In this guide
There are a hundred lists of exit interview questions and almost no evidence about which of them work. What follows is a question bank organised by what each question is actually testing, with the difference between an answer you can act on and an answer that only sounds like one, plus the part most lists skip, which is what to do with twenty of them once you have them.
Most exit interviews are theatre
The person has already gone. They have nothing to gain from precision, a reference to protect, and they are talking to the organisation they just left. Expect diplomacy, especially about anything involving a named individual.
You will see a figure claiming that 40 to 63 per cent of people change their stated reason for leaving when a neutral third party asks instead. It is published by a firm that sells third-party exit interviews, without a method, a sample or a definition, and we would not put it in a board pack. The oldest academic reference, Lefkowitz and Katzin 1969, found only weak convergence between reasons given at exit and reasons given in a later follow-up, which supports the concern without sizing it.
So: the distortion is real, nobody has credibly measured it, and it is worst exactly where you most want the truth. Design around that rather than pretending otherwise.
The one thing an exit interview establishes well
Timing. A departing employee who is being careful about why will usually still tell you when: when they started thinking about it, when they started looking, when it became definite. Dates are less loaded than judgments, so they survive the diplomacy.
Timing is also the most useful thing you could possibly learn, because it tells you how long the warning was visible. If the average answer is four months, then for four months there was something to notice and nobody was looking. That is a far more actionable finding than another tally of “career growth” as a leaving reason.
The twenty questions
Do not ask all of them. Eight to twelve, chosen for what you actually intend to act on, beats twenty asked out of completeness. The right-hand columns matter more than the questions: the same question gets you nothing or something depending on whether you probe.
Timing
Ask these first, while they are still being candid, and write the months down as numbers.
When did you first seriously start thinking about leaving?
Testing: The length of the decay window, and the most valuable answer in the interview.
Useless answer: “A while ago, I suppose.”
Usable answer: “Around March, after the replatform slipped the second time.”
Probe: What was happening at work around then?
When did you start actively looking?
Testing: The gap between disengaging and acting. That gap is your intervention window.
Useless answer: “Fairly recently.”
Usable answer: “June. About three months after I first started thinking about it.”
Probe: What changed between those two points?
Was there a moment where it became definite rather than an idea?
Testing: Whether there was a trigger or a slow accumulation. Both are fixable; they need different fixes.
Useless answer: “It just built up.”
Usable answer: “When the reorg was announced and my project moved without anyone telling me.”
Probe: Who could have told you, and when?
The push
What made leaving feel reasonable. Let them use their own categories before you offer yours.
What was the hardest part of working here in the last six months?
Testing: The dominant stressor, in their words.
Useless answer: “It was fine, just busy.”
Usable answer: “Being the only person who understood the billing service, and never getting cover.”
Probe: How long had that been true?
What did you stop doing, or stop caring about, before you decided to leave?
Testing: Disengagement behaviour. People describe this far more honestly than they describe feelings.
Useless answer: “Nothing really.”
Usable answer: “I stopped putting my hand up in planning. Around April.”
Probe: Did anyone notice?
Was what you experienced specific to you, or do others have it too?
Testing: Clustering risk, without inviting them to name individuals.
Useless answer: “Probably just me.”
Usable answer: “Everyone on the platform team is carrying it. I was the first to go.”
Probe: What would you look at to tell?
Workload and capacity
The questions that find burnout without using the word, which is important: people deny the word and describe the condition.
In a normal week, how many hours were you actually working?
Testing: Load as a number rather than an adjective. Ask, then wait.
Useless answer: “Quite a lot.”
Usable answer: “Fifty-five, and most Sundays for the last two months.”
Probe: When did that become normal rather than a crunch?
Was there a point where the load stopped feeling temporary?
Testing: The moment effort turned into exhaustion.
Useless answer: “It was always busy.”
Usable answer: “When the second launch date moved and nothing came off the plan.”
Probe: What would have had to come off the plan?
When did you last take leave and genuinely not check messages?
Testing: Whether recovery was possible at all. A long gap here predicts the rest.
Useless answer: “I took leave in the summer.”
Usable answer: “Last Diwali. Every day off since, I have been on Slack.”
Probe: What made it impossible to switch off?
Did your workload ever come up with your manager? What happened?
Testing: Whether the organisation had a chance to act, and what it did with it.
Useless answer: “I mentioned it once or twice.”
Usable answer: “Twice. Both times I was told it would ease after the launch. It did not.”
Probe: What would a useful response have looked like?
The manager, and the conversation that did not happen
Ask these even when the manager is not the reason. The absence of a conversation is itself the finding.
In the last three months, did anyone talk to you about your future here?
Testing: A yes or no you should count. Gallup found 51% of leavers say nobody did.
Useless answer: “We had one-to-ones.”
Usable answer: “No. We talked about tickets every week and never once about me.”
Probe: When was the last conversation about you rather than the work?
Did you feel able to raise problems? What happened the last time you did?
Testing: Whether the feedback path worked. The second half is where the answer is.
Useless answer: “Yes, the team is open.”
Usable answer: “I raised the on-call rota in January. It was noted. Nothing changed.”
Probe: What did you conclude from that?
What did your manager do well, and what would you have wanted more of?
Testing: Balance. Asked this way you get an answer; asked as “how was your manager” you get a reference.
Useless answer: “They were good.”
Usable answer: “Great at unblocking. Never once asked how I was actually doing.”
Probe: Would that have changed anything?
Preventability
Ask directly, then make them be specific. Vague agreement here is worthless.
Is there anything we could have done that would have kept you?
Testing: The addressable share of your attrition. Gallup put this at 52%.
Useless answer: “Not really, it was just time.”
Usable answer: “Yes. A real conversation about the lead role in February.”
