Employee engagement surveys became popular in the late ’90s after the Harvard Business Review published research showing a link between employee satisfaction at Sears stores and revenue growth. Today, U.S. firms spend about $720 million trying to improve employee engagement.

Unfortunately, it’s a waste of money, says a new paper in the Journal for Quality and Participation. “The dirty little secret of employee engagement surveys is that they’re largely junk science,” writes Robert Gerst, a partner at Calgary-based Converge Consulting.

The problem, says Gerst, is the attempt to use “statistical significance” to draw meaningful conclusions or patterns from broad survey data. “Just because you can detect a difference between two groups has nothing to do, at all, with whether that difference is in any way important,” he says. The surveys may even mislead managers and find problems where none exists.

Employee feedback is useful if done right. Gerst recommends qualitative research—speaking with actual employees—before developing targeted survey questions. As for surveys widely used now, Gerst writes: “Boiling employee engagement down to a single score means you don’t understand employee engagement.”