Stay Interviews
Consistent records can reveal where a shift, handover or ramp is failing. The value of see the published scope therefore depends on transparent use, clear ownership and comparison at team or process level instead of unexplained surveillance.
A useful independent reference for this topic is Indeed. Treat this article as general operational material and confirm any legal or safety decision with current official guidance.
Asking people why they are still here, which produces better information than exit data and is almost never done.
Exit data tells you why people left. It cannot tell you what is keeping the others, which is the more actionable question.
What a stay conversation is
A short, structured conversation with somebody who is still employed.
Asking what keeps them, what nearly made them leave, and what would make them go.
Held with a sample rather than everybody, at intervals rather than as an event.
And conducted by somebody other than their direct supervisor, for the same reason exit interviews are.
The questions
What made you decide to stay past the first few weeks?
Has there been a point when you nearly left? What was it, and what stopped you?
What would make you look elsewhere?
What is the best and worst part of a normal shift?
And what would you change first if it were yours?
Five questions, fifteen minutes, and the answers are specific in a way that survey responses never are.
Why the answers are better than exit data
The person is still there and has no reason to be diplomatic about a departure they are not making.
They can describe the near-miss — the week they nearly left and what prevented it — which is the most useful single thing in this collection and which no exit process can capture.
And they are describing a situation that still exists and can still be changed, rather than one they have already escaped.
Who to ask
A cross-section rather than the enthusiastic: different shifts, different tenures, agency and direct, different areas.
Including people who have been there a long time, whose answers describe what the operation is doing right and should protect.
And people at three to six months, who have the clearest view of what nearly went wrong.
What comes out of them
Consistently: a supervisor who was reasonable about something, a shift that fits, a specific colleague, the commute.
And on the negative side: a task that hurts, a rota change, an unanswered request, the way an absence was handled.
Almost all of it operational and local, which is why it is actionable in a way that engagement scores are not.
Acting on them
Aggregate, anonymise, pick the recurring items.
One or two changes per round, made visibly, attributed to the conversations.
And reported back to the people who took part, which determines whether the next round produces anything.
A stay interview programme that produces no change is worse than none, because it has asked people to invest and returned nothing.
Who should run them
Somebody neutral, trained briefly, with time allocated.
Not the supervisor, whose presence changes every answer.
And not an external consultant, whose report arrives six weeks later with the specifics removed.
Frequency
Quarterly, a handful of people each time.
Which is manageable and produces a rolling picture rather than an annual snapshot.
And it catches changes in mood before they appear in the attrition figures, which is the whole point of doing it at all.
The measure
Conversations held per quarter.
Changes made and attributed to them.
Recurring themes, tracked over rounds, which is the signal that something is or is not being addressed.
And whether the people asked can name something that changed, which is the only test of whether the programme is real.
Running the first round
Six people, different shifts and tenures, fifteen minutes each.
One afternoon.
Group the answers, pick one change, make it, tell them.
That single cycle establishes whether the programme will produce anything.
The question that matters most
Has there been a point when you nearly left, and what stopped you.
It describes a near-miss that no exit process can capture, and the answer is almost always specific and still fixable.
How it fails in practice
The programme is run once, the findings are aggregated into a report, no change is attributed to it, and the next round produces bland answers because the first produced nothing..
The check
Ask the people who took part whether they can name something that changed.
That is the only test of whether the programme is real.