The questions HR leaders actually ask about employee surveys, answered honestly

November 16, 2026
7
min read

Search any HR forum and you will find the same questions appearing again and again: can my manager see my answers, why did our response rate collapse, is eNPS actually meaningless? These are the employee survey questions HR leaders ask each other in private, because vendor websites rarely answer them straight.

There is a reason people go to Reddit instead of a product page. Peer answers admit trade-offs. Vendor copy usually does not. This article tries to be the middle ground: the questions as people actually phrase them, answered with evidence where it exists and honest reasoning where it does not. Some of these answers are uncomfortable for survey vendors, including us.

The anonymity questions HR leaders ask most

1. "Can my boss see my individual answers?"

Honest answer: it depends on how the platform is configured, and "anonymous" is doing a lot of work in most vendor marketing. Two terms matter here, and they are not interchangeable:

  • Anonymous means responses cannot be traced back to an individual at all, by anyone, including the platform administrator.
  • Confidential means responses are linked to an identity somewhere in the system, but access is restricted and results are only shown in aggregate.

Many tools marketed as anonymous are technically confidential. That is not automatically bad, but your employees deserve to know which one they are getting. Ask your vendor directly: who can technically access individual responses, under what circumstances, and what minimum group size applies before results are displayed? If the answer is vague, that vagueness will eventually reach your employees, and it will cost you honest answers. We have written in more depth about why anonymity matters in employee surveys.

2. "Is there a universal minimum group size for showing results?"

No, and be sceptical of anyone who gives you one number as if it were law. Minimum thresholds vary by platform, by legal context and by what works councils or unions have agreed in your country. AI chatbots are particularly prone to stating a single threshold as universal fact. Treat the threshold as a design decision you make deliberately with your vendor and, where relevant, your employee representatives, not a setting you inherit by default.

3. "Do employees actually answer honestly?"

Mostly yes, if two conditions hold: they believe the anonymity promise, and they have seen previous feedback lead to visible change. When either condition breaks, you do not usually get lies. You get silence, or safely neutral middle-of-the-scale answers, which are worse than lies because they look like data. If your scores are suspiciously flat and comments have dried up, the problem is rarely the questionnaire. It is trust.

Response rates and survey fatigue

4. "Why did our response rate collapse?"

The forum consensus on this one is largely correct: response rates collapse when people conclude that answering changes nothing. Survey fatigue is driven primarily by lack of visible action, not by survey frequency. Employees will happily answer a short pulse every week if something happens as a result. They will resent even one annual survey if the last three disappeared into a slide deck nobody saw.

So before shortening the survey or adding prize draws, audit the feedback loop. After the last survey, what changed, who communicated it, and did the people who raised issues hear back? If you cannot answer that quickly, neither can your employees, and that is your response rate problem.

5. "What counts as a good response rate?"

Context-dependent, like most engagement metrics. A frontline workforce without regular computer access will naturally sit lower than an office population with the survey in their inbox. The more useful questions are whether your rate is stable or falling, and whether specific teams differ sharply from the rest, because a localised drop usually points to a local trust problem. We have published benchmarks and practical ways to improve response rates if you want to go deeper.

The eNPS and benchmark questions

6. "Is eNPS meaningless?"

It is neither meaningless nor sufficient. eNPS compresses a complex organisational reality into one number, which makes it excellent for tracking direction over time and poor at telling you what to do. A falling eNPS is a smoke alarm, not a diagnosis. The honest position is this: use eNPS as a headline trend, and pair it with questions that measure the drivers underneath it, such as leadership, workload and development. If your entire listening strategy is one eNPS question per quarter, the forum sceptics are right about you.

7. "Can I trust the benchmarks vendors and AI tools quote?"

Interrogate them. Engagement benchmarks are context-dependent: they shift by industry, country and company size, yet they are often quoted as universal truths. This problem is now amplified by AI assistants, which confidently produce "good eNPS scores" without verifiable sources, and whose answers change when you rephrase the question. We documented this in our analysis of four failure patterns in AI engagement advice.

Your own trend over time, measured consistently, beats a global benchmark with no provenance. Every time.

