Compensation survey questions, and how to read the answers by group
A 12-question pay survey you can send this week. It asks whether pay feels fair, whether people understand how it was set, and whether it lands the same way twice. Answer it below, or open it in the editor.
- 12questions
- about 4 minto complete
- 3 scoresfair, explained, consistent
- 2 cutsband and time here
What these compensation survey questions measure
A pay survey and a pay analysis answer different questions. One asks people what they think. The other opens the payroll file. Run one without the other and you have half an answer.
Three comparisons, not one. One item that asks a single thing and calls it satisfaction cannot say which comparison went wrong. Three of the twelve split it, because a person measures their pay against three things: the work itself, the colleague doing similar work down the corridor, and what the same job pays somewhere else. The study behind the split links worker surveys to administrative pay records and treats the last two as separate variables, an internal reference wage inside the firm and an external one across firms.[2]
Feeling fair and being fair are two measurements. Validated against each other in about 500 German firms, people were as likely to call a fair wage unfair as to call an unfair wage fair.[1] That is the argument for running both, and the reason this template makes no claim about whether your pay is equal. It tells you whether it is explainable.
This is not the other kind of compensation survey. In its usual sense the phrase means an establishment-based collection of employer-reported wage and benefit costs, used to set market rates.[3] Nobody asks an employee anything. This one asks nobody else.
The 12 compensation survey questions
Every question, its answer format, and what to look at when the answer is low. Copy one block or the whole set.
What you are paid
Three comparisons, asked separately. Someone can be content with their own salary and certain that the person at the next desk is on more, and one item cannot hold both answers.
- My pay is fair for the work I do.The headline. It says how people feel, not why, which is what the next two are for.
- My pay is fair next to people doing similar work here.The comparison inside the building, treated as its own variable in the linked employer-employee work.[2]
- My pay is fair next to what the same job pays elsewhere.The comparison outside it. This is a belief about the market, and the belief is what drives the answer.
How it was decided
Understanding is the half you can fix without spending anything, so it is worth separating from the money.
- I understand how the pay range for my job was set.The band. A person can know their number and have no idea where it came from.
- I understand what would move my pay up.The route up. Low here beside a high score above is a progression question, not a pay question.
- When my pay was last set, someone explained the reasons to me.The last conversation. Everyone has had pay set at least once, so nobody is stranded on this item.
Whether it holds up
Consistency, then whether a person can raise the subject at all. Two different risks, so two items rather than one with an "or" in it.
- Pay decisions here are applied the same way from one person to the next.Read this one split by band. Consistency is the bucket where a gap between two groups IS the finding.
- I could ask my manager a direct question about my pay without it counting against me.One channel per item. For the HR answer as well, ask it as its own question instead of joining the two.
In their words, then the two cuts
One tick-all, one open answer, and the only two demographic questions this survey asks.
- Which of these do you think actually decides what people are paid here?Tick all, not pick one: asking for the single biggest factor forces a choice nobody actually holds.
- What would have to change for your pay to feel fair to you?The item that produces something you can act on. Read these before you average anything.
- Which pay band are you in?Put your own band names in before you send. A pile of answers on the last option is a transparency finding, not a data problem.
- How long have you worked here?Optional, like the band. People who skip both still count in the company total; they just appear in no cut.
Two swap-ins. Job family ("Which of these best describes your work?") and site ("Which location are you based at?") are the next cuts worth having, at the end beside the band, and only if the grid in the next section still holds up. If pay conversations here happen with HR rather than with managers, swap "my manager" for "HR" in question 8 and keep it to one channel.
You can create online survey free with all 12 questions already loaded, so the only editing left is your own band names.
Reading the answers by group
Three scores and two cuts. The cuts turn a company average into something a person can act on, and they also decide how much of it you are allowed to publish.
Read the three together, never one at a time. Average questions 1 to 3 as fairness, 4 to 6 as understanding, 7 and 8 as consistency. One number over twelve items hides the only thing worth knowing: which of the three is dragging.
| What the three scores look like | What that points at | Where to start |
|---|---|---|
| Fairness low, understanding fine | People follow the reasoning and still think the number is wrong. | The pay data, not the messaging. |
| Understanding low, fairness fine | People are content with the number and cannot explain it. | Publish the bands and the criteria. |
| Both low | A trust problem rather than either of the above. | Read the open answers before deciding anything. |
| Consistency low, the other two fine | Two people in the same position were treated differently. | Exceptions, off-cycle rises and new-hire offers. |
Your smallest box sets what you can publish. Cross the pay band with length of service and you are not making two lists but a grid: six bands by five tenure brackets is thirty boxes cut from one headcount, and a third cut makes it a hundred and fifty.
The floor, and where the number comes from. The US Office of Personnel Management runs the federal employee survey and publishes the rule it applies to itself: "No reports were produced when a subagency had fewer than 10 respondents."[4] For the file it releases publicly it goes further, treating a group as at risk of re-identification if there are "fewer than 5 respondents with the exact combination of demographics and work unit."[4] Ten before you report a group, five before a box belongs on a chart at all. Those are the only two numbers on this page.
So keep it to two cuts. Report band and length of service separately, and cross them only if every box clears ten. That is arithmetic rather than caution, and it is why this survey asks two demographic questions and not five.
What to promise about who sees the answers is set out under anonymous employee feedback. The wording of any cut you add is covered in how to write demographic answer options.
