Ask a room full of people leaders whether their employee survey is anonymous and every hand goes up. Ask them what would happen if someone in a team of 4 gave a 1 out of 5 and wrote a comment about their manager, and the room goes quiet.

This is the gap that we want to talk about. Not whether anonymous feedback is a good idea, because obviously it is, but what keeps it anonymous once real people in real teams start answering. Because the promise is easy to make and surprisingly easy to break, usually by accident, usually in the reporting rather than the survey itself.

Anonymous employee feedback means an employee can answer honestly without anyone being able to work out it was them. That last part is the bit that gets missed, it isn't enough to strip the name off a response. If the way you report the results lets someone reverse engineer who said what, the feedback wasn't anonymous, it just felt anonymous to the person answering. And once one person works that out and tells everyone else, you don't get honest answers again for a very long time.

Why people won't tell you the truth with their name on it

None of this is because your team is difficult, they are human beings and they are simply doing the maths.

If I'm 3 months into a new job and I think my manager is disorganised, saying so with my name attached is a bet. The upside is that maybe it gets better. The downside is that the person who decides my probation, my shifts and my next pay review now knows I said it. Most people, quite sensibly, don't take that bet. They write "all good" and get on with their day.

So what you get back from a named survey isn't a lie exactly. It's a carefully managed version of the truth, edited by someone who has to sit in a team meeting on Monday. You end up with a 4.2 out of 5 and no idea why three people left last quarter. Anonymity removes the reason to hide.

What breaks anonymity

Here's where most tools, and most people writing about this, stop short. Everyone agrees the survey should be anonymous. Very few people talk about the four ways it stops being anonymous after the responses come in.

Small groups. This is the obvious one and it's still the most common failure. A team of 4 gets a team-level report. Even with no names, a manager who knows their people can usually place a comment within about 30 seconds. Two of them are new, one has been off, so that leaves one.

Free text. People identify themselves without meaning to. "As the only person on the team who does the Saturday shift" is not an anonymous comment, no matter what the settings say.

What is a minimum group size, and why 5

A minimum group size is the smallest number of people a report will cover before the tool refuses to show it. Below that number, the scores are withheld entirely.

We use 5 at eeda. There's no law that says 5, and you'll see 3, 5 and 10 in different tools, but here's the reasoning. At 3 the group is small enough that a manager who knows their team can often place a comment. At 10 you protect people brilliantly and you also hide most of the useful detail, because plenty of real teams are 6 or 7 people and you've just made the tool useless for them. 5 is the point where it's genuinely hard to single someone out and the reporting is still worth reading.

A minimum group size on its own isn't enough. You also have to hide a group when the group you didn't select falls below the threshold, because otherwise it can be derived from the total.

How often should you ask?

Once a year is too slow. If the pool temperature drops in March and you ask in November, you've spent 8 months losing people over something you could have fixed in a week.

The other problem with annual surveys is that they're long, and long surveys get abandoned or rushed. A 40-question form at 4pm on a Friday produces a straight line of 3s from someone who wants their afternoon back.

Short and often beats long and annual. Five questions a week takes 60 seconds, keeps the habit alive, and gives you a trend line instead of a snapshot. It also means when something does go wrong, you see it in the data within a week after rather than at the next annual review.

What to do with what you hear

If people answer honestly, take the risk of saying something uncomfortable, and then nothing visibly changes, they stop. Not immediately, and not with a complaint. Response rates just quietly slide and you assume people are busy.

So close the loop out loud. "You told us the showers were broken, two are back in service this week." It doesn't have to be a big fix, and it doesn't have to be every item. It has to be visible and it has to be traceable back to something they said. One of those messages is worth more than a year of reminder emails asking people to please complete the survey.

Questions people ask us about this

Is anonymous employee feedback actually anonymous? It depends entirely on the reporting rules, not the survey. A response with the name stripped off is still identifiable if the tool will show you a breakdown covering 2 people. Ask any provider what their minimum group size is, and whether they also suppress a group when the complementary group falls below it.

What is a good minimum group size for employee surveys? 5 is a sensible default. Below 3 is not defensible in a small team. Above 10 and you lose most of the useful team-level detail. The number matters less than whether the tool also protects against working out a hidden group by subtraction.

Can employers see who wrote an anonymous comment? In a properly built platform, no. The response and the identity are stored separately. Ask your provider directly whether an administrator can export data that links a comment to a person, and get the answer in writing.

How do you stop people identifying themselves in free text? You can't entirely, and you shouldn't try too hard. What you can do is remind people at the point of writing that comments may be read by their leadership team, so they can decide how much detail to include.

How often should you run employee feedback surveys? Weekly, if the questions are short. Annual surveys tell you what happened, months after you could have done anything about it. A single question a week gives you a trend rather than a snapshot and takes seconds to answer.

The bit that matters

Anonymous feedback isn't a feature you switch on. It's a promise, and the promise is only as good as the least careful report your tool will draw.

If you take one thing from this, make it a question to ask your provider, or your own team if you've built it yourself: what's the smallest number of people you'll show me a score for, and what happens to the group I didn't select?

If they can answer that clearly, they've thought about it. If they can't, your people will work it out before you do.