the workers stack the products on the shelves. The shelves are closely arranged in the dedicated area of the robot. After the order is placed, Kiva sends the shelves to a picking station where the workers pick the right products, put them in a bucket, and send them to the next area Pack and ship.
In the past, managers used Excel and intuition to calculate the manpower required for each site. But from around 2014, statistics from Amazon warehouses will be transmitted to the Seattle headquarters, where software engineers will process and automate them.
David Glick, the former head of Amazon Logistics, who oversaw the initial development of the software, recalls that the engineers created a program called AutoFlow. At first, this software was not feasible, because it would make a judgement of arranging 0.5 employees.
In the spring of 2019, Amazon executives instructed employees operating BFI4 to adopt AutoFlow’s recommendation, which is refreshed every 15 minutes. If managers find that the system instructions are wrong, they can not follow the instruction form, but in most cases, managers are asked not to interfere.
Xiao Bo, who was in charge of the warehouse transportation department at the time, said: “In the beginning, we asked to allow the program to make mistakes and let it learn from it. Our team was operating in a much better situation than the software at the time, so this decision was really hard to understand at that time.”
At first, AutoFlow always overreacted to changes in demand, allowing workers to quickly move to a new location and then return to their previous location a few hours later. This method of task allocation not only consumes the energy of workers but also wastes time.
But it does continue to iterate and upgrade, becoming more and more usable. Amazon now does not need to send someone to troubleshoot problems in every warehouse. The operation of all warehouses can be monitored in front of the office computer.
Engineers have also created other ways to improve operational efficiency and accuracy. In the past, workers had to scan barcodes to know where the goods should be placed on the shelf. Now, video cameras can automatically identify the goods and emit a green light beam to determine where they should be placed.
The epidemic posed a major challenge to Amazon. At that time, Americans also had to stay at home, and shopping basically moved online. In the severe environment, Amazon has also recruited 400,000 new workers, which is also inseparable from the participation of automation. Amazon uses a program to screen out unqualified job applicants from a large number of resumes. It also uses video applications to train new employees and relies on software to guide new employees to complete simple and repetitive tasks. Automation has become a ready-to-use workforce for engineers, and adjustments can be made immediately when conditions change.
Amazon’s high-tech assembly line keeps the company’s efficiency higher, but it makes employees miserable. Last year, workers in a warehouse in Alabama tried to form a union but failed. The workers complained that they were forced to work with unreasonable goals, and these goals were recommended by the algorithm.
A worker who has worked in Chicago for a long time said that the most common guidance his manager gave him was to
“speed up, speed up, speed up.”
For the same task, the completion time is two seconds faster than the average time, and it can be praised by the manager. If it is two seconds slower, it will be warned. The employee requested anonymity because he was not authorized to accept media interviews.
In September, the California legislature passed a bill that would give warehouse workers the right to fight against the so-called speed quota. Supporters of the bill say that oversaturated work rhythms force employees to circumvent safety rules and give up rest time. The governor of the state has not yet signed the bill.
An employee who joined the Nevada warehouse in Amazon’s large-scale recruitment boom recalled that he was really worried that the shelves would collapse when his colleagues kept stacking goods on the shelves. The employee, who did not want to disclose his identity, said:
“Just to make the performance look good, even safety is not guaranteed, and I feel a lot of pressure at all times.”
Washington state regulators imposed a fine on Amazon’s DuPont warehouse earlier this year. The regulator said there was a direct link between the fast pace of work at the facility and injuries. Amazon is appealing the fine, saying it is revising its productivity tracking tool to better solve the problems faced by employees.
Many workers receive work tasks through computers or mobile phones all day long. They say this environment makes people feel isolated from their colleagues. Many front-line workers said in interviews that they often cannot name their managers, and it is not easy to establish relationships with colleagues. This situation is even worse under the epidemic. Such criticism made Xiao Bo feel frustrated, and he suggested that new employees still try to get to know everyone and actively socialize.
