Upward feedback is the survey most likely to go wrong. Writing twenty decent questions about managers is easy. The hard part is what you do with the answers.
Here’s the problem. Say a manager has four people on their team. You send the survey, three of them answer, and one of those three is unhappy. The manager knows who that is within a minute. The promise you made (“this is anonymous”) is broken, and the next survey gets polite, useless answers.
So this post gives you three things: a 12-question template, the four kinds of question that cause trouble, and what to do about teams too small to report on.
What this survey is for (and what it isn’t)
Use it to help managers get better. Don’t use it to decide pay, promotions or who gets managed out.
The reason is simple. If people think their answers can change someone’s pay, they soften them or skip them. They also remember it. You get one honest round at most.
Say this in the invitation, in plain words. If you’re running this around year-end reviews, keep the scores out of the review documents. Run the survey, give each manager their results, and keep it separate from the pay conversation.
The template: 12 questions in four groups
Use a five-point scale with every point labelled: Strongly disagree, Disagree, Neutral, Agree, Strongly agree. Every question describes something a manager does that an employee could actually see.
This set is deliberately different from our engagement survey questions. Those ask how people feel about the company. These ask about one person’s behaviour.
Clarity and direction
- My manager has told me what matters most in my job this quarter.
- When priorities change, my manager explains why.
- I know how my manager will judge whether my work is good.
Feedback and growth
- My manager has given me feedback on my work in the last month.
- The feedback I get from my manager is specific enough to act on.
- My manager has talked with me about my development in the last six months.
Support and workload
- When I tell my manager I’m overloaded, something changes.
- When I tell my manager something is blocking my work, they help clear it.
- My manager respects my time outside working hours.
Trust and safety
- I can disagree with my manager without worrying it will count against me.
- My manager admits it when they get something wrong.
- When I’ve brought my manager bad news, they’ve stayed constructive.
Then two open-text prompts:
- What is one thing your manager does that you’d like them to keep doing?
- What is one thing your manager could do differently that would make your work easier?
That’s 14 items in total. SurveyRock’s Free plan allows 10 questions per survey, and a rating grid counts as one question. Put the twelve statements in a single grid, as our manager feedback survey template does, and the whole survey is three questions, which fits on Free. Built as twelve separate questions, it needs Solo or above.
Notice what’s missing: “overall, how effective is your manager?” It feels like the headline question. It tells a manager nothing they can act on. A 3.4 on “effective” is just a number. A 3.4 on “when priorities change, my manager explains why” is a conversation.
The four questions that cause trouble
1. Questions that rate competence instead of behaviour
“My manager is good at their job.” “My manager is knowledgeable.” People answer these by how much they like the person. The result says nothing you can change.
Instead: ask about something observable. “My manager is a good communicator” becomes “When priorities change, my manager explains why.”
2. Questions where a low score points at one person
On a team of five, one person scoring “My manager gives me useful feedback” at 1 is a name. The question isn’t the problem. The size of the group reading the result is.
The fix is in how you report, and the next section covers it. Two related traps:
- Don’t add questions about tenure, role or location. Each one slices a team into smaller groups.
- Don’t offer the manager a view that shows results by any cut that leaves fewer than five people.
3. An open “anything else about your manager?” with no stated policy
Open comments are where anonymity usually dies. Nobody breaks it on purpose. Someone writes “when you moved me off the September project,” and everyone in the room knows who that was.
Instead: use specific prompts like the two above. Tell people before they answer who reads the comments, and whether managers see them word for word or in summary. Then do what you said.
4. Anything that will be read in the same room as a pay decision
Upward feedback that feeds a pay discussion gets softened. A manager who knows a low score affects their raise will also start asking who said it.
Instead: keep the survey and the pay process in separate meetings and separate documents, and say so in the invitation.
The small-team problem, and what to do about it
Reporting on a team of four breaks anonymity. One person’s answer is a quarter of the result. The manager can usually guess who. Even if they can’t, the team assumes they can, and that’s enough to change what people write.
That’s why anonymity needs a minimum reporting size: no results shown for a group smaller than some number. Five is the usual floor.
There’s a second, sneakier problem. Say a department of 20 shows an overall score, and three of its four teams show theirs. Subtract, and you have the fourth team’s number even though it was “hidden.” Good reporting hides the next-smallest group too.
