Employee retention survey: what to ask people who have not left
This one goes to everybody who still works for you, under no name, and asks how likely they are to be here in a year and what would change that. The eleven questions are already built. Open them in the editor, or copy the printed bank and paste it anywhere.
- 11questions to send
- About 3 minto answer
- 12questions to copy
- Anonymousread as counts, not names
Eleven employee retention survey questions, and the twelfth you write yourself
If the resignation is already in, none of this reaches it and the exit survey questions are the ones that still work. Everything below assumes the person answering has not decided.
Are they going
Two items, and they are what make this a retention survey rather than a satisfaction one. Both carry a window: a share of people who might go at some point cannot be ranked or acted on.
- I expect to still be working here twelve months from now.This is the headline. Report the share who disagree or strongly disagree, for the whole population and for every group at or above your reporting floor.
- Which of these is closest to where you are right now? (not looking / not looking but would listen / have started looking)Expecting to leave and doing something about it are different states. Someone in the third option has already made the decision the first item is still measuring.
What would move it
Six drivers, one idea in each stem, on one 5-point agreement scale. They are deliberately the six an employer can change, worded to be repeated unchanged next time so two runs compare.
- The amount of work I am given fits the hours I am paid for.Workload and hours are asked as one item and not two, because the complaint people mean is the ratio between them.
- I could take a real problem to my own manager.Kept separate from the next item on purpose. The old version of this survey joined manager and senior leadership into one stem, and a 2 could have meant either.
- I have confidence in the decisions senior leaders make about this organization.The other half of that split. A team can be low here and high on their own manager, which is a different problem with a different owner.
- My pay here is competitive with what I could earn elsewhere.Pay measured against the outside market, which is the comparison someone makes before they go. Whether pay is fair against the person at the next desk is a different question needing demographics this survey does not collect, and benefits are a separate market again.
- There is a job here I could move into that I would want.Asked as a concrete next job rather than as growth in general, so the answer names something you could put on a plan.
- Work that goes well here is noticed by the people who decide about my job.Written about the system rather than about an event in a window, so it does not turn into a count of thank-yous.
In your words
One forced choice, then one open box. The choice is countable; the box is where a reason nobody listed arrives, which is why it follows immediately and why no item above it offers a dead-end "Other".
- Which one of these would most affect whether you are still here in a year? (pay / how much work there is / my own manager / where my job can go from here / hours and flexibility / senior leadership)One answer, not several, because the point is the ranking. Edit the six to match the drivers you asked about above and keep the two lists aligned.
- What is the one change that would make you more likely to still be here in a year?The only optional item. Read these in full, grouped by whichever driver the person picked, and resist counting them: ten people writing the same sentence is not ten percent of anything.
The cut you report by, and the one you write yourself
Two cuts is the most a survey this short can carry without becoming identifying. The first ships with the template because it means the same thing everywhere. The second cannot.
- How long have you worked here? (under a year / 1 to 2 / 3 to 5 / 6 to 10 / more than 10)Closed bands, not a date. This is the cut the template ships with, because a first-year answer and a tenth-year answer can produce the same average from opposite reasons.
- Which team do you work in? [replace the options with the groups you already report by]Question 12, and the only one not in the survey you can open. Write it as a closed list. A free-text team box cannot be counted, cannot be suppressed by a minimum-group rule until somebody hand-codes it, and is the field people type their own squad name into.
One scale, six drivers, one idea in each stem. The published evidence on scale length converges on five to seven points and expressly declines to name a default;[4] this template uses five, so the same item can be repeated next time and the two numbers compared.
The twelfth question is the one a template cannot write. Only you know which groups you already report by, and the shape matters more than the wording: a closed list can be counted and suppressed, a free-text box cannot. Two closed cuts is as far as an eleven-item survey should go before the answers stop being anonymous in practice.
This survey is deliberately narrow. It asks about one decision and the six drivers behind it, so the staff satisfaction survey is the better read when you want the whole job covered and nothing about leaving.
A retention number is a decision, not a mood. There is a window in question 1 and none anywhere in a morale instrument, which is the whole difference: the employee morale survey questions report how a team feels now; this one reports what they expect to do about it.
What an intent-to-leave answer is actually worth
One number justifies running this survey and forbids the thing most pages about it imply. It is worth knowing before you send anything.
The people who say they are going do go, at nearly five times the rate. Among 102,970 US public school teachers, 33 percent of those who said they intended to leave as soon as possible had left the profession a year later, against 7 percent of everyone else.[2]
And two thirds of them were still there. The same paper puts the correlation between saying it and doing it at 0.15 and calls the predictive validity of intent-to-leave tenuous.[2] Both halves are the finding: across a group of forty the difference is worth acting on, and against one person it is close to nothing. That is why this survey carries no name and is reported only in groups.
Most departures are a decision somebody made. Of all separations in the United States in 2025, 60.6 percent were quits, against 33.8 percent layoffs and discharges and 5.6 percent retirements, deaths and transfers.[1] Quits are the only component a survey of current employees can reach, and the only one that was reversible for a while.
The named version of this conversation is a different instrument. One person, their own manager, on the record, ending in something they both agreed: that is what stay interviews are for. This one is anonymous, goes to everybody at once, and is built to be counted.
The moment before people start weighing it up is worth its own read. When somebody has just changed role, manager or team, the employee experience questionnaire asks whether that change has actually landed.
