An employee engagement survey asks the people who work somewhere how they feel about working there, in a form that can be counted and compared to the last time you asked. That is the whole definition, and everything interesting is in the qualifications: who answers, whether they believe it is confidential, what is done with the result, and whether the same questions get asked again. This page covers what the instrument is for, what it is genuinely good at, and the honest answer to whether surveys are effective, which is that they are effective exactly to the extent that something visible follows them.
What it is for, stated plainly
Three things. To find out where the problems are, at a level specific enough to act on, which means by team rather than by company. To give people a route to say something they would not say to their manager's face. And to create a record you can compare against next time, so that a change can be shown to have worked or not. Anything a survey is used for beyond those three, and there are many things, is a secondary use and should not be allowed to distort the design.
Are they effective? Only with the second half
The evidence a company can actually see is its own: the response rate on the second round, which reflects whether the first one produced anything. Surveys in companies where visible change followed hold their response rates and produce increasingly candid comments. Surveys in companies where nothing followed decay, first in participation, then in honesty, and the score often drifts up as the people with something to say stop bothering. That is the failure mode to watch, because it looks like improvement.
Confidentiality is the thing that makes it work
An engagement survey collects opinions people believe carry a risk. If they think results can be traced back, they answer politely, and a politely answered survey is worse than no survey because it produces false comfort. That means a minimum reporting size applied without exception, a clear statement of what administrators can see, and never breaking either. One breach, once, ends the usefulness of the instrument in that company for years.
What HR gets out of it that nothing else provides
A comparison over time on a constant instrument. Exit interviews are biased by who leaves, one-to-ones are filtered by the relationship, and manager reports are filtered again. A survey with stable questions, run to the same population, is the only routine measure a people function has whose movement means something. That is the argument for keeping the questions stable even when somebody wants to redesign them, and for resisting the annual urge to freshen the wording.
Questions people ask about what is an employee engagement survey
How is engagement different from satisfaction?
Satisfaction asks whether people are content; engagement asks whether they are willing to put something in. The distinction matters because a comfortable, disengaged workplace scores well on satisfaction and produces nothing, which is precisely the case a satisfaction survey cannot see.
Should the survey be anonymous or confidential?
Most are confidential rather than anonymous: responses are identified to the system so results can be cut by team, but not reported in any form that identifies a person. Say which of the two you are running, in those words, because people asked to trust a promise deserve to know its exact shape.
What is a reasonable question count?
Short enough that people finish it and you can act on all of it. The planner on this site prices each extra question in employee minutes across the whole company, which is usually a more persuasive argument for cutting the list than any principle.