The Amazon Way: Why the data-driven workplace is a nightmare

Jeff Bezos is either a liar or a fool. The Amazon CEO responded to the lengthy and damning New York Times story on his company’s brutal office culture of surveillance and exhaustion by saying he didn’t recognize the portrait of Amazon as “a soulless, dystopian workplace.” He went on, “I don’t think any company adopting the approach portrayed could survive, much less thrive, in today’s highly competitive tech hiring market.” I’d venture that the multi-billionaire, sitting atop this very market, knows he’s spewing face-saving bunk.

The management approach depicted in the Times piece is precisely how companies in the tech sector, and elsewhere, now function and thrive. The story of a workplace governed by data-driven worker surveillance, vicious, humanity-effacing competitiveness and intolerable hours is no fiction.

Bezos’ defensive response—disingenuous or not—misses the actual dystopian thread of the story. He condemned any possible lack of empathy from employees, and argued that his offices are sites of (at least occasional) “laughter” and “fun.” While the environment depicted certainly seemed to foster cruelty and selfish ambition, Bezos’ focus on certain, potentially un-empathetic managers and employees is misplaced. Amazon is not a good company infected by mean-spirited individuals. Bezos’ words sidestepped the fact that the Amazon Way is at fault.

The protestant frugality and celebration of endurance is part of it—no free snacks, no Google-style gyms, nor sleep pods. But the undergirding framework is the “continual performance improvement algorithm” that a former continual marketer said is run on the staff. Bezos’ management dream is a cybernetic nightmare: every task, and task-doer, measured as data. Workers, fiercely monitored and pitted against each other in the form of data points, work to maintain their place in the punishing ecosystem, and thus sustaining it as increasingly competitive. Empathy, meanwhile, is not valorized in the feedback loop. There’s nothing neutral about data; it serves the purpose of counting what we choose to enumerate. Amazon’s model is about efficiency and bulk, it doesn’t read for well-being. And those who say that efficient work is predicated on well-being are sadly not paying enough attention to late capitalism. Workers can burn out and be happily replaced.

Bezos is not the originator of this management model, but he may be its foremost apostle and extremist. The father of scientific management, as it is known, was 19th century mechanical engineer Frederick Winslow Taylor. Taylor worked to concretize performance monitoring and enhancement by dissecting task completion into measurable fragments that could be judged for speed and correctness. Production targets could be reformed and reformed accordingly. Crucially, Taylorism aimed to remove individual workers from deciding how their work is done. Managers, too, are deployed to oversee metric-driven, regimented piecework. When the head of finance for Amazon Web Services, Sean Boyle, told the Times, “Data creates a lot of clarity for decision-making,” Taylor’s ghost spoke through him. Then Boyle added, “data is very liberating,” and Michel Foucault’s ghost wept.

It is liberating, of course, for upper management to bypass the vagaries of worker autonomy. But the surveilled, monitored and performance-judged worker is everywhere in chains. At worst (or best of Taylorist management) they are the “docile bodies” Foucault warned would be the subjects of a mode of governmentality that monitors and manages every daily task. There’s no room for deviation, if keeping a job relies on performing every task according to data-driven targets.

A primary example of this, of course, is the well-documented cruelty of Amazon’s shipping warehouses, or “fulfillment centers” in its Orwellian corporate jargon. Distribution centers, paying around an allegedly decent $11 per hour, keep a draconian production line, enforcing long hours in harsh conditions. In a Pennsylvania warehouse, for example, during a heat wave, so many ambulances were called to the facility that the retailer paid Cetronia Ambulance Corps to have paramedics and ambulances stationed outside. Fifteen workers collapsed in the 102 degree heat.

The warehouses aren’t simply brutal sweatshops. They’re a Taylorist dream in which movement, shelving, and packaging of goods are broken down into subtasks, usually measured in seconds. Data then determines the best, and thus applied, workflow. Workers are subjected to a “three strikes you’re out system,” through a demerits model relating to task completion. One employee told the Times, in a separate report, that the Pennsylvania employees sent home during the heat wave were penalized with disciplinary demerits. This example alone gives lie to the techno-dreamer assumption that data-precise employee surveillance lends towards better, safer practices.

Karen Levy, a fellow at the Data and Society Research Institute, who focuses on technology, surveillance and regulation, has noted how data-driven monitoring systems promote worker active compliance beyond regulated task completion. She carried out a three-year study on how performance tracking was affecting the trucking industry. She observed that truckers would keep working during federally required breaks, racking up points on scoreboards, which were posted in break rooms and even mailed home to spouses. The irony is that the tracking devices in trucks were initially installed under the pretext of combatting worker fatigue. Levy noted that the game-like structure of a points system can diminish worker resistance to problematic conditions. “If you distract workers with the idea that they are playing the game, they don’t challenge the rules of the game,” she said.

Levy also noted that resistance in a surveilled working environment­—even small acts of taking time for oneself—is often foreclosed. Some truckers, unable to foil the monitoring systems, ended up smashing them with a hammer. An important paper on workplace surveillance, written by Levy’s colleagues at the Data and Society Research Institute, noted that “resistance to monitoring in general is complicated by the fact that some degree of monitoring is acceptable to most workers, and can effect or normalize compliance.”

The compliance of a lower waged labor subjected to problematic Taylorist management is different to the compliance of a well-paid office employee by a matter of scale—everyone is trying to keep their jobs, but the warehouse worker has fewer options and resources than the business school graduate. And Amazonians are certainly not the only privileged workers complicit in their own (and so each other’s) data-driven oppression. This is what I believe the late great poet George Oppen meant when, nearly 50 years ago, he wrote, “we are pressed, pressed on each other…We have chosen the meaning Of being numerous.”