Probe: What would we have had to change, and by when?
If that one thing had gone differently, would you still be here?
Testing: Whether the stated reason is the real one. A hesitation is itself information.
Useless answer: “Maybe.”
Usable answer: “Honestly, yes. In February. By June it was too late.”
Probe: What made June too late?
The pull
People are far more candid about what they are going to than about what they are leaving.
What does the new role give you that this one did not?
Testing: The actual gap, described as a positive.
Useless answer: “It is a good opportunity.”
Usable answer: “I will own a service end to end, and there are three other senior engineers on it.”
Probe: Could that have existed here?
Was pay the main reason, or the reason leaving became possible?
Testing: The most important distinction in the interview. Pay is often the enabler, not the cause, and the two imply completely different fixes.
Useless answer: “The money is better.”
Usable answer: “The money made it possible. I would not have looked if the last year had been different.”
Probe: What would you have needed instead of the money?
Closing
Two questions, and then silence.
Would you work here again, and would you recommend us to someone you respect?
Testing: Residual goodwill and your referral pipeline. Two different questions, so ask both.
Useless answer: “Sure, maybe.”
Usable answer: “I would come back to a different team. I would not send a friend to mine.”
Probe: What would have to be true?
What should I have asked you that I did not?
Testing: The highest-yield closing question there is.
Useless answer: “No, that covers it.”
Usable answer: “Whether I ever told anyone. I did, twice, and nothing happened.”
Probe: Nothing. Just wait.
Who should ask, and when
Never the direct manager. If the manager is any part of the reason, having them run the interview guarantees you will not hear it. HR or a skip-level is the practical answer, and you should say who it will be in the invitation so nobody is surprised.
On timing, every source contradicts the others (the last day, the final week, halfway through the notice period, ninety days after they start the next job) and none of them cites evidence. Pick one and keep it consistent, so that your own answers are comparable across exits. Indian notice periods of 30 to 90 days give you more room to choose than most markets.
Say up front who will read it, whether it is attributed, whether it is recorded, and how long it is kept. India’s DPDP Act applies to employee personal data and an exit interview is full of it, so settle your notice and retention position with your legal team before running this at scale. None of this is legal advice.
Coding the answers so twenty exits become a pattern
A folder of free text is not data. Before you run the first interview, decide the small set of categories every exit will be coded into, and code each one the same way. Ten categories is plenty: workload or burnout, manager, career growth, compensation, culture, role fit, job security, location, personal, and the pull of a better offer.
Force a single primary reason. “A bit of everything” is how a pattern gets lost. Keep the free text for the quote you will repeat in a leadership meeting, which is what it is good for, and keep the categories for counting.
One exit is a story. Ten are a signal. Do not restructure a team on the strength of one articulate leaver.
The two numbers worth more than the rest
If you are building a business case for doing something about retention, two numbers decide the answer, and neither of them can be borrowed from a published benchmark because no credible one exists for India:
- The share of voluntary exits that are burnout-driven. Ask directly whether workload was a factor, record yes, no or partly, and count it.
- How many months the decay ran before the resignation. Straight from the timing questions.
Those two feed the cost of burnout attrition model directly. They are the two inputs that move its answer most, and the only two it refuses to supply a default for. Read them together with the proportion who say nobody discussed their future: if the decay ran four months and half your leavers had no such conversation, the information existed and nobody went looking for it.
Ask them six months earlier
Everything above documents a loss you have already taken. The same questions, asked of someone who has not resigned, can still change the outcome, and most of this bank works unchanged as a stay interview.
Run them with the people carrying the most load rather than the people who seem unhappy; those are different lists, and the first is the one that surprises you. Run them after a launch, a reorg, or a departure from the team, because the weeks after a colleague leaves are the highest-risk window there is. And do not ask unless you intend to do something: a stay interview that produces no visible change has demonstrated that telling you is pointless.
The exit interview kit packages all of this as a workbook: the question bank, a tracker with the coding built in, a rollup that computes the two numbers above, and the stay-interview version of the questions.
Questions people ask
What are the best exit interview questions?
The ones that establish timing, because those are the questions a departing employee answers accurately even when they are being diplomatic about everything else. "When did you first seriously start thinking about leaving?" and "when did you start actively looking?" tell you how long the warning was visible and how big your intervention window was. Questions about feelings get you politeness; questions about dates and behaviour get you facts.
Who should conduct an exit interview?
Not the direct manager. If the manager is any part of the reason the person is leaving, having them run the interview guarantees you will not hear it. HR or a skip-level is the practical answer for most Indian companies, and you should say who it will be when you send the invitation. Third-party interviewers are sometimes recommended, but the research supporting that recommendation is published almost entirely by firms that sell third-party exit interviews.
Do employees tell the truth in exit interviews?
Partly, and the honest answer is that nobody knows how partly. The figure you will see quoted (that 40 to 63 per cent of people change their stated reason when asked by a neutral third party) comes from a vendor and is published without a method, sample or definition. The oldest academic reference, Lefkowitz and Katz in 1969, found only weak convergence between reasons given at exit and reasons given in a later follow-up. Treat the distortion as real, unquantified, and worst on anything involving a named person.
When should an exit interview happen?
Every source gives a different answer: the last day, the final week, halfway through the notice period, or ninety days after they have started the next job. None of them cites evidence. Pick one, keep it consistent so your own data is comparable across exits, and judge it by whether the answers you get are specific. In India, where notice periods run 30 to 90 days, there is more room to choose than in most markets.
What is the difference between an exit interview and a stay interview?
Six months, and all of the value. A stay interview asks substantially the same questions of someone who has not resigned, which means the answers can still change something. An exit interview documents a loss you have already taken. If you only have the appetite to run one of them properly, run the stay interview.