The Transparency Trap

Ethan Bernstein is an assistant professor of leadership and organizational behavior at Harvard Business School. Follow him on Twitter: @ethanbernstein.

     Transparency” is a watchword in management these days, and it’s easy to understand why. After all, if people conduct their work in plain view, won’t they be more open and accountable? Won’t they flag and fix problems more easily, and share information and their good ideas more freely?

      That’s certainly what I expected to discover a few years ago, when I went in search of empirical evidence that transparency improves performance in organizations. But through rigorous field research and experiments, and observations by embedded researchers, I learned that it’s not that simple. My findings, which complement various studies on open workspaces (see “Balancing ‘We’ and ‘Me’” in this issue), suggest that more-transparent environments are not always better. Privacy is just as essential for performance.

     Here’s the paradox: For all that transparency does to drive out wasteful practices and promote collaboration and shared learning, too much of it can trigger distortions of fact and counterproductive inhibitions. Unrehearsed, experimental behaviors sometimes cease altogether. Wide-open workspaces and copious real-time data on how individuals spend their time can leave employees feeling exposed and vulnerable. Being observed changes their conduct. They start going to great lengths to keep what they’re doing under wraps, even if they have nothing bad to hide. If executives pick up on signs of covert activity, they instinctively start to monitor employee behavior even more intensely. And that just aggravates the problem.

      If all this seems vaguely Orwellian, so did some of the activities I saw in leading companies where intense visibility and tracking were making things worse, not better. For instance, at one of the world’s largest mobile phone factories, which is in China and is owned by a global contract manufacturer, the workers on one line were hiding processimprovements they had made—not just from managers but from their peers on other lines. Why? Because, as one experienced worker explained, “it’s most efficient to hide it now and discuss it later. Everyone is happy: They see what they expect to see, and we meet our targets.”

      This was not an isolated example. In my research, I found that individuals and groups routinely wasted significant resources in an effort to conceal beneficial activities, because they believed that bosses, peers, and external observers who might see them would have “no idea” how to “properly understand” them. Even when everyone involved had only the best of intentions, being observed distorted behavior instead of improving it.

      Some organizations, however, had found the sweet spot between privacy and transparency, getting the benefits of both. They used four types of boundaries to establish certain zones of privacy within open environments: They created boundaries around individual teams—zones of attention—to avoid exposing every little action to the scrutiny of a crowd. They drew boundaries between feedback and evaluation—delineating zones of judgment—to avoid politicking and efforts wasted on managing impressions. They set boundaries between decision rights and improvement rights—establishing zones of slack—to avoid driving out tinkering. And they put boundaries around carefully defined periods of experimentation—zones of time—to avoid both too frequent and too infrequent interruptions. Across several studies involving different industries, cultures, and types of work, the companies that had done all this were the ones that consistently got the most innovative, productive, and thoughtful work from their employees.

Type 1: Boundaries Around Teams

As social media platforms, wearable devices, and other tools for transparency become more advanced, our sense of being “onstage” is growing. And so, in keeping with the sociologist Erving Goffman’s insights about interpersonal behavior, we spend more time acting, trying to control others’ impressions and avoid embarrassment—particularly at work. We cater to our audience, doing what’s expected.

That was the case at the Chinese mobile phone factory, which had 14,000 workers. When I began studying that work environment, it seemed like the epitome of transparency: Each floor—roughly the size of a football field, with no walls or other divisions—held as many as 2,000 workers across shifts.

By embedding into the lines five Chinese-born Harvard undergraduate researchers—who worked, ate, and lived alongside the employees, who knew them only as coworkers—I quickly learned that the production teams hid a great deal from observers, despite the open environment. For example, to speed up assembly, workers scanned multiple bar codes into the system at once instead of scanning each one individually after applying it to a metal shield in a phone, as standard operating procedure required. And team members cross-trained on tasks during downtime—it looked like fooling around from the outside—so they could cover for one another when an operator fell behind. There was no ill intent—only a rational calculation about how to be most productive without having to waste time on explanations.

Such subterfuge is problematic for a host of reasons, though, ranging from increased risk of compliance-related defects to a lack of shared learning. To test some basic interventions that might address it, I set up a few field experiments. On one floor, where 32 production lines made similar mobile data cards, I randomly selected four lines on which to experiment, leaving 28 “controls” to work as they always had.

Because one of the experimental lines was very close to a control line, engineering put up a curtain between the two. When it was raised, one of the embedded students overheard a worker say, “Wouldn’t it be nice if they hung up curtains all around the line, so we could be completely closed off? We could be so much more productive if they did that.” Curious to see if that would be true, I asked engineering to fully encircle each experimental line with the equivalent of a hospital bed curtain. Over the next five months, to my surprise, the lines with curtains were 10% to 15% more productive than the rest, even when I controlled for other influences (such as the Hawthorne effect, whereby subjects improve simply in response to being studied).

