Fourteen drivers, 42 rated items and 42 open questions, each one shown with the anchor pair it was written for. Copy a driver, take 25 to 35 items into your survey tool, and keep the outcome wording frozen so next year compares to this one.
Get the 84 questionsSr. Content Strategist
Covers HR technology, employee engagement and survey methodology at CultureMonkey, translating people science research into guidance HR teams can use.
People Science Team
Research team analyzing engagement data across 15+ industries globally.
First published Jun 2023
Last updated · 13 min read · Fact-checked
The short answer
What is a questionnaire for an employee engagement survey?
It is the fixed set of items an engagement survey asks: rated statements grouped by driver, each with a scale and anchor labels, plus a few open questions. A working questionnaire runs 25 to 35 rated items across the drivers you can act on, holds a small block of outcome items at frozen wording so cycles stay comparable, and states its anonymity threshold before anybody answers.
An engagement survey measures commitment and willingness to give extra effort, then explains that score through the drivers behind it: clarity, recognition, growth, leadership, communication, enablement, workload, belonging, autonomy and purpose. It is not a satisfaction survey. Satisfied employees are comfortable; engaged employees are invested, and the two come apart often enough that measuring one tells you little about the other.
The questionnaire is the instrument: the items, their order, their scales and their anchors. The survey is everything around it, including who is invited, how it is communicated, the anonymity rules and what happens to the results. Two companies can send an identical questionnaire and learn different things, because the survey around it was run differently.
The reason to get the instrument right is that engagement is both low and still falling, so the margin for a noisy measurement is thin. Gallup’s 2026 State of the Global Workplace puts global engagement at its weakest reading in years, with managers declining fastest, and prices the productivity gap in trillions. A questionnaire that cannot tell a real drop from a rewording problem cannot help you respond to either.
Global engagement
20%
Share of employees worldwide engaged at work in 2025, down from the previous year.
Manager engagement
22%
Share of managers engaged in 2025, down nine points since 2022. Managers set the ceiling for their team’s scores, so ask about them directly.
Cost of disengagement
$10T
Estimated global productivity lost to disengagement in 2025, about 9% of global GDP.
United States
30%
Share of U.S. employees engaged in the first quarter of 2024, the lowest reading since 2013.
Source: Gallup, April 2024
If you are still deciding how the wider programme should run, the companion guide to the employee engagement questionnaire covers scoring and benchmarks, and engagement survey strategy covers cadence and ownership.
Fourteen drivers, six questions each: three rated items with their anchor labels and three open questions that explain the ratings. Copy a driver into your survey tool, or take the whole bank and cut it down to the 25 to 35 rated items that map to a decision somebody has agreed to make.
Selection rule: pick 25 to 35 rated items and two or three open questions. Keep the outcome items in Motivation and pride at frozen wording, and rotate the rest.
Role clarity is the first thing to test, because an unclear role contaminates every other answer. An employee who cannot say what success looks like will rate their manager, their workload and their growth against a target nobody agreed on.
If clarity scores below the company average, read the rest of that team's results as provisional and fix the role definition first.
Recognition is the driver most often measured badly, because a single item on whether someone feels valued cannot separate frequency from fairness. These three split it: whether recognition happens, whether the effort behind a result is seen, and whether it lands on the right people.
A high frequency score with a low fairness score is a distribution problem, not a volume problem. More praise will not fix it.
Growth questions predict voluntary exits earlier than satisfaction questions do. The item that matters most is whether an employee can name a realistic next step, not whether training exists.
Pair this driver with your regretted-attrition data. Teams that score low here usually lose people six to nine months later.
Leadership items measure whether decisions are understood, not whether they are popular. Keep them about senior leadership as a group, because an item that points at one named person stops being anonymous.
Never ask employees to rate an individual leader by name on a company-wide survey. It identifies the respondent as much as the leader.
Communication is worth measuring in two directions, because a company can brief people well and still give them nowhere to answer. These items separate the quality of what arrives from whether it arrives in time to be useful.
Cross-team communication almost always scores lower than manager communication. Compare each against its own history, not against each other.
Feedback questions work best when they ask about usefulness and frequency separately from comfort. Someone can receive plenty of feedback and still not know where they stand.
Read this driver beside Leadership and trust. Low scores on both usually mean a psychological safety problem rather than a process gap.
