TL;DR
- The real failure: most exit interviews are designed to be answered, not to be useful. A question that asks someone to generalise gets a generalisation. A question that asks for a specific occasion gets evidence.
- When doing nothing is right: if the leaver cohort is small, the role is short-tenure, and you've no pending structural change in that function, an annual sample conversation outperforms a per-leaver ritual you can't resource properly.
- What has to be true: the interviewer must be outside the leaver's reporting line, the notes must be disclosable to the leaver on request, and at least one person in the room must be able to convert a finding into a decision.
- How the options split: generic open prompts, structured rating scales, and incident-based questions each pull different signals, and each fails in a different way.
- Decision rule: if you can write down the decision a useful answer would change, your question set is shaped right. If you can't, you're running a satisfaction survey that happens to be timed to someone's last day.
- Outcome to expect: a smaller set of specific, dated, named incidents, written in the leaver's words, owned by someone outside HR, and acted on within a defined window. Everything else is ceremony.
The Tuesday After
A senior product designer hands in her notice on a Wednesday afternoon. By Friday her line manager has booked the standard thirty-minute exit slot for the following Tuesday. The HR business partner who runs the slot has a template printed out. It opens with the question everyone opens with. Why are you leaving. The designer, who has worked there for four years and survived two reorgs, looks at the template and gives the answer she has already told her manager twice. Better opportunity, more senior scope, fully remote. The HR business partner writes it down. The interview runs fifteen minutes under slot. The notes go into the leaver file. Nothing else moves.
Most exit interviews in most organisations end like that one. Not because the leaver is being evasive. Because the question was built to be answered, and the answer arrived as planned.
But the question that would have surfaced something different sat two lines further down on the same template, and nobody asked it. It asks for a specific moment. A meeting. A decision. A Tuesday in March when the redesign was paused and never restarted. The designer would have answered that one differently. She would have given a date, a name, and the thing that was said in the room. That answer would have been evidence. The first answer was courtesy.
Best tools for Employee Retention
The mechanism is simple and almost no template respects it. A question that asks someone to generalise gets a generalisation, and a generalisation can't be checked, assigned, or acted on. A question that asks for a specific occasion gets evidence, and evidence can. The work below is about rebuilding the question set so it pulls the second kind of answer out of the room.
When You Do Not Need to Act
Plenty of organisations run an exit conversation well enough that redesigning it would be a misuse of time. Pretending otherwise makes the rest of this piece look like a sales argument rather than a working tool. The honest test is the size and shape of the leaver cohort, not the rhetoric you've read about exit interviews. Four stages describe most of what you'll see.
Your current setup is genuinely fine. You're a fifty-person company losing four or five people a year. Most leavers are short-tenure. Your HR lead runs a thirty-minute conversation in week one of the notice period, writes a one-page summary in the leaver's own words, and reads the summaries back at the quarterly leadership meeting. The summaries are short, specific, and have caused a manager to change something twice in the past three years. You don't need a new question set. You need to keep doing that one.
There's friction worth relieving. You run exit interviews. They produce notes. The notes sit in a folder that nobody opens between cycles. You're doing the work, but the work has nowhere to land. The fix here's not a better question. It's a documented owner for each exit finding and a deadline for the response. The question set can stay exactly as it's.
There's a real risk of misreading the signal. Your leaver cohort is large enough to matter, but the conversation is run by the line manager, not by an independent interviewer. The notes are thick. Decisions never change. The risk isn't that the leavers are lying. The risk is that the interviewer is the subject of the finding, and the notes have been quietly edited. You need a structural change, not a question change.
The edge case. You operate in a country where works councils or employee representatives have consultation rights around leavers, or you handle large numbers of redundancies, or you sit in a sector where a single exit finding can trigger a regulator's interest. The conversation itself becomes a piece of evidence. The questions have to be defensible under that lens, and the notes have to be written with the assumption that the leaver will read them. Most published question sets were not built for that.
If you sit in the first stage, stop reading and go close a real finding instead. If you sit in the second or third, the rest of this piece is for you. If you sit in the fourth, take local advice before changing anything, because the shape of the exposure changes by jurisdiction.
Five Questions You Are Asking Yourself at 11pm
Do leavers even tell the truth on the way out. Not always, and not in the way the article you read last week implied. They tell a version that protects the reference, the relationships they still need, and the story they want future employers to hear. They don't, as a rule, fabricate. The job is to design questions that make the truthful version easier to say than the safe one. A question about a specific moment lets the leaver describe what happened without indicting anyone by name. A question about feelings forces them to either indict or soften, and most choose to soften.
Is the interview I am running the interview I would want to run. Sit with the last twelve summaries you wrote. Underline every sentence that contains a date, a name, or a verb that implies action by someone inside the company. If the underlined sentences are less than a quarter of the whole, you're collecting mood. Mood is interesting. Mood isn't actionable.
Who in this organisation will actually do something with what I write down. Name the person. If you can't, the exit interview is a ritual, and you're the priest. The interview can still be worth running for the leaver's experience, but don't pretend it produces operational signal.
What happens to the notes if the leaver asks for them. In most jurisdictions, the notes are personal data about a living individual, and the individual has rights of access. Write them as if the leaver will read them next week. That single habit eliminates the worst of the editorial drift and makes the document defensible.
If the line manager runs it, what does the leaver not say. The answer to that question is the reason most published guidance recommends one. It doesn't have to be HR. It has to be outside the reporting line, and it has to be briefed on the questions and the limits.