By shielding employees from observation, the curtains supported local problem solving, experimentation, and focus. But within the curtains work became much more transparent. Partly for that reason, defects remained extremely low, even as throughput rose. And over time the camaraderie within boundaries made the workers more likely to share—as a group—their privately worked-out solutions with other lines.

Traditionally, people in organizations expect full transparency within teams but not necessarily beyond them. Team boundaries can allow for productive, selective opacities within starkly transparent environments—as was clear at Valve Software, a top PC game developer I studied with Francesca Gino and Bradley Staats. Valve’s 400-plus employees are allowed to allocate 100% of their time to projects they feel are valuable to customers. When they collaborate on new products or features, they form teams called cabals and move their desks (which are set on wheels) together into clusters. The office layout is so fluid, with some individuals rolling their desks to different cabals multiple times a week, that Valve even has an internal application to track desk location.

Valve’s cabals choose their own workspaces, creating privacy by positioning themselves at a distance from others. Though transparency is high within them, it’s moderate at best across the company because of the physical separation and Valve’s distaste for managerial oversight. (No one has the role of keeping tabs on the cabals or shuttling information back and forth.) This gives the cabals more freedom to investigate ideas.

When one employee started a cabal to explore how Valve could get into hardware, the team was initially tiny. Had it immediately tried to rally the support of the entire organization of software engineers, the hardware concept might have been dead on arrival—it’s hard to persuade lots of people at once to embrace anything new, even at Valve. But acquiring a few followers with whom to experiment and create prototypes was doable. Gradually, the hardware cabal accreted people and resources, gaining scale and momentum. To recruit more people to join it, early members eventually had to tell others what they were up to. In other words, they increased their transparency outside the group—but in their own way and when they were ready.

Is Valve providing an innovative, productive work environment? Its success suggests that it is. In its 18-year existence, Valve has produced a large share of top PC games. According to its founder, Valve has grown sales by more than 50% every year and brought in more revenue per employee than Apple or Microsoft. Its game platform consumes more bandwidth than most countries do. The cabals help the company compete in a market where creativity and rapid prototype and launch capabilities are critical.

Though Valve is an extreme case (and its success is a product of many factors), other firms are similarly fostering innovation and productivity by allowing privacy within team boundaries. For instance, Google doesn’t track when and where its engineers spend the 20% of their time that they devote to projects that interest them personally—but they feel transparently accountable to others within the self-organized teams in which the work gets done. And that protected 20% time has been credited with the incubation of more than half of Google’s current product portfolio, including Gmail, AdSense, Google Talk, Google News, Google Transit, Google Now, and the Google Transparency Report.

Team boundaries have a big impact on performance for service providers as well. In a recent Harvard Business School study, Melissa Valentine and Amy Edmondson show how such boundaries (in their case, counters delimiting small and very fluid groupings of nurses and physicians) improved teamwork and efficiency in a hospital’s emergency department. Transparency and accountability among people working within the boundaries increased. As a result, average patient time in the department fell by more than 40%, with no decrease in quality. Remarkably, the department sustained that improvement for over a year (the length of the study) even though its daily patient volume rose by more than 25%.

Although tools for observation (see the sidebar “Tracking Every Move”) and collaboration have become more powerful, making it easier for individuals to do much of their work without formal teams to support them, teams are actually proliferating rather than dying off. Longitudinal surveys show that today nearly all Fortune 1000 firms have formal team structures, compared with fewer than 20% in 1980. Though a number of factors are driving that trend, my research suggests it has something to do with the value of boundaries. Workers today can tackle problems in cooperation with large networks—and even crowds—of people, but as teams scholar Richard Hackman demonstrated, they frequently do it better on clearly bounded teams. Boundaries create a focus on “us” and “our work together,” liberated from external noise, whether it’s unproductive interference or chaotic workflow. No matter what the work is, some observers will increase productivity, but others will undermine it. Whether boundaries are spatial or psychological, they can limit observation to a zone of people. It happens with curtains, cabals, counters—even nominal team boundaries mitigate the pressures of being onstage by keeping the audience small.

Type 2: Boundaries Between Feedback and Evaluation

Organizations are incorporating more and more real-time data—all those electronic bread crumbs we leave behind as we do our work—into performance assessments. In response, employees waste a lot of valuable energy managing impressions. But tools that separate data-informed feedback from the evaluation process help lower people’s defenses and put the focus squarely on productivity and problem solving, where you want it.

