Employee Engagement Survey Template

Before you draft a question, list every promise made after the last survey: what was promised, to which team, who owned it and what employees can see today. Mark each one delivered, in progress or dropped. Anything still "in planning" a year later counts as dropped.

Dropping a promise can be the right call; staying silent about it never is. Gallup, which sells its own engagement survey, lists inconsistent follow-through¹ among the common reasons engagement efforts fail. A team that was promised a change and saw nothing has little reason to answer again.

For a first survey, list this year's decisions that employees' views could shape, such as publishing pay ranges or capping workloads, and who makes each one. Those decisions set your questions.

Pick a minimum number of responses below which no group's result is shown, state it in the invitation, and never lower it after the survey closes. The UK Civil Service People Survey reports results only where at least 10 people have answered the same question². The template uses five, which gives more small teams a result but protects each person less.

Count responses, not headcount. A team of six clears a threshold of five on paper, then misses it if two people don't answer. Forecast every group before launch:

1. 1

   List every breakdown requested

   Results by team, level, location or tenure, plus any combinations, and who asked for each.

2. 2

   Forecast each group's responses

   Use each team's own response rate from last time, not the company average.

3. 3

   Sort the groups

   Mark each one reportable, borderline (headcount up to about 1.5 times the threshold) or never reportable.

4. 4

   Tell people now

   Give borderline managers the exact number they need, and tell anyone whose breakdown won't exist.

Combinations such as team by gender split a company into groups so small that people can be identified. In its 2024 public data, the US Federal Employee Viewpoint Survey blanked out demographic answers wherever fewer than five respondents shared the same work unit and demographics³. Check subtraction too: publish a department and all but one of its teams, and the missing team's result can be worked out.

Beside each draft question, write the decision its answer would change and the one person who owns it. "The leadership team" and "HR" don't count as owners. Then cut three kinds of question.

Values questions, such as "Which of our values do you identify with most?", usually fail the owner test too. Test values against real examples in a company values document instead.

Write each question as one plain statement about one thing, on one five-point agree scale. Pew Research Center warns that questions asking people to evaluate more than one concept⁴ are hard to answer and often produce results that are hard to interpret. "My manager gives me useful feedback and recognition" is two questions.

Pew also advises identical wording when you compare results with earlier surveys⁴, because slight changes in wording can affect answers. So leave last year's wording alone, and compare with an outside benchmark only if you copy its wording and scale exactly. Add two to four open questions and keep the survey to about 10 minutes.

Open the invitation with what happened to last year's promises, drops included, before you mention this year's survey. Then give the length, the closing date, the reporting threshold, who can see raw answers and the settled topics you aren't asking about.

Take level, location and tenure from HR records rather than asking. The survey tool then knows who answered, so call the survey confidential, not anonymous. Be honest that a threshold can't protect comments either. The UK Civil Service People Survey exempts comments from its 10-person rule and advises people not to include details that could identify them².

Send two reminders at most, each telling a team how many more responses it needs for its own result. Never show anyone a list of who hasn't answered, and never judge managers on their team's response rate.

Report each question as percent favorable, the share who agree or strongly agree. The federal survey calls this percent positive³. Gallup warns that percent favorable is a low-bar metric that inflates scores¹, so check the share who strongly agree as well.

Gallup estimates that managers account for at least 70% of the variance in engagement scores across business units⁶. A department average can hide one struggling team behind several doing fine, so show the team range beside the department range.

Survey numbers show where to look, not why. OPM, which runs the federal survey, suggests other tools such as focus groups⁷ to understand what lies behind the results.

Give managers their team's results two days before the company readout, so nobody hears their score in front of colleagues. Then agree one to three commitments per team, and a dozen at most company-wide. Each needs one owner, a date and something employees will see, such as a published on-call schedule.

Say which findings you won't act on this year, and why. When a commitment slips or is dropped, tell the affected team within two weeks, not at the next survey. OPM advises keeping employees aware of the actions being implemented and informed of progress⁷.

For the readout, the survey results report tool shows the response count beside every group's figure. Between surveys, stay interviews tell you why individual people stay. The template on this page sets out each step above, filled in for a made-up company.