How Amazon's algorithms determine who should fire

How Amazon's algorithms determine who should fire

Amazon has 110 large warehouses, 75 executive centers, 45 sorting centers and 50 delivery stations in the USA. Robots cannot cope with all this good: the company says that it will take at least another 10 years for full automation. Therefore, all these facilities employ more than 125,000 full-time employees of the company. All of them need to be controlled so that not a single product lingers on the way to the buyer. How to do it? While the algorithms cannot cope with physical work, they turn out to be occupied with the most important: they fire people.

Amazon executive centers are the main driving force of the company. Huge warehouses of 100,000 m2 each, where employees receive, distribute, sort, track, pack and ship tens of thousands of parcels every day. Critics say that in these centers many people experience almost inhuman working conditions every day for fear of losing their jobs. A journalist from the UK worked at Amazon for a month, and reported that many of his colleagues were forced to write in bottles to fulfill the norm. If you do not have time to pack 240 (sometimes up to 400) boxes per hour or you do not pick up the goods quickly enough, you will immediately become a candidate for dismissal.

"Their logistic system is built on people working themselves to the bone and then replacing them with others," tells one of the Amazon warehouse managers, a former US Air Force veteran. - “And there is always someone behind you who is ready to get your job.”

Amazon has always found answers to critics. Including stating that her conditions and salary is better than that of other retailers, and information about the constant fear of layoffs is greatly exaggerated. But the other day in the hands of Vox was a document calling such statements into question. In fact, layoffs due to "underperformance" occur even more often than insiders thought. Just few people linger so long to gather statistics.

The full document (35 pages, mostly tables with productivity values ​​for each of the employees, which are filled out for privacy reasons) is available right here . This is a letter from a lawyer from Morgan Lewis, representing Amazon in one of the cases opened against the company due to “unfair dismissal”. He has a lawyer, Cary Crystal, that the rules for employees are issued by the algorithm from the outside and apply to the entire center, that is, for each employee in it, and cannot be modified inside, even if you didn’t like (or, on the contrary, liked) the manager. And the system issuing warnings and submitting an order for dismissal is “extremely objective.”

Muslims working in the Amazon warehouse in Minneapolis are protesting because they don’t have time to pray (they are given 2 breaks of 15 minutes for Ramadan)

The document reveals some of the secrets of the company. For example, it turns out that a warehouse in Baltimore (where Barbara Duvall worked with a complaint to Amazon) worked about 300 full-time employees (that is, interned and working on an ongoing basis, not under Prime Day or Christmas) worked for the year . This is from 2500 people working there. If such a pace of layoffs keeps constantly, it means that Amazon annually fires more than 10% of its employees.The company's staff in America is more than 125,000 employees, which means that thousands of people lose their jobs every month due to the insufficiently rapid movement of goods and boxes.

And these are just layoffs for reasons of non-compliance with the productivity standards that track algorithms. In fact, Amazon has much more turnover. Employees can also be fired by managers (for example, for antisocial behavior) or they can quit themselves without sustaining the pace of work.

Critics of the system (mostly Amazon workers themselves) say that this machine only sees the numbers, not the people behind them. Ctesy Mitchell, director of the Institute for Self-Sufficiency, says that executive center employees who regularly seek material or social support come to her center regularly. “I constantly hear from them that they feel completely interchangeable, like cogs in the system. They are treated like robots, because they are constantly monitored and evaluated by these automatic systems. Robots want the same robots to work for them. "

The main indicator of your effectiveness in the system, according to an Amazon lawyer’s document, is “time off task” (“time off task”, ToT). The lower it is, the less you rest, the better. Other performance indicators are secondary. Even if you are the soul of the company and did not take ten days off, it’s enough to miss two hours and you are fired. In 30 minutes you receive a written warning, 1 hour - the last warning. From here and workers who are afraid to take a break in the toilet. If you have been busy for too long, it can be a warning, and if you get several of these in one month - goodbye, work.

Amazon says that the pace of work is set for all employees of the warehouse, and reduced only if more than 75% of them can not cope with their norms. The worst 5% are sent to the training program, and if their results do not improve there, then dismissal also follows. If the employee does not agree with the decision, you can try to challenge it within the company.

Actually, they fired Barbara Duvall, who went to court: for several weeks she performed tasks at a rate of 72-93% of the norm. It was especially tight with the packing and scanning speed of bar codes of goods. The algorithm sent her to practice, where she was shown a faster way to scan. She did not follow him, the pace improved, but not much. At the same time, Barbara tried to fight for “workers' rights” (Amazon says it’s not to work for its rights). Soon she was fired, along with several hundred other "lazy". Barbara is sure that she was fired for her protest activity, but Amazon has all the numbers, her performance for each week.

A company representative in the letter responded in defense of Amazon:

Only about three hundred people in Baltimore were fired during this time due to lack of productivity. In general, the number of laid-off employees in the past two years has decreased both at this facility and in North America as a whole. Our systems automatically set standards based on parameters such as warehouse location and demand in this region. We do not demand the impossible from people.

Perhaps the most surprising thing about all this is that people are still surprised when software starts to deal with issues of dismissal. AI has for several years selects job seekers , works by the project manager , and by Walmart - and manages the store . Working time accounting systems have long been used to assess the activity of office workers.On-board computers in waggons in the United States check speed, gear changes and engine rpm to monitor the quality of work of truckers. Cash terminals in large stores monitor the pace of the cashier, and if he scans the goods too slowly, give him a signal.

All these systems are already helping employers decide who should be fired. Amazon simply moved to the next logical step, and stopped demanding permission from human managers, finally putting numbers at the forefront. Heartless? But effective!

Own employees for the company - as much as possible consumable goods, as far as the laws allow. For all questions in the priority are buyers. Only a few days ago, Bezos announced that he would spend $ 800 million so that the company's customers in the United States who signed up for Prime received the goods during the same day. This will be achieved including by opening new executive centers. This is despite the fact that on our warehouse in the States, goods from Amazon and so on average come in two or three days, this is the fastest delivery from stores. But Bezos believes that the speed of their work is the main competitive advantage of the company. And a minute delay (for example, because of the employee who decided to rest) can prevent him from realizing this advantage.

This strategy brings the company good results. April 25th Amazon published a quarterly report, according to which in January-March the Internet giant more than doubled its net profit. It reached a historic high of $ 3.561 billion (compared with $ 1.629 billion a year earlier). Revenue rose by 17% and reached $ 59.7 billion. On such news, Jeff Bezos's fortune exceeded $ 160 billion. This is more than the 40 million poorest Americans.

Source text: How Amazon's algorithms determine who should fire