DARPA has a strategy for maintaining its innovative culture that may seem controversial: Roughly a quarter of its workforce leaves every year.

The Defense Advanced Research Project Agency released a report this month that says most of the organization’s nonadministrative employees vacate their jobs every four to five years. Many even view the expiration date on their security clearances, which are reissued every two years, as a “personal expiration date,” said DARPA spokesman Rick Weiss said.

To be clear, the agency is touting this practice. Advocates say it helps create a sense of urgency among current employees, while constantly bringing in new talent with fresh ideas. As a result, the agency is not only able to constantly imagine the next big innovation, but recruit those with the technical know-how to make those innovations a reality.

Weiss said that the time’s-up-you’re-out rule is more tradition than policy. While program managers and directors aren’t exactly canned after a period of time, there is social and institutional pressure for them to move on after about four years. “If you want to stay longer than that, it starts to get more difficult,” he said.

This of course begs the question, who would accept a job with a fixed end date? Well, DARPA isn’t exactly a conventional workplace. Program managers often come from academia or industry and find the prospect of a stint in one of the government’s best-known innovation hubs appealing, Weiss said. They also typically oversee external research projects that last only a few years, the culmination of which provides a natural exit ramp. In situations in which a project may last longer, Weiss said DARPA has processes in place to smooth the transition.

“They don’t always want to leave the work they’re doing [outside the agency] to take a permanent position in the government, but they relish an opportunity to do a few years at a place like DARPA, where you have the resources and long leash to do something extraordinary,” Weiss said.

What apparently works for DARPA may not work in every office. Turnover does not come without significant disruption to an organization, said Peter Cappelli, a management professor at University of Pennsylvania’s Wharton School. It can mean an expensive and time-consuming process of bringing new employees into the fold, while also encouraging existing employees to always be thinking about their next career move.

“It seems to me that they’re making sense of high turnover rather than this driving innovation,” Cappelli said. High employee turnover “just doesn’t square with what we know about innovation.”

DARPA certainly knows about innovation. The agency has had a hand in most of the major inventions we know and use today, from the Internet to flat-screen monitors to GPS technology. Founded in 1958 under former president Dwight D. Eisenhower, DARPA was given the mission to ensure that the technological prowess of the U.S. did not fall behind its rivals. Today, DARPA pours millions of dollars into public- and private-sector research projects each year.

The “Innovation at DARPA” report asserts that long-time employees can become complacent and unimaginative, letting the agency’s past failures and an outdated understanding of technology shape their ideas for what innovations are possible. By contrast, new employees may have a deeper knowledge of cutting-edge technology and feel unencumbered by the agency’s history.

“The longer you’re in one place, the more tendency you have to become risk-averse. You start to refine what you’re doing as opposed to throwing out what you’re doing and starting fresh,” Defense Sciences Office Director Stefanie Tompkins said in the report.

Cappelli at Wharton maintains that longtime employees can still generate and carry out innovative ideas when they work in collaborative environments and have enough job security to take calculated risks. To that end, the DARPA report also emphasizes the agency’s risk-taking and failure-tolerant culture, as well as the high degree of autonomy its employees enjoy.

“People become complacent typically because of the way they’re managed,” Cappelli said. “I don’t think that moving, per se, is going to shake things up. It has more to do with the way people operate.”

Josh Bersin, a principal at Bersin by Deloitte, a research and advisory practice focused on human resources, said whether the DARPA model works for an organization may ultimately depend on how it defines innovation. If you’re merely looking for a fresh trove of ideas, then the constant inflow of new people can provide just that. Turning those ideas into revenue, however, may require more consistency.

“There’s a whole discipline to managing innovation through continual cycles of customer feedback, and when people keep leaving they don’t get that customer feedback,” Bersin said. “I don’t think most profit-making companies would be very innovative if everyone left every two or three years.”