Belonging items are the ones most sensitive to anonymity. Ask them only where your reporting threshold is large enough that no answer can be traced back to a single person on a small team.
Hold belonging results to a minimum group size of five responses, and report anything smaller only at the level above.
Enablement is the cheapest driver to act on, because the answers name specific tools and steps. It is also the one that most often explains a low engagement score in an otherwise healthy team.
The open answers here are your fastest visible win. Fixing two named blockers within a month does more for the next response rate than any reminder email.
Wellbeing items must stay about work conditions. Questions about health, finances or family life belong to a clinician, not to an engagement survey, and asking them costs trust for the whole questionnaire.
Say the word normal in the workload item. Without it, whoever answers during a launch week rates the launch rather than the job.
Collaboration questions are most useful when they cover the seams. Work rarely fails inside a team; it fails in the handover between two of them.
Break this driver out by team and by the teams each one depends on. A single company average hides every seam worth fixing.
Autonomy is about the match between the decisions a role requires and the authority it carries. Ask about that gap rather than about empowerment as a feeling.
Autonomy scores drop sharply one or two levels below a new leader. Cut this driver by reporting line before you conclude anything about seniority.
These are outcome items: they measure engagement itself rather than a driver of it. Freeze their wording, because they are the ones your trend line and your eNPS rest on.
eNPS is % Promoters (9-10) − % Detractors (0-6), and it runs −100 to +100. Keep it on its own axis and never average it into a 5-point driver score.
This driver measures whether the survey itself is believed. If employees have told you something twice and nothing happened, these three items are where that shows up, and where your next response rate is decided.
Track the acts-on-feedback item cycle over cycle. It is the single best early warning that participation is about to fall.
Purpose items connect the daily task to the company outcome. They tend to score well in small companies and decay with headcount, which makes them a useful check on how far your strategy actually travels.
If purpose scores high while clarity scores low, people believe in the mission and cannot see their part in it. That is a cascade problem, not a motivation problem.
Need a narrower set? There are dedicated banks for eNPS questions, open-ended pulse questions and employee satisfaction questions.
Measuring with a stretchy ruler gives you numbers and no answers. The scale on an engagement questionnaire decides whether results are comparable between teams and cycles, so pick one 5-point family, match the labels to the verb in each statement, and change it as rarely as you change the wording.
One scale family across the questionnaire means a 4 in sales means what a 4 in support means. Mixed point counts break that, and a company average built on them is not an average of anything.
Agreement labels fit a statement of belief. A clarity statement needs unclear to clear, a frequency statement needs never to always. Forcing everything onto agree and disagree makes some items read as double negatives.
Five points separate mild from strong feeling while staying quick to answer on a phone. Three flattens everything into a shrug; seven and up adds effort without adding signal for most engagement items.
Equal numbers of favorable and unfavorable options with a true midpoint keep the distribution honest. An uneven scale bends results toward whichever end has more room.
With a fixed 5-point scale, favorable is the top two options and that definition holds all year. Half the arguments about engagement reporting are really arguments about an undeclared favorable rule.
Keep the recommend item on 0 to 10 and report it as % Promoters (9-10) − % Detractors (0-6), on a −100 to +100 range. Never average it into a 5-point driver score.
For the mechanics of writing balanced statements and label sets, see the guide to Likert scale questions.
Neither alone. Rated items tell you what moved and let you compare; open questions tell you why, in words a manager can act on. The strongest questionnaire blends them, with rated items carrying the trend and two or three open questions carrying the explanation.
| Aspect | Open-ended | Multiple choice | Blended |
|---|---|---|---|
| Depth | Surfaces themes nobody thought to ask about. | Limited nuance, but every answer is countable. | Keeps the free-text detail beside a trend you can chart. |
| Effort to answer | Slow. Too many open items is the most common cause of drop-off. | Fast, which is what keeps a long questionnaire finishing. | Two or three open questions at the end of the rated blocks. |
| Comparability | Hard to compare between teams or cycles. | Clean benchmarks by team, location and quarter. | Comparable scores, with context for the ones that moved. |
| Analysis cost | Needs topic and sentiment grouping to be usable at scale. | Automatic. The scale does the coding for you. | Rated items report themselves; only the comments need reading. |
| Anonymity risk | Higher. Writing style and specifics can identify a person. | Low. A number reveals nothing on its own. | Hold comments to the same reporting threshold as scores. |
| Action | Names the specific blocker to fix. | Shows which driver to prioritize. | Gives a manager both the priority and the first step. |
A practical split: 25 to 35 rated items, two or three open questions, and one optional free comment at the end.