Former General Motors executive Alicia Boler Davis, who is in charge of Amazon’s fulfilment centre fleet, believes that higher levels of automation will give managers time to make more contact with workers. She said: “I hope that managers can spend their time understanding employees and ensuring safety. What I am thinking about is how can we reduce the burden? Our goal is to simplify the work process, reduce the difficulty of work, and reduce the need to simplify the process so that Decision-making becomes easier, but it is also necessary to reduce the physical burden and allow employees to handle some other tasks.”
Amazon’s reputation as a “devil factory” dissatisfied the people who built the business for the company. A former Amazon employee who was responsible for the development of warehouse technology said:
“Amazon is not without humanity, nor is it intended to abuse employees. It’s just that when one is devoted to solving mathematical problems, there is a real reminder. Don’t forget that work arrangements need to consider humanistic care.”
It is not only frontline workers who are trapped in the algorithm but Amazon managers are also constrained everywhere. A warehouse manager in Oregon, USA, said that he actually wanted to get to know the hundreds of employees who reported to him, but time pressure kept him busy at work all day, unable to discuss career goals with employees. He said that he has almost no option to “rest”. The manager left Amazon last year because he signed a non-disclosure agreement and requested anonymity. He said that after a 12-hour shift, he occasionally took a nap in the car to avoid getting stuck on the way home.
But Amazon said these are not common phenomena. In July, it announced that it would make employee benefits one of its guiding principles, and promised to be the best and safest employer in the world. Amazon recently stated that it will spend 1.2 billion U.S. dollars to provide courses and job training for front-line workers, as well as reimbursement of university tuition for some employees.
Xiao Bo believes that new employees will be better trained to understand their promotion channels on Amazon. He said:
“Not everyone can find their own development path in the fulfilment centre. We have to ‘take the initiative’ to the employees, such as telling him that there is a new device here. If you are interested, I can teach you how to use it. “
What kind of future will industry benchmarks lead?
The Amazon assembly line has attracted other companies in the logistics industry to follow suit.
“Amazon is the industry benchmark, far ahead,”
said Glick, the former head of Amazon logistics, who is now the CEO of warehousing startup Flexe.
Amazon’s number one rival, Wal-Mart, is also introducing highly automated warehouses to existing stores. The US grocery chain giant Kroger Co. is trying to build a robotic warehouse for grocery delivery. Instacart Inc., the second-largest unicorn in the United States, has established its own grocery delivery business through an army of robots, and the company is building its own robotic warehouse.
Even a smaller company like Cargo Cove feels it can match Amazon’s automation efficiency. Cargo Cove is a 4-year-old warehousing startup with only 80 employees. The company is planning to introduce robots and software in the next few months to automatically route orders and monitor employee output. Robert McFaul, the founder of Cargo Cove, said:
“This is the same principle as Amazon, which is to establish a very simple standardized process.”
Back to the BFI4 warehouse, Xiao Bo was walking fast along the corridor, watching the workers untie a yellow suitcase of goods stuck on the conveyor belt. He clicked a few keys on an application and noted when the stuck area was cleaned up. This is a safety check he needs to conduct regularly in order to reduce the number of injured in the warehouse.
Amazon’s technical team has a long-term goal to build a fully automated, without human intervention, fulfilment centre. This desire will take several years. The current challenge is mainly to let the robot arm grasp objects of different sizes and textures. Amazon executives say that manpower is still essential for the foreseeable future.
Amazon engineers are currently studying how to move more products in each warehouse to further increase the speed of product delivery, which is undoubtedly good news for customers.
It’s just that the hearts of the employees will be cold again, and I don’t know when they will be able to fly out of the walls of machines and algorithms.
This article originally appeared here.
Keep visiting for more such awesome posts, internet tips, gadget reviews, and remember we cover,
“Everything under the Sun!”