Sending it, and who can see the answers
Two decisions, and neither of them is made for you by anything else on this search.
Decide the smallest group you will report, and put the number in the invitation. There is no statutory minimum for a workplace survey: US Department of Education guidance on aggregate reporting treats a cell of three as the absolute floor and notes that larger minimums such as 5 or 10 cut the risk further.[3] The threshold is a choice about risk rather than a constant, which is why it belongs in the invitation. This template uses five.
A link only reaches people who open a link. That under-reaches the group this survey most wants to hear from, and no wording fixes it. Send it from a person rather than a system, leave it open long enough to catch a week of leave, and say in the invitation what will happen to the answers.
All eleven questions are already built. Open them in the editor, change any wording before you send, or start from an empty page in the free form builder instead.
Reading what comes back, in three passes
In this order, because the second pass is where the wrong conclusion gets drawn and the third is what stops it.
| The pass | What you compute | What it does not tell you |
|---|---|---|
| The headline | The share who disagree or strongly disagree with question 1, for everyone and then for each group at or above your reporting floor. | Who. Nothing here identifies a person, and at the level of one person the signal is weak.[2] |
| The gaps | The mean of each of the six drivers, for everyone, then the same six inside the groups where the headline is worst. | Cause. A low driver sitting beside a high leaving share is a place to look, not a finding. |
| The words | Nothing. Read every answer to question 10 in full, sorted by the driver the person picked in question 9. | A percentage. Open answers are for what the ratings could not have asked. |
Do not build a list of names. There is none to build it from, and the paper above is why the design is that way.[2] What these answers support is a decision about a team, a policy or a manager layer.
When the next-job item is the lowest driver, that is a heading and not a plan. The training needs assessment asks which capability is missing in which job family.
When workload and hours are the two lowest, this survey has reached its limit. A wellness survey for employees measures the working conditions themselves rather than how people feel about them.
The people you wanted and did not get never appear in these counts. The candidate experience survey asks them once, after they leave the process.
Then publish it. Pick two or three things, put a name and a date on each, and show them on a 30, 60 and 90 day plan people can see. Running this twice without doing that teaches everyone that answering changes nothing.
Questions people ask about employee retention surveys
What is an employee retention survey?
A short anonymous survey sent to everyone who still works for you, asking how likely they are to be here in a stated number of months and which parts of the job would change that answer. It is read as counts by group, never as a file on a person, and nobody answering it has resigned.
How many questions should it have?
Eleven here, answered in about three minutes, plus a twelfth you write yourself. The long lists on this subject are menus to pick from, and a menu is a fine thing to read and a poor thing to email to four hundred people.
Can this survey tell you who is about to quit?
No. Among 102,970 US public school teachers, 33 percent of those who said they intended to leave as soon as possible had left the profession a year later, against 7 percent of everyone else. Two thirds of them were still there, and the correlation between saying it and doing it was 0.15.[2] Across a team that is a strong signal; against one person it is close to nothing.
How small can a group be before you report it?
There is no statutory minimum for a workplace survey. US Department of Education guidance on aggregate reporting treats a cell of three as the floor and notes that 5 or 10 are used to cut the risk further.[3] This template uses five, and says so in the invitation.
How often should you run it?
Twice a year, plus one extra run about a month after anything structural: a reorganization, a change of manager, a pay round. Keep the two intent items and the six drivers word for word identical between runs, because a number you have reworded cannot be compared with the one before it.
Methods and sources
What this template is based on
The design follows one measured finding: intent to leave separates groups strongly and individuals weakly.[2] The base rates are the 2025 annual JOLTS estimates.[1] The reporting floor and the five-point scale are our choices, taken with the two sources that describe the tradeoffs.[3][4]
Question licensing
No licensed instrument governs retention questions, and none is reproduced here. Every one of the twelve is written for this page and is yours to copy, edit and translate.
Limits and disclosure
The teacher study is US public school teachers followed for one year, and the page says so wherever the figure appears; it is not a general employment statistic and it is not a rate for your organization. The JOLTS quits rate is monthly and is not an annual probability of anyone leaving. A link-based survey reaches people who open a link, which under-reaches exactly the group it most wants.
SuperSurvey makes the editor this template opens in. The twelve questions work with or without it.
References
- U.S. Bureau of Labor Statistics. Job Openings and Labor Turnover, January 2026, news release USDL-26-0439, 13 March 2026, carrying the 2025 annual estimates. bls.gov. Public domain. Note for link checkers: bls.gov returns 403 to a scripted fetcher and is live in a browser.
- Nguyen, T. D., Bettini, E., Redding, C., and Gilmour, A. F. (2022). Comparing Turnover Intentions and Actual Turnover in the Public Sector Workforce: Evidence from Public School Teachers. EdWorkingPaper 22-537, Annenberg Institute at Brown University. doi.org/10.26300/3aq0-pv52. Analytic sample 102,970 teacher-year observations.
- U.S. Department of Education, Privacy Technical Assistance Center. Frequently Asked Questions: Disclosure Avoidance. studentprivacy.ed.gov. Guidance about aggregate reporting of student data, quoted here for the reasoning rather than as an HR standard.
- Maitland, A. (2009). "How Many Scale Points Should I Include for Attitudinal Questions?" Survey Practice, 2(5). surveypractice.org
Ask the people who have not left yet
Eleven questions, about three minutes, no name attached to an answer. Open them, change what you need and send it today.
Use this template