Before you ask about pay
Four decisions, then the note that goes in front of question one.
Pick the moment. Three work: before the pay round opens, straight after you publish a band, or once you have acted. The first tells you what to fix while the money can still move, the second whether the explanation landed, the third is the only one that closes a loop.
Say what it cannot do. Answering will not change one person's salary, and putting that in the invite is what stops the survey reading as a negotiation. Name the decision it feeds and the date that decision gets made.
Keep benefits out of it. Health cover, leave and pension sit on a different budget line, and this survey is deliberately about the paycheck; the package belongs in the benefits survey template. Merge the two and neither question gets answered.
Ask about pay, not about leaving. Whether someone would go over money belongs in an employee retention survey, where the rest of the reason gets asked too. Next to question 1 it changes what question 1 is asking.
One thing did change in June 2026: the EU pay transparency directive has had to be in national law since the seventh of that month, and it gives a worker the right to ask in writing for the average pay, broken down by sex, for people doing the same work as them.[5] Where that applies, knowing what your staff think before they ask is worth more than finding out after.
Pay is not the only thing a low answer can be about. The workplace satisfaction survey questions put pay next to the manager, the hours and the work, so you can see which one is really driving the answer. On the way out the same subject is a different instrument, and the exit survey questions ask it of somebody who has already decided.
A survey is not a pay audit and cannot stand in for one. Perception and payroll do not track each other[1], so whether people doing equal work are actually paid equally is answered from compensation data and not from twelve ratings.
What to put in front of people before question one
Before you start This survey asks what you think about your pay: whether it is fair, whether you understand how it was set, and whether pay decisions land the same way for everyone. What it can change: how we set and explain pay from [DATE]. What it cannot change: any individual salary, including yours. Answering it is not a request for a raise and will not be read as one. We ask for your pay band and how long you have been here so the answers can be read by group. Both are optional. We report a group only at ten answers or more, and we never publish a smaller one. [NAME] in [TEAM] owns this survey and can open the responses. Twelve questions, about four minutes. It closes on [DATE] and we publish what we decided by [DATE].
Compensation survey FAQ
What questions should a compensation survey ask?
Three about the money itself, three about how it was decided, two about whether it is applied the same way for everyone, one tick-all on what people believe sets pay, one open answer, and the cuts you will read the results by. Twelve in all.
How many compensation survey questions is too many?
Some question lists for this subject run to thirty or fifty items. None of those is a survey anyone can send; they are menus to pick from. Choose the twelve you will read, and keep the rest for the next round.
Should a compensation survey be anonymous?
It asks for no name and no email address. It does ask for a pay band and length of service, because without a cut you cannot tell a company-wide grumble from one team with a real problem. Whoever owns the survey can open every response as a row, so name that person in the invite and report groups rather than rows.
Should we ask about race or gender in a pay survey?
Not in this one. The categories come from a national standard, they differ by country, and they change, so a second copy on a page whose subject is not identity goes stale the moment the standard moves. The DEI survey template carries the current US and UK wording and the reporting rules with it.
Is this the same as a compensation survey from a firm like Mercer?
No. Those collect wage and benefit costs from employers to set market rates; the US federal one is described as an establishment-based survey of businesses and government agencies.[3] This asks your own staff what they think, which is a different question with a different answer.
Methods and sources
What this template is based on
Twelve questions written for this page. The three fairness items split one judgement into the comparisons a linked employer-employee study treats as separate variables: the wage itself, the internal reference wage inside the firm, and the external reference wage across firms.[2] Wording and order follow published practice on clarity, ordering and bias.[7]
Question licensing
No item comes from a proprietary or licensed instrument, and none needs to. The Gallup Q12, the Utrecht Work Engagement Scale and the Maslach Burnout Inventory are absent by wording and by structure.
Limits and disclosure
This measures what people believe about their pay, which is not the same measurement as their pay.[1] There are no quotas and no sample balancing, so report who answered rather than implying who was asked. Suppressing small groups on a published table is a discipline in its own right.[6]
SuperSurvey builds the editor this template opens in. Nothing on this page depends on using it.
References
- Diehl, C., Lang, J., Strauss, S. and Brueggemann, O. "Is My Wage Fair? Validating Fairness Perceptions among Women and Men." Socius 11 (2025). doi.org. Open working paper: University of Konstanz Cluster of Excellence "The Politics of Inequality" WP 33, econstor.eu
- Mohrenweiser, J. and Pfeifer, C. "Wage Structures, Fairness Perceptions, and Job Satisfaction: Evidence from Linked Employer-Employee Data." Journal of Happiness Studies 24(7), 2291-2308 (2023). Open version: IZA Discussion Paper 12821, docs.iza.org
- US Bureau of Labor Statistics. Handbook of Methods: National Compensation Measures. bls.gov
- US Office of Personnel Management. 2022 OPM Federal Employee Viewpoint Survey: Technical Report. opm.gov
- Directive (EU) 2023/970 of the European Parliament and of the Council of 10 May 2023 on pay transparency and enforcement mechanisms. eur-lex.europa.eu
- UK Office for National Statistics. "Statistical disclosure control." ons.gov.uk
- American Association for Public Opinion Research. "Best Practices for Survey Research." aapor.org
Send the compensation survey while the pay round is still open
Twelve questions, about four minutes, and three scores at the end that say which part of pay is the problem. Put your own band names in and send it.
Use this template