Why "rate your manager 1–5" produces nothing useful
Most anonymous review questions are written for the wrong audience. They are written for the spreadsheet that will receive the answers, not for the person sitting at their desk trying to give an honest one. "Rate your manager's communication on a scale of 1–5" produces a number that is easy to chart and impossible to act on. The chart will say 3.4. Three point four what? Of which behaviors? In which situations? You cannot manage what the question did not ask about.
The pattern repeats across the standard review-question canon. "Does your manager give you the support you need?" Yes or no, the answer carries no information. "How satisfied are you with leadership?" The same. The questions feel rigorous because they have a scale and a label, but the answers cannot tell you what to change on Monday morning, which is the only test that matters.
The fix is not to drop the structure. The fix is to write questions that the answerer can give a useful answer to, and that the leader can act on without guessing what the number meant.
The four question shapes that surface signal
Useful anonymous review questions tend to share four properties. Not every question has all four, but the best ones have at least three.
Specific. The question names a behavior, a moment, or a kind of work, instead of a general trait. "In a meeting in the last month, was there a moment you wished you had said something but did not?" surfaces concrete content. "Do you feel psychologically safe?" produces a yes-or-no that the answerer interprets differently than the asker meant.
Recent. The question is anchored in a window — the last week, the last sprint, the last quarter. Recent events are easier to recall accurately, less colored by retrospective justification, and more likely to be still actionable. "In the last sprint" is a far better frame than "in general."
Behavioral. The question asks about something that was said or done, not something that was felt. Feelings are real and worth knowing, but a feeling-based question ("how supported do you feel?") usually returns a feeling-based answer ("sometimes I feel unsupported") that is hard to translate into action. A behavioral question ("the last time you needed help and could not get it, what was happening?") returns a story you can do something with.
Forward-looking. The question gives the answerer permission to ask for something rather than only describe what is wrong. "What would you want to happen differently next quarter?" surfaces requests, not just complaints. Both are useful, but most review templates lean heavily on the second and skip the first, which leaves the leader with a list of problems and no list of remedies the team would actually be glad to receive.
Question library: 12 questions you can use verbatim
The twelve below are organized by what you are trying to learn. Pick three to five for any given review cycle. More than five and you start producing form fatigue, which lowers response rate and shortens the answers.
What is working that we should not change
1. In the last quarter, what is one thing your manager or team did that made your work meaningfully easier? Be specific about the moment.
2. Of the rituals or meetings you regularly attend, which one would you protect if leadership tried to remove it, and why?
3. What is something you have learned at this company in the last six months that you would not have learned somewhere else?
What is not working that you would change
4. In the last sprint, was there a moment you wished a decision had been made differently? What was the moment, and what would you have wanted to happen?
5. Where in your week do you most consistently lose time to something that feels avoidable?
6. Is there a piece of feedback you have wanted to give but have not, and what has been holding you back from giving it?
How safe the channels feel
7. In a meeting in the last month, was there a moment you wanted to say something but chose not to? What was the topic, and what would you have wanted to say?
8. When you have raised concerns in the past, has anything changed as a result? Tell us about a specific case.
9. If you had a serious concern about how something was being handled at a team or company level, who would you tell first, and why that person?
What you are asking for next
10. What is one thing you would want to learn or get better at in the next six months that the company could help with?
11. If you could change one specific thing about how decisions are communicated to the team, what would it be?
12. What is one experiment you would like the team to try in the next quarter that you do not think will be tried unless you ask for it?
The wording is intentional. The questions are long because the framing is doing work. Shorter questions are tempting but tend to produce shorter, less useful answers.
Pre-commit to acting on what you hear
The single largest determinant of whether an anonymous review cycle produces honest answers is what people remember happened the last time they answered honestly. If the previous cycle's feedback disappeared into a slide deck and was never mentioned again, the response rate this cycle will reflect that. If the previous cycle led to two visible changes that were attributed to the feedback, the response rate will reflect that too.
The pre-commit is the single most useful thing a leader can do before sending the review questions. In the same message that includes the link to the form (or the prompt from the bot), name three commitments: (1) what cadence you will respond on — usually a single follow-up message within two weeks summarizing what you heard, (2) what you will and will not do with the data — for example, that you will share aggregate themes with the team but no individual responses, and (3) at least one specific thing you are willing to change if the feedback supports it. The third commitment is the one that signals seriousness. "If a majority of responses ask for shorter all-hands meetings, we will move them to thirty minutes for the next quarter" is a real promise. Your team will remember whether you kept it.
One footnote on the technology side: HushAsk's anonymous channel hashes the sender's user ID with a 256-bit SHA-256 digest at send time, so even if you wanted to identify a respondent, you architecturally could not. On Slack Enterprise Grid plans, Slack's audit log records bot interactions, so for those teams the anonymity guarantee from HushAsk applies to the data HushAsk stores but not to Slack's own logs — worth raising with your IT lead before launch. For Free, Pro, or Business+ teams the caveat does not apply.
Most teams that ask the right questions and pre-commit to acting on the answers see response rates between 70 and 90 percent in the first cycle, and durable rates above 60 percent in subsequent ones. The questions are doing some of that work. The pre-commit is doing the rest.