The engagement survey that made things worse

A services business of three hundred people ran its first engagement survey. The response rate was eighty-one per cent, which everyone took as a good sign, and it was. The report ran to forty slides. It went to the leadership team, who discussed it for an hour, agreed it was useful, and moved on to the quarterly numbers.

Nothing was said to employees. Six months later the second survey went out. The response rate was thirty-four per cent, and the comments that did come back were noticeably angrier than the first time. One of them read, in full: what happened to the last one.

The survey had not failed to measure engagement. It had reduced it.

Why asking creates an obligation

A survey is not a neutral instrument. The act of asking people what is wrong tells them somebody intends to do something about it — that is the only reasonable inference from being asked. When nothing visible follows, employees do not conclude that the business is busy. They conclude that it asked, heard, and decided not to bother.

That is a worse position than never having asked, because it converts a diffuse dissatisfaction into a specific piece of evidence that management does not act. It is also very hard to recover from, since the next survey is read through it.

The practical rule that follows is uncomfortable and worth stating plainly: do not run a survey unless you have decided in advance that you will publish something and change something. If that commitment cannot be made, the survey is a risk with no upside.

The score is the least useful thing in the report

Most engagement reports lead with a number — an index, a score out of five, a percentage favourable — and most conversations about them are conversations about whether the number went up.

The number is nearly useless on its own. It compresses several hundred people's very different experiences into one figure, and the movement between years is usually within the range that could be explained by who happened to respond. Worse, it invites the wrong question. "How do we raise the score?" produces initiatives aimed at the measurement. "What are people telling us?" produces work aimed at the problem.

1  THE FREE TEXT, ALL OF IT
     Someone senior reads every comment, unfiltered and
     unsummarised. It is a few hours. It is the survey.

2  THE SPREAD, NOT THE AVERAGE
     A department at 4.2 and one at 2.1 average to a
     comfortable 3.2 that describes nobody.

3  THE QUESTIONS WITH THE WIDEST DISAGREEMENT
     Where half the business says yes and half says no,
     something structural differs between them.

4  WHO DID NOT RESPOND
     A team at thirty per cent response has told you
     something, and it is not neutral.

5  THE HEADLINE SCORE
     Last. For the board slide. Not for the decisions.

The instruction to read every comment personally meets resistance from senior people with full calendars, and it is the single thing most worth insisting on. A summarised comment has had its specificity removed, and specificity is the whole value — "communication could be better" is a theme, while "we found out the office was moving from a client" is a fact somebody can act on.

Two things, visibly

The common response to a survey is an action plan with eleven workstreams. Eleven workstreams produce eleven slow, partial, largely invisible efforts, and by the time any of them lands nobody connects it to the survey.

Two things, done quickly and named as coming from the survey, change more than eleven done properly. The mechanism is not the fix itself; it is the demonstration that saying something led to something. That is what determines whether people answer honestly next time, and it is the part that separates real employee engagement and culture work from an annual measurement exercise.

1. 1

   Pick one thing that is cheap and visible

   Something that can be done in weeks. The shift roster published a week ahead instead of on Friday evening. It does not have to be the biggest problem, and it should not be — the biggest problems take a year.

2. 2

   Pick one thing that is genuinely hard

   Named publicly, with an honest timeline. "Career progression is unclear. We are building job levels, and it will take until March." People are far more tolerant of slow than of silent.

3. 3

   Say what you are not doing, and why

   The most-skipped step and the most valuable. If people asked for something the business will not do, say so and give the reason. An honest no is a better answer than a year of nothing, and it stops the same request returning.

4. 4

   Report back at the timeline you gave

   Even if the answer is that it slipped. A missed date acknowledged keeps credibility; a missed date ignored spends it.

Do not ask about things you cannot change

Standard survey templates ask about pay, about senior leadership, about strategy. If the business has no intention of revisiting pay this year and no mechanism for employees to influence strategy, asking produces data that can only disappoint.

This is not an argument for a survey that only asks safe questions. It is an argument for asking about the space where action is actually possible, and being honest about the rest. A shorter survey about things that can move is more useful than a comprehensive one whose findings sit in a drawer.

Pay deserves a specific note, because it appears near the top of almost every engagement survey and is almost never a pure pay problem. Where people cannot see how pay is decided, they read every outcome as arbitrary, and the survey records it as dissatisfaction with the amount. The fix is usually structural rather than financial, which is why this finding tends to hand off to compensation and rewards strategy work rather than to a budget conversation.

Anonymity, and why people do not believe it

Every survey promises anonymity. A significant proportion of employees do not believe it, particularly in smaller teams, and particularly where the demographic questions would identify them anyway — a team of four, split by gender and tenure, is not anonymous by any practical definition.

Two practical measures. Do not report any group smaller than five, and say so before people answer rather than afterwards. And cut demographic questions you have no intention of analysing; every additional one raises the perceived risk of identification and lowers response honesty for no gain.

Where trust is already low, the first survey will understate the problem. That is worth knowing in advance so the results are not read as reassuring. A business whose first survey comes back bland and positive after a difficult year has usually measured caution rather than contentment.

What to do if you are already in the hole

If a survey has been run and nothing followed, the recovery is not a better survey. It is an acknowledgement.

Say that the last one produced no visible action, say what the findings were, and say what happens now. It is an awkward five minutes and it is the only route back. Running a second survey without addressing the first is what took the response rate in the opening example from eighty-one to thirty-four.

Then run the smaller version. Three questions, one team, acted on within a month. Rebuilding the belief that feedback goes somewhere is a sequence of small demonstrations, not a single large gesture, and it is the actual substance of employee engagement and culture work — the survey is an instrument, not the programme. Where the underlying finding turns out to be that people leave before anyone asks them anything, the more urgent conversation is about retention and attrition rather than engagement measurement.