The employee survey: what you ask, how often and what happens afterwards

Echipa HR 365 · reviewed 2026-09-04 · 5 min read

The rule that decides whether the survey is worth running: do not send it unless you are ready to communicate the results and change one thing from them. A survey with no consequences teaches people that their view gets collected, not used — and by the second wave the response rate collapses and never recovers.

What you measure: the dimensions that produce action

A bad satisfaction survey is a list of questions whose answers cannot change anything. Every dimension has to have an owner and a lever: if the score is low, somebody specific can do something specific.

The cadence: how many times a year

Twice a year for the full survey, with eight to twelve questions. More often becomes noise: these dimensions do not change in six weeks, and the variation you measure is Wednesday’s mood, not the organisation.

In between, if you want a denser rhythm, use a single item repeated monthly — usually the most predictive one is workload. A monthly single item gets a high response rate precisely because it is one item.

The three things you must do afterwards

Within three weeks of the survey closing

1. Leadership or HR — communicate the results to everybody, the weak scores included — especially the weak ones
2. Leadership — name one single thing that will change, with an owner and a deadline
3. HR — say explicitly what will not change and why; silence on a raised subject is read as refusal

The third step is missing almost everywhere and is the cheapest. People accept a “no, and here is why” far more readily than they accept no answer at all — and at the next wave they carry on responding.

Anonymity: guaranteed, not promised

The promise of anonymity collapses the moment results can be sliced to a team of three. The practical rule: no report is shown for groups under five respondents, and open questions are never shown by department — they are read only in aggregate, at company level.

If the tool asks for a name or is completed from a personal account, say so upfront. A survey presented as anonymous that turns out to be identifiable does more harm than one openly attributed.

The open questions: two, not ten

Open questions produce the best information and are the most expensive to process. Two are enough, and the phrasing matters more than the number.

The second phrasing is the one worth using if you pick only one. It splits the answer into three categories you can count and compare between waves, unlike free comments, which get read once and lost.

The response rate: what it actually means

A rate under 50% makes the results hard to interpret, because the people who respond are not a random sample: the very satisfied and the very dissatisfied answer, and the middle stays silent. The rate itself is the first indicator you read, before any score.

What raises it, in order of effectiveness: time allocated during the working day, communicating the results of the previous wave, a survey under ten minutes, and evidence that something changed last time. The last is the strongest and the only one that accumulates over time.

Where to start

Before building the survey, answer one question: what are you willing to change on the basis of the results? If the answer is nothing for now, postpone it. Nothing is lost by postponing; by sending it, you lose the instrument.

If the answer is something concrete, cut the dimensions down to those with a clear owner. A survey of eight useful questions beats one of twenty nobody can move.

Fix the date for communicating the results before sending the survey, and announce it alongside. It is the only commitment that raises the response rate at the next wave.