What SurveyRock does here: the five-person floor, enforced
SurveyRock enforces that floor for you. A filtered view that would cover fewer than five people is refused, not just hidden. Turn on Collect responses anonymously (every plan) before you send. It can’t be turned off once responses exist, and if you switch it on late, responses already collected keep their links to the person.
On an anonymous survey, the response isn’t linked to the person’s contact record, IP address or device. Any filtered view of the results (by team, by answer, by date) is refused unless at least five completed responses match. You can raise that minimum to 7, 10, 15 or 20. Once responses exist, you can’t lower it.
For team-by-team reports, SurveyRock’s Team reports (Growth and up) tag each response with a team, never a person. A team view is held to that minimum. Comments and individual responses are held to a separate comment minimum, ten by default. When exactly one team under a department is below the minimum, the next-smallest team is withheld with it, so nobody can subtract their way to it. A department can still be reported with all its teams rolled in.
One setting to leave alone: Staff analysis access lets owners, admins and editors read below the minimums. It’s off by default, it can only be switched on before the first response arrives, and every read below a minimum is logged. If you promised anonymity, keep it off.
Where the limits are
We won’t call this a guarantee, because it isn’t one. Two filtered views that differ by a single person, both above the minimum, can still point at that person.
The minimum also only covers filtered views. Whole-survey results are not gated, and every open-text comment is readable by anyone with access to them. So don’t send one survey to one manager’s team of four and call it anonymous. Run one survey across many managers and report by team.
Your three real options for a team under five
- Roll it up. Combine small teams into a larger group, such as the department, and report on that. The manager of a team of three sees the department’s results, not a number that is basically their team.
- Replace it with a conversation. Have the manager ask their team two or three of the questions out loud, in one-on-ones, and make it clear it’s not anonymous. That’s honest, and it’s often more useful than a survey.
- Don’t run it. For a team of two or three, skipping is a legitimate choice.
Whichever you pick, say it before you send: “Results are reported only for groups of five or more. Smaller teams are combined with [group].” If people hear it after, they’ll assume the worst.
Scale and wording
Use five points, all labelled with words. Here’s why that beats a 1–10 for this survey.
With a 1–10 scale, people spend effort deciding between a 6 and a 7, and each person has their own idea of what a 7 means. A manager who sees an average of 7 thinks that’s fine. Their team may have meant “not great.” Labels fix it, because everyone reads “Disagree” the same way.
Report each question as the share of people who chose Agree or Strongly agree, not as an average. “62% agree” is easier to discuss than “3.8.” For more on when a Likert scale fits, see when to use a Likert scale and the Likert scale glossary entry.
Keep each question to one thing. “My manager is clear and supportive” is two questions, and nobody can answer it honestly if only one is true. Avoid wording that nudges the answer. A leading question like “How much do you value your manager’s helpful feedback?” assumes the answer.
What happens after: the manager conversation
The survey only works if the next step is handled well. Three rules:
- A manager sees their own report before anyone discusses it. Give them a day or two with it. Nobody should hear about their own results in a meeting.
- Someone talks them through it. That’s usually HR or the manager’s own manager. Tell them first what the report is and isn’t: development input, not a verdict.
- The manager shares back what they heard and names one thing they’ll change. That’s the habit that makes the survey work a second time. If people answer and nothing visible changes, they stop answering honestly.
One thing, not five. A manager who says “you said priority changes aren’t explained, so I’ll add two minutes to Monday’s meeting for that” has done more for next year’s response than any report.
For the whole engagement programme around it, see the complete guide to employee engagement surveys. For more on sharing results by team, see how to share employee survey results by team without breaking anonymity. If you’re planning the survey around an engagement cycle, the eight-week timeline shows where this fits.
Copy and adapt: the words around the questions
For the invitation to employees:
This survey asks about your manager’s day-to-day behaviour. It’s to help managers improve. It isn’t used for pay, promotion or performance ratings. Your answers are anonymous. We report results only for groups of five or more. Teams smaller than that are combined with [group]. Comments are read by [who] and shared with managers [word for word / in summary]. It takes about five minutes.
For the note to managers:
Your team was asked 12 questions about your behaviour, plus two comments. You’ll see your results before anyone discusses them, and [name] will go through them with you. This is for development, not evaluation. We’ll ask you to share back what you heard and name one thing you’ll change.
Add the twelve questions above, the two prompts, and the five-point scale. Adapt the wording to your own company. The behaviours matter more than the exact sentence.
Start from our manager feedback survey template, swap in the twelve statements and the two prompts above, and run it anonymously. With the statements in one grid it fits on the Free plan.