Within any large organization, some employee teams always perform better than others—but the reasons are often murky. In 2010 then doctoral student Ben Waber embedded himself in the offices of Travelco, an online travel agency, to figure out what factors make a team productive. He noted that all the employees worked in the same headquarters, sat in the same meeting rooms, and enjoyed the same snacks and Foosball tables. But when he observed certain interactions among them, he says, “there was something really weird”: he saw members of the most effective teams tended to eat in groups of 12, while employees from lower-performing teams usually ate in groups of four.

When Waber thought about this, it made sense. “If you ate lunch with someone, you were much more likely to talk to them later in the week,” he says, and more communication is often linked to higher performance. That is because people who eat with more colleagues at lunch forge more in-person connections, which they can leverage if they get stuck or need help on a task. So from a company’s perspective, millions of dollars could hinge on differences in lunch-table size. Travelco immediately changed its cafeteria seating.

Many companies are eager for useful observations like these, and new technologies—such as ever smaller sensors, cameras, microphones and radio-frequency identification (RFID) chips for identification and tracking—are allowing them to constantly and automatically monitor their employees, in ever more detail, in search of similar breakthroughs.

Waber reported his insights in his dissertation and spun them into his own company, Humanyze, which he co-founded in 2011. Humanyze now works with clients around the world, from oil firms to banks to start-ups, analyzing their employee data in an effort to make them more productive. This process often involves workplace sensors that read employees’ badges to track where they go, how long they stay there and with whom they interact. Waber’s company is even experimenting with sound detectors that tell it if two people standing next to each other are speaking, although the system cannot reconstruct what anybody is actually saying. “I don’t care about what people are talking about,” Weber explains. “I just want to know if people are talking.”

Humanyze is hardly alone—employee-tracking technology is big business. Steelcase, a company that manufactures office furniture and helps design workplaces, offers employers an online  platform called Workplace Advisor that, it says, “uses strategically placed sensors and gateways to track precise, real-time space usage.” Along with calendar data, e-mails and other digital inputs, Workplace Advisor also employs heat sensors to determine how many people are inside a room. “If the space is designed to have six users in it, does it?” asks Brandon Buckingham, director of a Steelcase unit called Smart + Connected. If people are not using rooms the way they were intended, he says, it might be time for a company to rethink its office design. Beyond sophisticated equipment such as heat sensors, companies from McDonald’s to Amazon employ algorithms and passive tracking software to watch their employees, all with an eye to increasing efficiency and productivity.

It might seem intuitive that gathering such information will lead to insights that ultimately boost productivity. But researchers who study organizations are not necessarily convinced that this is so. Surveillance does seem to change employee behavior, but whether that improves the bottom line is less certain. “There’s almost no evidence on it,” says Lamar Pierce, a professor of organization and strategy at Washington University in St. Louis. And there are documented downsides to monitoring employees’ every move.

What researchers have found so far is mostly indirect evidence of a payoff, such as identifying worker characteristics that correspond to productivity. One study, for example, found that employees who use technology tend to be connected to more projects (although they also tend to do more multitasking, which can decrease efficiency). Another suggested that having more unpleasant interactions with supervisors causes burnout, which leads to decreased performance. (Waber himself is author of a handful of studies about tracking and productivity, as is Taemie Kim, another co-founder of Humanyze and its chief scientist. But many of their papers were not peer-reviewed, and they both have clear incentives to show that this kind of technology works.) What is missing, Pierce says, is research that links everything together, showing the cause-and-effect chain from sensor data to analysis to specific decisions that improve business outcomes.

A 2015 study published in Management Science by Pierce and his colleague Daniel Snow exemplifies the problem. They examined the effects of a new tracking system that made it harder for restaurant employees to give out free items or skim cash off the top and found that the system did, in fact, lead to less theft. But the researchers could not identify a mechanism by which it did so. In another field, Pierce says, that paper would not be robust enough to make it through peer review. Given the dearth of existing evidence in the literature, however, he and Snow were able to publish it. “The reason that paper was still published in a top journal was because even if we can’t say why, there’s still so little evidence about whether [tracking] works at all,” Pierce says.

Even in the example of the lunch seating at Travelco, the actual insight is hard to pinpoint. Is it that reclusive, less happy workers self-select into smaller groups of four? Or does the table size itself drive performance, suggesting that moving a low-achieving employee to a 12-seater would improve that person’s work? Waber admits that researchers do not know. “These are always hypotheses,” he says. “But how expensive is it to add some lunch tables and test it out?” Moreover eight years after Waber’s lunch-table observation, it is still unclear whether Travelco’s cafeteria change impacted worker productivity.

The lack of clear research data does not mean there is no power in tracking; it just means that its ability to increase profits remains mostly unproved. This is partly because proof would require an extensive chain of events: a company would have to track its employees, analyze the resulting data, use that analysis to change the workplace and then analyze the outcome in a way that removes all doubt that it was the tracking that made the difference. Before such a case study could even get started, researchers would need to persuade a company to agree to every step—which is harder than it sounds. “No company wants to be published as bad management, says Elizabeth Lyons, a management researcher at the University of California, San Diego. “And as researchers, we can’t sign NDAs [nondisclosure agreements] that say you can only publish this result if it’s positive.”

What researchers do know, however, is that the very act of observation can have an impact—and installing monitoring devices can immediately shift the way employees feel and behave. Lyons recently completed an as yet unpublished study looking at remote workers in northern Kenya who were tasked with reporting the types of vegetation they observed. She found that simply making these laborers aware that their work was being monitored altered their behavior. “We changed how visible the monitors’ presence was,” she says, and the workers’ performance improved.

Monitoring does not always have positive effects, however. In one study, Michel Anteby, a researcher at Boston University Questrom School of Business, observed Transportation Security Administration agents after surveillance cameras were installed all over their work areas. Initially these cameras were meant to protect the workers from theft accusations. But supervisors quickly began using the technology to catch agents for minor infringements such as dress-code violations or use of their personal cell phones during breaks. As a result, the workers began to engage in what Anteby calls “invisibility practices.” Essentially they changed their behavior to avoid the cameras as much as possible, “getting lost” on purpose and dawdling when transferring between terminals. These agents also reported feeling like the cameras reduced them from human beings to cogs in a machine. “They spoke about being interchangeable, being part of the woodwork,” Anteby says.

They are not alone. Amazon employees who are tracked as they “pick” items in warehouses say that being constantly timed is demoralizing at best and terrifying at worst. When the British newspaper Daily Telegraph secretly installed motion sensors under its employees’ desks in 2016, the staff pushed back, citing privacy concerns, and the paper removed them. The same technology (made by a company called OccupEye) was installed at the U.K. banking giant Barclays in 2017, inspiring similar reactions. And in some cases, employees have sued for invasion of privacy.

So the question still remains: Is the money companies might invest in all these passive tracking systems worth it? Right now, experts in the field of management science say the jury is out. “It’s nonconclusive that more monitoring is valuable,” Lyons says. And the conclusive downsides should not be underestimated.