Framing decides whether a score means anything. Five rules cover most of it: one idea per item, neutral wording, plain language, a balanced scale, and a pilot before launch. Each one exists because breaking it produces data that looks fine and cannot be acted on.
| Do | Do not | Why it matters |
|---|---|---|
| Keep each item to one sentence in plain language | Stack clauses or qualifiers | An item read two ways produces two different scores that get averaged into one meaningless number. |
| Ask about one idea | Combine two ideas in one item | A middling score on valued and fairly paid cannot be traced to either half, so nobody can act on it. |
| Use neutral wording | Signal the answer you want | Leading items inflate scores and hide the problem you ran the survey to find. |
| Offer a balanced scale with a true midpoint | Offer more favorable options than unfavorable ones | An uneven scale bends the distribution, which makes benchmarking and trend lines unreliable. |
| Write in the first person and the present tense | Ask people to speak for their team or to predict | My manager gives me useful feedback is observable. Hypotheticals measure imagination, not experience. |
| Pilot before rollout | Launch untested | A pilot catches confusing items, broken devices and missing translations while fixing them is still free. |
Seven kinds of item do more damage than the insight they promise. Each one below shows the phrasing to cut, the replacement that measures the same thing, and what the bad version actually costs you.
How great is your manager at motivating you?
Ask instead: My manager helps me stay motivated at work.
A question that carries its own answer inflates the score and tells you nothing you did not already assume.
I feel valued and fairly paid.
Ask instead: Split it into I feel valued at work and I am paid fairly for my work.
When two ideas share one item, a middling score cannot be traced to either of them.
Do you like your work?
Ask instead: My work gives me a sense of accomplishment.
Broad items produce broad answers. Nobody can build an action plan on whether people like things.
Any item about finances, health, religion, politics or family circumstances.
Ask instead: Keep every item on work conditions, and route personal support through a benefits channel instead.
One intrusive item lowers trust in the whole questionnaire, including the parts employees were willing to answer.
Leadership communication efficacy meets my needs.
Ask instead: Senior leaders explain the reasons behind their decisions.
If two employees read an item differently, their scores are not comparable and the driver average is meaningless.
Which manager do you disagree with most?
Ask instead: Leadership decisions are explained to me, not just announced.
An item that only a handful of people could answer identifies them, and the next survey pays for it in participation.
Changing I am proud to work here to I feel proud of this company between cycles.
Ask instead: Keep outcome wording frozen and add new items as new rows.
A reworded item resets its own history. The movement you see afterwards is the edit, not the workforce.
A pilot is a quiet run-through with 10 to 20 employees, or 5 to 10 in a company under 200 people. It takes about a week and catches the faults that are expensive after launch: items read two ways, a version that breaks on a phone, a translation nobody checked.
Invite 10 to 20 employees across departments, seniority levels, shifts and locations, or 5 to 10 in a company under 200 people. A pilot drawn from one office tells you nothing about how the questionnaire reads on a factory floor.
Have testers flag any item they had to read twice, any pair that felt repetitive, and the point at which they wanted to stop. Those three signals catch most drop-off before launch.
Check whether testers picked the same option down a whole driver or skipped an item entirely. Both usually mean the item is vague or leading rather than that the tester was careless.
Open the questionnaire on a phone, on a shared shop-floor device, and in every language you plan to send. Confirm a link or QR route works for employees without a company email address.
Remove low-value items, split anything double-barreled, and match each anchor set to the verb in its statement. A shorter questionnaire that finishes beats a complete one that does not.
Freeze the wording of the outcome items so future cycles stay comparable, record the version, set your minimum reporting group size, and tell employees the date they will see results before you open the survey.
Say the threshold out loud: tell employees in the invitation that no group smaller than about five responses will be reported separately. Anonymity that is promised but not explained is not believed, and participation is where that shows.
For the announcement itself, there is a set of pre-launch communication samples you can adapt.
It is a fair challenge, and it is usually raised by an executive who has seen two surveys produce nothing. The answer is not that surveys lift engagement on their own. It is that a falling number you cannot break down by driver and team is a number you cannot respond to.
What the data says
Global employee engagement fell to 20% in 2025, and manager engagement fell to 22%, down nine points since 2022.
