Do Employee Engagement Surveys Work?

Evidence grade: Misapplied. Good evidence exists and is routinely cited for a claim it does not support. This grade describes the published evidence. It is not a prediction about your workforce and not a result RetainScore has produced.

The evidence is real. The claim made from it is not.

There is a large, credible body of work linking employee engagement to business outcomes including turnover. The best known is the meta-analysis by Harter, Schmidt and Hayes, published in the Journal of Applied Psychology in 2002, which pooled 7,939 business units across 36 companies to examine "the relationship at the business-unit level" between satisfaction and engagement and outcomes including employee turnover. It found relationships large enough to matter in practice.

That finding is about engagement levels. It is not about surveys. Nothing in it shows that measuring engagement changes engagement, and nothing in it shows that running a survey reduces turnover. Those are three separate claims, and the vendor pitch routinely collapses them into one.

Two things the paper says that the pitch leaves out

First, the direction of causation is open. The authors write that "although findings of causal direction are unresolved, preliminary evidence in individual studies generally suggests that aggregate employee attitudes have positive relations" with the outcomes studied. Units with low engagement and high turnover may be badly managed, in which case both are symptoms and neither causes the other.

Second, the paper's own conclusion is conditional and it is about management practice, not measurement. Its stated implication is that "changes in management practices that increase employee satisfaction may increase business-unit outcomes, including profit." The active ingredient in that sentence is the change in management practice. The survey is how you found out where to make it.

Why the survey so often changes nothing

This is the version of the story most people searching for this actually have. Engagement scores went up, or the survey ran on schedule for three years, and turnover did not move.

The mechanism is not mysterious. A survey is an instrument. Instruments do not have effects. Whatever effect exists comes from what changes after the reading, and the chain from a score to a changed working condition has several places to break:

• Aggregation hides the problem. A company-level score of 74 can contain one team at 40. The research linking engagement to turnover is business-unit-level for a reason, and reporting that stops at the company or division loses the resolution that made the finding useful.
• The result goes to the wrong person. If the team's score is a judgment on the manager, and the manager owns the response, the incentive is to manage the score.
• Nothing is resourced. An action plan with no budget, no owner, and no decision rights is a document, and employees can tell the difference within one cycle. The second survey is answered by people who watched what happened to the first.
• The cadence outruns the change. Annual working conditions do not respond to quarterly pulses. Measuring more often is frequently a substitute for acting once.

How to read your own survey honestly

Treat the score as a pointer, not an outcome. The question worth asking is not did engagement go up but which specific working condition did we change, in which team, and did that team's turnover move. If nothing was changed, the survey has told you where to look and has not yet done anything else, which is exactly what the evidence says it can do.

Be most skeptical of the engagement number quoted as a retention result. Engagement and turnover being correlated across thousands of business units is a real finding. Your engagement score rising is not a retention outcome, and the literature the vendor is citing does not say that it is.

Work through the retention diagnostic to see which factors the research finds predictive, or read the companion grades on exit interviews and retention bonuses.