In general, any information that goes into a formal performance review tends to put people on edge. Nevertheless, most employees are keenly interested in improving their skills. Just look at the popularity of Rypple, a social media platform created to allow members of organizations to give and gather anonymous feedback. (Salesforce.com purchased Rypple within three years of its launch for $60 million. It’s now called Work.com.) “You simply had to ask, ‘How am I doing at X?’” explains Rypple cofounder Daniel Debow, “and the answers were purely for you.” Because only the recipients had access to their feedback, fear of repercussions was removed from the equation. Further, Debow notes, those giving the feedback submitted honest, useful appraisals—with assurance of privacy, they didn’t have to worry that candid criticism might damage colleagues’ reputations.

Another way of allowing employees to learn from their day-to-day actions without having every little mistake exposed to management is to deliver feedback within a protective bubble. A large U.S. trucking company did this when it installed a DriveCam at the top of each tractor cab’s windshield to improve driver safety and performance. The small video camera points both outside and at the driver, gathering and wirelessly transmitting data that analysts can use to flag risky behaviors and prevent accidents. A green light tells the driver that all is well. But during a “G-force event” (any erratic driving incident that causes gravitational force, such as excessive speeding, slamming of brakes, or sudden swerving), the light blinks red and green. If the force is strong enough, the light turns red and the camera stores footage from eight seconds before and four seconds after the event. (On average, each vehicle’s DriveCam stores about five minutes’ worth of video a month.) The DriveCam also records key metrics, like the truck’s speed and location.

A small group of coaches who oversee fleet safety review any events deemed preventable. Only in a situation involving damage or a willful breaking of the law—for instance, failing to use a seat belt or texting while driving—would the coaches share footage with management. And the supervisors who evaluate the truckers aren’t privy to the coaching.

When the DriveCams were installed, drivers initially dreaded “being watched by Big Brother.” Some got distracted when the red light came on, which made safety worse. But drivers have since warmed to the cams, because they now trust that management won’t use the videos to evaluate or reprimand them. As one coach explains, the collaboration is helping drivers “turn bad habits into good habits” and improving their safety record. When coaches look at the footage with drivers, “it really does help,” another says. “It changes people’s perspectives.” Sometimes it’s just a simple realization: “Wow, you know, I was following a little too closely.”

Type 3: Boundaries Between Decision Rights and Improvement Rights

Managers work hard to clarifydecision rights, and for good reason. Spelling out who gets to make which calls helps organizations run more smoothly. It prevents duplicated effort, for example, and decision gridlock. But the empowerment of a select few can leave the other people in the organization feeling voiceless, especially if they aren’t explicitly invited to improve systems, processes, roles, and tasks. Employees may withhold their ideas or implement them on the sly. When organizations don’t grant improvement rights to those without decision rights, innovation by those who see solutions where others don’t—known as productive, or positive, deviance—is effectively squashed in favor of conformity and compliance.

It’s important to draw a line between the two kinds of rights, because the people exercising them have different needs. Holders of decision rights benefit from a transparent environment, where “every small fact becomes the subject of careful, scientific investigation,” as Frederick Winslow Taylor put it more than a century ago. But while holders of decision rights want perfect visibility, which requires transparency from everyone, that kind of visibility gets in the way of employees’ striving to make things better, because it curtails the experimentation necessary for improvements, as seen in the mobile phone factory and other settings.

In fact, a long stream of research tells us that in the presence of others, people do better on repetitive, practiced tasks—what psychologists call dominant responses—but worse on learning tasks that call for creative thinking. The visibility created by transparency conjures up self-consciousness and inhibitions. That’s why musicians perform in front of an audience but practice without one—they need privacy to noodle and make discoveries. So, the right level of transparency—and thus oversight—depends on the activity and the observer. While musicians may practice in front of a teacher, that teacher is an invited coach, not a consumer of their work. Technology is making close scrutiny by large audiences of consumers possible to a degree that Taylor could never have imagined, and clear decision rights amplify its effects. If you’re under the spotlight in front of such an audience, the last thing you want to do is to make unpracticed improvements while being held to a performance standard. All that transparency can create yearning for a closed door with a sign that says, “I’m in rehearsal!”

Organizations that understand all this are giving employees a reprieve from total transparency in order to make “slack” (excess resources) more productive rather than more scarce. Take Flextronics, a company that Willy Shih, Nina Bilimoria Angelo, and I have studied. By setting up a “moonshine shop,” Flextronics has turned its factory floor in Guadalajara into a veritable Legoland for workers. The shop gives employees a place to develop tools and fixtures for their lines in periods of downtime—creative work that imparts a sense of ownership. (Manufacturing companies often facilitate improvement rights in this way.) Made of simple pipes, connectors, and recycled materials, the designs produced in the shop can cost a tenth of what it takes to produce the more complex, specially sourced fixtures provided by vendors. The quality makes IKEA look high-end, but the designs do the job efficiently, safely, and effectively. More important, the shop encourages continual innovation by the operators, creating efficiencies that would otherwise remain in the imagination of workers.

