The famous engagement instruments are proprietary, the enterprise platforms start with a sales call, and almost nobody talks about the two things that decide whether the results mean anything: whether employees trust the anonymity, and who didn't answer. Here is the whole method — usable questions included.
By the Wavefield Research team · Published Aug 12, 2026 · Updated Aug 13, 2026
An employee survey is unlike every other survey a company runs: the respondents know you, work for you, and wonder what happens to their answers. That changes the method. A customer who distrusts your survey ignores it; an employee who distrusts it answers carefully — and careful answers are worthless data.
So the design goal is not maximizing responses; it is making honesty feel safe. Everything that follows — question wording, anonymity rules, group-size thresholds, what leadership does after the results — serves that one goal. Get it right and engagement data is the earliest warning system a company has. Get it wrong and you have a well-formatted record of what employees think you want to hear.
I would recommend this company as a place to work.
The classic advocacy item — a fully labeled agreement scale, reported as the top-two-box share by team, and only for groups large enough to protect anonymity.
The best-known engagement questionnaires are trade secrets — the famous twelve-item instrument is licensed, not shared. The unglamorous truth is that well-worded items covering six areas measure the same territory. Here they are, free to use and adapt: agreement scales (Strongly disagree to Strongly agree) unless noted, two examples per area.
I would recommend this company as a place to work.
I rarely think about looking for a job somewhere else.
Advocacy, retention intent, and pride are the spine of every serious engagement instrument. Three to five agreement items, averaged into an index, tracked over waves. Keep the wording identical between waves — a changed word makes the trend meaningless.
How likely are you to recommend this company as a place to work? (0–10)
The employee version of NPS: promoters (9–10) minus detractors (0–6). Useful as a trended headline for executives, nearly useless alone — a small company's eNPS swings wildly on a handful of answers, so report it with its base size and read the index items for the real story.
My manager gives me feedback that helps me improve.
I can raise a problem with my manager without fear.
Engagement research keeps finding the same thing: the manager is the single biggest lever. These items localize problems — a strong company score can hide one struggling team, which is exactly what department-level reporting exists to catch.
I have the tools and information I need to do my job well.
Most days, the way we work helps me get things done rather than getting in the way.
Often the most actionable area in the whole survey: enablement problems have concrete fixes (a tool, a process, a decision), unlike culture problems. Low enablement scores with high engagement scores describe committed people being worn down — the most urgent pattern to catch.
I can see a path to grow my career here.
When I do good work, someone notices.
The strongest predictors of regretted attrition after the manager items. Recognition is also the cheapest thing on this list to fix, which makes it a reliable first win for teams new to acting on survey results.
What one change would most improve your experience of working here?
One open-end, at the end, always optional. Forcing it kills honesty; skipping it wastes the survey. Coded into themes, it becomes the action list — and it is where problems you did not think to ask about surface.
The wording rules are the same ones that govern every questionnaire — one idea per item, balanced scales, no leading language — covered with examples in survey question types.
Frequency is an architecture question, not a preference. The standard structure uses three layers, each with a different job:
| Layer | Cadence | Length | What it does |
|---|---|---|---|
| Annual census | Once a year | Full instrument (12–20 items + open end) | The benchmark wave — full picture, all teams, trend anchor |
| Pulse survey | Monthly or quarterly | 5–8 items, 2 minutes | Tracks whether actions from the census are working |
| Lifecycle surveys | Event-triggered | Short and specific | Onboarding (day 30/90) and exit — the moments of sharpest insight |
The binding constraint is action, not survey fatigue: employees will answer a monthly pulse survey indefinitely if things visibly change, and will abandon an annual survey after two years if nothing does. Delivery mechanics — email invitations with single-use links, SMS for frontline staff who never open email — are the same as any internal study; survey data collection methods covers the trade-offs.
Read the marketing pages for the big engagement platforms and you will find benchmarks, dashboards, and action plans — and almost nothing about anonymity mechanics or response bias, the two forces that quietly decide whether the numbers deserve trust:
Employees do not believe 'this survey is anonymous' — they believe arithmetic. The working rule: never report any group smaller than five respondents, and never combine filters (team × tenure × role) into a slice small enough to identify someone. Say the rule out loud in the invitation. Trust in the mechanism, not the promise, is what honesty depends on.
The disengaged are the least likely to answer an engagement survey, which biases every score upward. A team with a 45% response rate and a good average is not a good result — it is a team where most people did not bother. Report response rates next to scores, always, and treat a falling response rate as the earliest warning the survey produces.
Every wave that produces no visible change teaches employees the survey is theater, and response rates fall accordingly. The discipline: each wave ends with a small number of commitments, named owners, and a 'you said, we did' note before the next wave opens. If leadership will not commit to that, running the survey does more harm than not running it.
An eNPS that moved eight points on a 40-person company moved by noise. Engagement numbers deserve the same statistical discipline as any survey result: base sizes on every chart, margins of error on the headline numbers, and significance testing before anyone celebrates or panics over a trend.
On Wavefield, an engagement study runs like any other project: the agent drafts the questionnaire from a brief, invitations go out as single-use links by email or SMS, and reports pair every percentage with its unweighted base — with significance testing on effective bases for wave-over-wave comparisons via tracker waves. Group-size discipline in reporting (nothing under five, no identifying filter combinations) is a rule you enforce in how you read and share results — we'd rather tell you that plainly than pretend software can substitute for it. The analysis workflow is in how to analyze survey results.
A core of 12 to 20 agreement-scale items covering engagement itself (advocacy, pride, intent to stay), the manager relationship, enablement (tools and processes), and growth and recognition — plus eNPS as a trended headline and one optional open-ended question. The wording matters less than two rules: keep every item about one thing only, and keep wording identical between waves so trends mean something.
The standard architecture is one full annual census (the complete instrument, all staff) plus short pulses monthly or quarterly (5–8 items) to check whether the actions from the census are working. More frequency only helps if action keeps pace — surveying more often than you act erodes trust and response rates.
Know the difference: anonymous means responses cannot be linked to individuals at all; confidential means a system links them but reporting is aggregated. Most professional programs are confidential (it enables demographic breakdowns and wave-over-wave analysis) with strict reporting rules — no group under five respondents, no filter combinations that isolate individuals. Whichever you run, state it plainly in the invitation; vagueness reads as surveillance.
Healthy internal surveys reach 70–85%. Below 60%, treat the results as directional at best — and remember the bias runs one way, because disengaged people skip the survey. The score you see is almost always rosier than the truth. Response rate by team is itself a finding: the teams that do not answer are often the teams with the problem.
Related: customer feedback surveys · survey question types · sample size calculator · how to analyze survey results
Upload a brief and the agent drafts your engagement survey — clean items, single-use links, honest reporting. Free to build, from $99 to field, $99 per repeat wave.
Start a project