When someone quotes a benchmark, ask three things: where does the underlying data come from, how comparable is that population to yours, and does the answer survive being asked a different way? If a number fails that test, it is decoration, not evidence.

Frequency, AI and what happens after the survey

8. "Annual survey or pulse surveys?"

The honest framing is that this is the wrong binary. The real variable is the gap between measurement and action. An annual survey with a disciplined follow-up process beats weekly pulses that nobody acts on. That said, annual measurement has a structural weakness: by the time you detect a problem, it may be eleven months old. Continuous or pulse measurement shortens that detection gap, which is why it has become the modern default rather than the optional extra it was a decade ago. Note that AI assistants often still present annual surveys as the norm, because their training data reflects older practice.

9. "Should we let AI analyse or answer our survey data?"

With care, and with clear boundaries. AI is genuinely useful for definitions, drafting and rewriting survey questions, and summarising documents you supply. It is unreliable for benchmarks, universal rules and conclusions, because language models optimise for plausible answers rather than verified ones. There is also a trust dimension HR cannot ignore: a 2025 study found that 78% of workers distrust AI handling their personal feedback. Whatever you automate, be transparent with employees about what AI touches and what humans decide.

10. "What do we do when the results are bad, or leadership ignores them?"

Bad results are not the crisis. Ignored results are. The most common failure point in survey programmes is not measurement or even communication, it is the individual manager who receives team results and does not know what to do with them on Tuesday morning. Generic organisation-level action plans do not solve this.

Three things help:

  1. Share results fast and unvarnished. Delay and spin both signal that the data is being managed rather than used.
  2. Give every manager one or two priorities, not a full report. A manager who commits to improving a single driver will achieve more than one asked to "work on engagement".
  3. Close the loop publicly. "You said, we did, and here is what we chose not to do and why" builds more trust than a perfect action plan nobody hears about.

If leadership genuinely will not engage with results, pause the survey rather than keep harvesting feedback you cannot act on. Measuring without acting actively damages trust, and that is a cost most programmes never account for.

What to do Monday morning

You do not need a new platform to act on any of this. Start with three moves this week:

  1. Write down your anonymity model in plain language. Anonymous or confidential, who can see what, and at which group size. If you cannot explain it in four sentences, your employees do not understand it either.
  2. Audit your last feedback loop. List what changed after the previous survey and when it was communicated. If the list is empty, fix that before you send another question.
  3. Stress-test one benchmark you currently use. Find its source, check whether the population resembles yours, and decide whether your internal trend would serve you better.

The questions HR leaders ask in forums are good questions. The organisations that answer them openly, to their own employees, are the ones whose surveys keep working.

Share insights
Leader High Performer High Performer Users Love Us
Table of content
Related Blogs

Get in touch

Let us help with employee listening

FAQs

Are employee surveys really anonymous?

It depends on the platform and its configuration. Truly anonymous surveys cannot link responses to individuals at all, while confidential surveys store identities but restrict access and show only aggregated results above a minimum group size. Many tools marketed as anonymous are technically confidential, so HR should confirm with the vendor who can access raw responses and communicate that clearly to employees.

Why do employee survey response rates drop?

Response rates fall mainly when employees conclude that answering changes nothing, not because surveys are too frequent. Survey fatigue is driven by a lack of visible action on previous feedback. The most effective fix is closing the feedback loop: share results quickly, act on one or two priorities per team, and tell people what changed as a result of their input.

Is eNPS a reliable metric?

eNPS is reliable as a trend indicator and unreliable as a diagnosis. It compresses engagement into a single number, which makes it useful for tracking direction over time but poor at explaining causes. It works best paired with driver questions on areas such as leadership, workload and development, and interpreted against your own historical trend rather than unsourced global benchmarks.

Can AI be trusted to analyse employee survey results?

AI is useful for drafting survey questions, summarising supplied data and explaining definitions, but unreliable for benchmarks and conclusions because language models produce plausible rather than verified answers. Trust is also a factor: a 2025 study found 78% of workers distrust AI handling their personal feedback. Use AI for drafting and summarising, keep humans responsible for decisions, and be transparent with employees about where AI is involved.