Manufacturers aren’t the only organizations that have made slack more productive by protecting improvement rights. Saravanan Kesavan, Bradley Staats, and I saw this happen at the U.S. retailer Belk when it upgraded a mostly manual labor-scheduling system for its 24,000 employees and 300-plus department stores. Belk could have followed the lead of large retailers that have automated nearly all the scheduling task, increasing the efficiency of labor with complex algorithms based on minute-by-minute sales figures, real-time weather predictions, activity-based time studies, and other data. But Belk wanted to give its store managers and schedulers the flexibility to account for staffing variations and other local factors, since retail labor is a key driver of customer experience—and, therefore, sales. So its managers chose the simplest form of the technology and allowed local store managers and schedulers to exercise judgment, revising the schedules proposed by the system without having to seek corporate-level approval.

In the early days they revised more than 70% of the scheduling. Now that rate is below 50%—a more efficient, productive range. And while at least one of Belk’s competitors recently suffered well-publicized challenges in getting a return on its new fully automated scheduling system, Belk’s pilot stores showed a 2% lift in gross profit by the end of 2013, several months after implementing the version that allowed for overrides.

Which employees should be given improvement rights in order to create productive zones of slack? That depends on the organization and its leadership. In a lean environment everyone may be responsible for improvement. But other companies might treat it as an opportunity, not a mandate, perhaps vesting improvement rights in an R&D unit, a heavyweight team of senior managers, or frontline workers. Or an organization might outsource improvements to suppliers, contractors, or consultants. In any case, the assignment of improvement rights both reflects and influences strategy, so leaders must protect them by puttingskunkworks activities inside zones of privacy.

Type 4: Boundaries Around Time

Another way to strike the right balance between transparency and privacy is to experiment within limited blocks of time. With this approach, executives give employees more freedom for a specified period, so people can prepare for—and make the most of—their window of privacy.

This type of boundary complements the other three. A company might set up temporary teams for idea gestation, provide a onetime stream of developmental feedback (such as a 360-degree evaluation) that won’t make its way into performance reviews, or assign improvement rights to a certain group for a quarter. Some biotech and consulting firms have borrowed the concept of the sabbatical from education and offered employees periods of relatively opaque slack. Googlers often use their 20% time on Fridays.

Tony Lo, the CEO of Giant Bicycles, granted CFO Bonnie Tu time-limited decision-making authority when he asked her to develop a business model to better meet female customers’ needs. Lo saw Tu as a perfect leader for this initiative: Her seniority, reputation, and financial acumen gave her the freedom to break the usual rules. Lo—who was used to checking on important projects monthly, weekly, or even daily—left Tu alone for six months, the amount of time he thought it should take to develop and launch the idea. Tu and her team delivered brilliantly: They created a store in Taipei for women only—which reached profitability faster than any other Giant store. It has spawned a number of innovative products and serves as a model for similar stores around the globe.

In the same spirit, several major retailers have supplemented corporate planograms with “flexograms,” enabling individual store managers to change how and where they display products in response to customer behavior. While some retailers, like H&M, have made flexograms a standard practice, most limit them to times of year when local customization and ingenuity are likely to maximize sales. They frequently allow experimentation during the December holiday season, for example. CVS put its sun care displays on wheels so that store managers could easily reposition them to take advantage of fast-changing weather and buying trends at specified times during key periods in the summer.Sir Alex Ferguson, the former manager of the Manchester United football club, who is widely believed to be one of the greatest coaches in history, had an interesting take on transparency and its effects on performance. Though he championed the use of vests fitted with GPS sensors, which allowed analysis just 20 minutes after a training session, he said he would “never criticize a player during a training session. That’s where they try the irreverent things that will, and won’t, work during a match.”

It’s an important point: Irreverence increases our willingness to test how we do things and to deviate from the norm. But total transparency heightens the risk that our irreverence will come back to haunt us—and thus has a chilling effect on experimentation. Advanced sensing and tracking technologies make behavior highly visible in real time. How all that information should be used—by individuals, their teams, their supervisors—is a management question, not a technology question. Organizational cultures that foster psychological safety, trust, balanced power dynamics, and collaboration can help. But it’s also critical for leaders to mitigate transparency with zones of privacy, enabling just the right amount of deviance to foster innovation and productivity.