A director at a large insurer told me her company had run the same engagement survey for nine years. She could recite the lowest-scoring item from memory, because it had been the lowest-scoring item every year she worked there. The score was not the part that stayed with me. It was that in nine years nobody had asked her what it would take to move the thing. They asked her to rate it, annually and reliably, with a reminder email from a shared HR mailbox.

The instrument has a problem before the follow-through does

Most of the complaint aimed at employee surveys is about what happens after the results come back, and that complaint is fair. Sitting underneath it is an earlier problem, and I find it the more uncomfortable of the two, because it suggests the results were never going to contain the thing leaders were looking for.

Researchers at Binghamton University's School of Management put roughly 200 working adults through questions about harmful leadership behavior, prompts along the lines of a manager who publicly belittles subordinates or has explosive outbursts. About one in ten recalled anything negative about their supervisor at all. Most reached instead for something positive about the person.1 The researchers' reading was that these ratings may be capturing whether someone likes their leader rather than what that leader actually did.1 The work was supported by a U.S. Army Research Institute grant and was selected for the Academy of Management's Best Papers list out of roughly three thousand submissions.1

A survey that asks people to characterize their manager will reliably tell you who is pleasant. It was never built to tell you who is doing damage.

Consider what that does to the leader who is warm in the hallway and corrosive in the room where decisions get made. He is agreeable most of the time, and agreeable is what people reach for when a survey asks them to describe him in general terms. The two meetings a quarter where he takes someone apart never surface, because nothing in the instrument asks about a particular behavior on a particular day. His scores come back clean. His attrition does not, though attrition is a lagging number that lands in a separate report, read by different people, two quarters later.

What does surface tends to sit there

Then there is the part most people already suspect. Perceptyx, a company that sells employee listening software and therefore has a stake in the conclusion, put questions to more than 750 senior HR decision-makers at organizations of a thousand people or more.2 Seventy-one percent of employees say their organization shares survey results with them. Fifty-nine percent say action plans get written afterward. Only fifty-one percent say an actual improvement came out of any of it.2 Twenty points separate being shown the results and watching anything change, and across all five years of this study, action planning and follow-through has never left the top five barriers.2 Among the most mature organizations in the sample, the distance between holding the data and doing something with it now outranks every other barrier they name.2

The cost lands somewhere specific. When employees report seeing no change at the local level after a listening event, they become two and a half times more likely to disagree that senior leaders' actions line up with the organization's stated values.2 The exercise did not simply fail to help. It handed people evidence about the leadership above them, and they filed it.

The question that fell off the list

The direction of travel makes both problems worse. Asked to name the top talent priority for 2026, 37 percent of HR leaders said employee performance and productivity, up from 23 percent in 2024. Psychological safety fell from 17 percent in 2025 to 11 percent.2 In a year of shrinking budgets the first number makes sense on its own terms. Put the two beside each other though, and what large organizations are trying to learn about has narrowed around output, which is not the category a belittling manager shows up in. Meanwhile the share running an isolated survey event out of HR, with little connection to the business, climbed to 22 percent, the highest since the study began.2

What actually moves

I have watched executive teams treat the annual survey as a low-risk obligation, a few hours of HR time in exchange for something to show the board. The research says the risk is real and that it points back at them. What works turns out to be smaller than most transformation plans. Choose one item and name two specific actions against it, then put the follow-up conversation on the calendar before the meeting breaks up. The same data favors that version over the ambitious one, showing engagement gains in 59 percent of cases where managers are equipped to build and track plans against 28 percent where no structured approach exists, and 69 percent of organizations seeing engagement rise where employees report visible behavior change.2

On the measurement problem, the fix is narrower still and costs nothing at all. Stop asking people to characterize their manager and start asking them what happened. What did you see in the last month that you would not want repeated. A general question collects an impression, and impressions are mostly a record of whether the person is liked. A specific question collects an event, and an event is the only thing anybody can act on.

So the question I would put to an executive team about to launch this year's survey is a plain one. Of everything you learned last time, what changed that an employee could point to without your help? If that takes you more than a moment to answer, the survey has stopped telling you about your organization. It is telling you how long people will keep filling it in.