Gallup, State of the Global Workplace 2026. Gallup puts the associated productivity loss at roughly $10 trillion, about 9% of global GDP.
A decline that steep is an argument for better measurement, not less of it. A well-framed questionnaire tells you which driver is falling, in which teams, and under which managers, which is the difference between a budget line for manager coaching and a slide that says engagement is down. Managers falling fastest is itself a finding you only get from an instrument that asks about them directly.
What does not work is surveying without acting. The item to watch is Management acts on the feedback employees give: when it drops, the next cycle’s response rate drops with it, and no amount of reminder email recovers it.
CultureMonkey ships pre-built engagement questionnaires so HR teams start from a researched item set instead of a blank page, then adapt it. What follows is what the templates cover and where they stop, so you can judge the fit.
CultureMonkey templates cover the same drivers as the bank above, from leadership trust to enablement, so no driver is missed because nobody thought to ask.
Every item can be reworded, removed or replaced, and CultureMonkey keeps a version record so you know which wording produced which cycle’s score.
CultureMonkey enforces a minimum response count before any group is reported separately, including for open comments, which is what makes the anonymity promise real.
CultureMonkey sends a single questionnaire in every language you add, and reaches frontline sites by link or QR code for employees without a company email address.
CultureMonkey groups open answers into topics with sentiment, so the free-text block stays usable at a few thousand responses instead of being skimmed.
CultureMonkey puts an earlier cycle beside the latest one on a team heatmap, which is where a manager sees whether what they changed actually moved anything.
If you are still comparing platforms, the roundups of employee engagement survey tools and pulse survey tools set out the selection criteria.
A questionnaire for an employee engagement survey earns its place by being specific, neutral and stable. Specific so each item points at something a manager can change, neutral so the score is not decided by the wording, and stable so this year can be compared with last year. The 84 questions above give you the raw material; the framing rules, the list of items to cut and the pilot checklist are what turn a selection from it into an instrument.
Before you launch
CultureMonkey runs that questionnaire without the manual work: one survey in every language you add, link or QR access for frontline sites, anonymity held to a response threshold, open comments read into topics and sentiment, and an earlier cycle beside the latest one on a team heatmap so managers can act within the month.
Eight questions HR teams ask most often when they are putting an engagement questionnaire together.
Good questions state one idea, in the first person, in words every employee reads the same way, and they sit on a scale whose labels match the statement's verb. A workable questionnaire covers the drivers you can act on, such as role clarity, recognition, growth, leadership, enablement and workload, and keeps a small fixed set of outcome items like I am proud to tell people where I work so the score stays comparable between cycles.
Use 25 to 35 rated items plus two or three open questions for an annual engagement survey, and 5 to 10 items for a pulse. The 84 questions on this page are a bank to select from, not a questionnaire to send whole. Every item you add costs completion, so keep only the ones tied to a decision somebody has agreed to make with the result.
Yes, and you should reuse most of it. Year-over-year comparison is the main reason a questionnaire is worth running, and it survives only while the wording stays fixed. Keep the outcome items and the core driver items identical, then rotate a small block of items to match this year's priorities. Change the core wording and you reset your own trend line.
Yes. Anonymous responses are more candid and more complete, and participation is measurably higher when employees believe the promise. Anonymity is only real when it is enforced by a reporting threshold, so results for any group smaller than about five responses roll up to the level above instead of being shown. Say that threshold in the invitation, before people answer.
Keep a common core so the company can be compared, then add a short department block. Frontline teams need items on scheduling, equipment and shift handover; engineering teams need items on focus time and technical debt. Ask CultureMonkey or your survey tool to route those blocks by team so nobody answers items about tools they do not use.
The questionnaire is the instrument: the ordered set of items, their scales and their anchors. The survey is the whole exercise around it, including who is invited, how it is communicated, the anonymity rules, and what happens to the results. Two companies can send the same questionnaire and get different data because the survey around it was run differently.
Translate professionally rather than automatically, keep the anchor labels consistent across languages, and pilot each version with employees who work in that language. Watch for anchor sets that do not carry, since strongly agree has a different force in some cultures. Report each region against its own history rather than against a single global average.
Check three things. Rated items inside a driver should move together; the driver scores should relate to something real you already track, such as regretted attrition or absence; and the open comments should explain the ratings rather than contradict them. If a driver moves on its own while nothing else does, the items are probably measuring the wording rather than the workplace.