The System Works Because People Work Around It

Whenever a manager approached the production line, the workers became less productive.

They did not stop working. They started working the way they had been told to.

This was one of the more surprising findings in Ethan Bernstein’s study of a large mobile phone factory in China. Workers had developed “little tricks” that allowed them to perform their jobs faster and sometimes more safely. They shared these techniques quietly with one another. They also learned to abandon them whenever a manager, customer, or other outsider came near.

Under observation, they returned to the official procedures posted at each station. Those procedures often worked less well than the methods employees had developed through experience. The factory could see almost everything its workers did, but what it saw was often a performance staged for the people watching. The techniques that actually kept production moving disappeared whenever management approached.

Bernstein called this the transparency paradox. Managers believed that making work more visible would improve control and, in turn, help the organization learn. Instead, visibility encouraged workers to conceal what they knew.

He later tested what happened when several production lines were surrounded by a curtain. Managers could still enter, and production remained electronically monitored. The curtain simply protected workers from the continual possibility that someone passing by might inspect what they were doing. Performance improved. Workers became more willing to adjust how tasks were divided, help one another, and try new ways of solving problems.

The most immediate lesson is that privacy improved productivity. But the more interesting lesson may be that it improved organizational learning. The workers had already discovered better ways to perform parts of their jobs. Their ingenuity benefited the factory whenever managers were not nearby. Yet what they learned remained hidden within small groups. A useful technique could improve one production line without ever being evaluated or shared elsewhere.

The system worked because people worked around it. It also failed to learn because they had to hide how they did so.

Organizations often assume that making work more visible will help them improve. The transparency paradox suggests the opposite can happen. When employees expect every departure from the official process to be scrutinized, they become less willing to reveal the experimentation through which organizations actually learn.

We already recognize the value of protected experimentation in some settings. Google’s well-known “20 percent time” is the clearest example. The point was not that employees should spend one day each week doing whatever they wanted. It reflected a different assumption. Some employees might discover something valuable if they were given room to depart from their assigned work.

The factory suggests that this need for experimentation is not limited to software engineers or product designers. The workers in his study were not inventing a new mobile phone. They were assembling one. Yet their work still required judgment. The official procedures could not anticipate every bottleneck or capture everything workers learned from performing the same task thousands of times.

Experimentation was already happening. The only question was whether the organization would learn from it. Whether it did depended on what the organization assumed when employees departed from the official process. Were they making mistakes? Or were they discovering something the process itself had missed?

Douglas McGregor made a similar point decades ago in his distinction between Theory X and Theory Y. His larger argument was that management systems inevitably reflect assumptions about employees. Organizations that assume people cannot be trusted tend to design work very differently from organizations that assume people are capable of exercising judgment.

Organizations also tend to reserve this kind of trust for a relatively narrow category of workers. Autonomy is celebrated when it is given to software engineers, scientists, or product designers. In many other occupations, professional judgment is increasingly supplemented, and sometimes displaced, by procedures and metrics that make performance easier for outsiders to monitor.

K-12 education illustrates the tension. Standardized testing can reveal important differences in student achievement and identify schools that need additional support. But once test scores become the primary evidence of success, teachers and schools adapt to what the system rewards. In a study of high-stakes testing in the Chicago Public Schools, Brian Jacob found substantial gains on the high-stakes math and reading assessments without comparable gains on a separate low-stakes state exam. Schools also shifted attention away from subjects such as science and social studies that were not part of the accountability system.

Such outcomes do not necessarily mean teachers are lazy or trying to game the system. They show that people respond to what an organization makes consequential. The same pattern can emerge almost anywhere.

People closest to the work often discover things the formal system does not know. As research on organizational routines has shown, routines are continually recreated through the way people perform them. Employees identify unnecessary steps, solve recurring problems, and quietly adapt procedures to reality.

Technology is making this tension more salient because organizations can monitor work more closely than ever before. AI will make it easier to identify departures from established procedures. Whether it also helps organizations recognize useful patterns in those departures will depend less on the technology than on what leaders ask it to look for.

None of this means that every workaround is wise or should become official policy. Some shortcuts create safety risks, while others solve one problem by creating another somewhere else. Bernstein found that not every experiment behind the curtain improved performance. The curtain also did not remove accountability. Managers continued monitoring production, and workers remained responsible for quality and results.

What changed was that workers had enough room to try something before it was judged. Organizations often treat accountability and autonomy as opposites, but Bernstein’s study suggests that some autonomy may be necessary for learning. An organization cannot evaluate a better way of working unless employees first have enough freedom to discover one.

Most organizations do not need their own version of Google’s 20 percent time. They do need to ask whether too few people are trusted to improve the work they already do. Employees performing ordinary work constantly discover where official processes fall short. They solve problems that designers did not anticipate and develop knowledge that exists nowhere in a manual or dashboard. Organizations benefit from that knowledge whenever the work gets done, but they do not necessarily learn from it.

Greater visibility may help organizations identify errors and enforce standards. It may also make employees less willing to reveal the experiments and adaptations through which better methods emerge. The challenge is not simply deciding how much an organization can see. It is deciding whether employees are trusted enough to show what the organization still has left to learn.