Why your engagement survey and your exit interviews disagree.
The survey says people are broadly happy. The leavers say otherwise. Both are telling the truth, and the gap between them is the most useful thing either dataset can show you.
On this page
Surveys ask the people who stayed how good things are; exit questionnaires ask the people who left what mattered. Compare the same group over the same window, respect small numbers and separate signals from findings.
Many HR directors have sat through this meeting. The annual engagement survey has come back at a respectable level, a little up on last year, and the slide after it shows the exit data: career progression, personal growth, recognition and well-being, from people who have just resigned. Someone asks which one is right. The honest answer is both, and the reasons they differ are worth understanding before anyone tries to reconcile them.
They ask different people
An engagement survey measures the people who are still there, including the many who are content and the few who are about to leave but have not said so. An exit questionnaire measures only the people who have decided to go. Across the 30 UK organisations in our leaver benchmark, the median share of leavers who left with dissatisfaction a factor runs between 44% and 52% depending on sector. Those people were in last year’s survey too, diluted by everyone who stayed, and many of them answered it politely. The survey was not wrong; it was averaging over a population that included its own future leavers.
They ask at different moments
A survey is a snapshot taken on a date the organisation chose. A resignation is a decision taken on a date the employee chose, often after months of something not being resolved. By the time someone fills in an exit questionnaire they have had the conversation with their partner, weighed the offer and stopped hoping. They are clearer about what mattered than they were in February, and they have less reason to soften it. In our experience the exit answers are not harsher versions of the survey answers; they are more specific ones.
They ask different questions
Most engagement surveys ask one question of each topic: how good is it. A good exit process asks three: how good was it, how much did it matter to you, and what would have made a difference. The second question changes the ranking. A low survey score on something few people care about drives wasted action; a middling score on the thing that decides whether people stay is the real priority, and a satisfaction-only survey cannot tell the two apart. Exit data, which is built around what mattered, can. If your survey already asks importance alongside experience, the two datasets will disagree much less. If it does not, that is the first thing to change.
They average over different groups
An organisation-wide engagement score is a weighted average of every team. One directorate, one site or one staff group can be losing people fast while the total barely moves, because it is small, or because the rest of the organisation is fine. Leaver data is concentrated where the leaving is, so it points at the hot spot by default. The survey has the same information in it, but only if someone cuts it to the same group and the group is large enough to report.
How to read them together
When a client asks us whether a particular area is a problem, we do not start from either dataset. We start from the question, and then apply the same discipline to both. It is a method any HR team can follow.
Start from a specific question, not a general worry
“Are we losing experienced hires from the northern sites inside two years?” can be answered. “Is morale a problem?” cannot.
Compare the group with the rest of the organisation over the same window
A group’s figures mean little on their own. The question is always whether this group differs from everyone else, measured the same way, over the same twelve months.
Read both ends of the lifecycle
Put the survey, the new-starter feedback and the exit data for the same group side by side. If new starters in that group are already reporting an expectation gap, and leavers from it cite the same thing, you have a pattern rather than two complaints.
Respect the small numbers
Do not report a group of fewer than five people; a single answer can identify someone and a single answer can swing the figure. We only rank the reasons for leaving where a group has twenty or more leavers. The reporting threshold checker does this arithmetic for you.
Separate a signal from a finding
A difference that chance could easily produce is a signal to watch, not a finding to act on. Say which it is in the report. In our experience the most expensive mistakes come from treating a loud anecdote as a pattern, and the second most expensive from dismissing a real pattern because the group was small.
Code the comments properly
In our experience, free text is where the two datasets most often agree. Read every comment, assign each person to the themes they raise once, and count. Then check the counts against the scores. When the words and the numbers point the same way, you have something a board will accept.
What usually comes out
Done this way, the disagreement tends to resolve into one of three findings. The organisation as a whole is fine and one group is not, in which case the survey was right about the whole and the exit data was right about the part. The thing people are leaving over is something the survey never asked about in a way that mattered, in which case the survey needs an importance question. Or the two datasets agree after all, once the survey is cut to the leavers’ groups and tenure bands, and the problem was that nobody had cut it. All three are more useful than the argument about which dataset to believe.
We run engagement surveys and exit questionnaires on the same platform so that the two can be read together by team, tenure and staff group. WinningFormula asks how good, how important and what to do; LastOpinion asks leavers what mattered and what would have kept them; and our analysts do the reading in between.
Drawn from great{with}talent’s survey and exit method and the way we run hot-spot reviews for clients (2026), with no client figure or detail used. The leaver benchmark figures are computed from LastOpinion exit questionnaires at 30 UK organisations over the latest twelve months, no organisation under fifty responses and none named. See also Why people leave: a year of UK exit data and How to read a leaver analysis report.
The clearest thirty minutes in your week.
Send us your last engagement survey and we’ll show you what it’s missing – no charge, no deck, no obligation. Or start with a pilot or a 30-minute conversation.