Three Honest Categories the Approaches Split Into
Generic open prompts. The why-are-you-leaving family. What could we've done differently. What did you like most about working here. Cheap to write, easy to defend, almost impossible to act on. They produce the leaver's curated narrative, which is usually true and usually incomplete. They're right when the goal is to give the leaver a courteous hearing and surface obvious red flags like harassment, where the answer doesn't need to be specific to be actionable. They fail when the goal is operational signal about the function, the manager, or the work itself, because the leaver's curated narrative is built to understate those.
Structured rating scales. Rate your manager from one to ten. Rate your workload. Rate your tools. Easy to aggregate, easy to chart, easy to compare against last quarter. The numbers make it into a slide. The slide makes it into a leadership meeting. The leadership meeting doesn't change anything. Rating scales are right when you've a stable leaver cohort and you're tracking a small number of dimensions over time, and when the dimensions are concrete enough that the leaver's intuition is the right signal. They fail when the scale becomes the question, because the leaver is being asked to summarise a complex situation into a single integer, and the integer then has to defend itself against a manager who wants to argue with it.
Incident-based and counterfactual questions. Walk me through a specific moment when you decided to start looking. What would have had to be different for you to have stayed. What did your manager do in the last month that worked, and what didn't. These take more interviewer skill, produce longer notes, and yield the only kind of answer that changes a decision. They're right when the goal is to learn something the organisation didn't already know. They fail when the interviewer can't hold space for the answer, or when the answer is so specific that it identifies a person the leaver didn't mean to name. Both failure modes are real, and both are teachable.
Five Diagnostic Questions You Can Self-Assess Against
Can you name, from memory, three specific changes your organisation made in the last two years because of an exit interview finding. If you can, the system is producing signal and the question is whether to extend it. If you can't, the system is producing summaries, and the summaries aren't landing. To answer this for real, pull the last twenty exit summaries and read them with a pen. Underline every sentence that implies an action someone inside the company took or could have taken.
What proportion of last year's leavers sat with an interviewer outside their reporting line. A rough count is enough. If the answer is less than half, the notes are softer than they look, and the softness is structural.
When in the notice period does the conversation happen. Week one captures the moment of leaving and the relief of having decided. Week three captures the handover and any bitterness that built up during it. Neither is wrong. Both produce different answers. Pick one, write the choice down, and tell the leaver which window they're in.
What is the longest note in your last batch and what is the shortest. If the longest is over a page, you're interviewing well and writing too much. If the shortest is under half a page, you're not interviewing at all.
If a leaver asked for their notes tomorrow, what would they read. Read three summaries back to yourself as if you were the leaver. The exercise is unpleasant. It's the only one that finds the editorial drift.
Six Question Types, and What Each One Actually Yields
The open why-are-you-leaving question
What it's. The opening prompt on most templates. Why are you leaving. What brought you to this decision. Tell me about your decision to move on. It's usually the first thing the interviewer says and almost always the worst thing they ask.
Why it earns a place. It gives the leaver permission to set the frame. It produces a first sentence, and the first sentence often contains the answer the leaver has been rehearsing. That rehearsed answer is worth recording, because the gap between it and the answers to later questions is itself diagnostic. If the rehearsed answer and the specific answer agree, the leaver is being candid. If they diverge, something is being managed.
Where it genuinely falls short. It produces a generalisation, and a generalisation is a closed loop. There's nothing to verify, nothing to assign, and nothing to action. The leaver says career growth, the interviewer writes career growth, the leadership team reads career growth, and no one asks what career growth means in this organisation and on this team. Used as the whole interview, this question is a courtesy. Used as the door into the interview, with specific-moment and counterfactual questions waiting behind it, it earns its slot.
The rating scale
What it's. Rate your manager from one to ten. Rate the clarity of your goals. Rate your workload. Rate your tooling. Sometimes anonymised, sometimes not. Always aggregated.
Why it earns a place. It produces a number. Numbers travel. A number makes it into a deck. A deck makes it into a meeting. A meeting is where, occasionally, a decision gets made. The scale is the only format that lets you compare one cohort of leavers against another, which is the only way to know whether the intervention you made after the last batch actually moved anything.
Where it genuinely falls short. It asks the leaver to do the hardest part of the analysis themselves. They have to summarise a year of working life into a single integer, and they have to do it under conditions of partial candour. Worse, the integer then has to defend itself against a manager who wants to argue with it, and the manager usually wins, because the manager has more context and more at stake. Rating scales also systematically overweight the most recent months, because that's what the leaver remembers, which is exactly the period that least explains why they left. The scale is a tracking instrument, not a diagnostic one.
The would-you-recommend-us question
What it's. The exit interview version of the employee net promoter question. Would you recommend us as a place to work. Would you apply again. Would you take this job again if offered.
Why it earns a place. It produces a comparable signal across cohorts and across teams. It survives the aggregator. It looks good in a quarterly review. Used as one data point among several, it's honest.
Where it genuinely falls short. The leaver is asked to express an opinion about a complex organisation in a single binary, and the answer is shaped by how the leaver wants to be seen rather than by what the leaver knows. A leaver who is angry says no and means it. A leaver who is leaving on good terms says yes and means almost nothing by it. The question's real failure is that it tempts the interviewer to stop. A yes lets the interviewer move on. A no triggers a follow-up that the leaver, having given the rehearsed answer, is now under pressure to perform.
The specific-moment question
What it's. Walk me through a specific moment when you decided to start looking. Tell me about a meeting in the last six months where something changed. Describe a conversation with your manager that you remember.
Why it earns a place. It's the only question type that produces evidence. A date, a place, a name, an event. Evidence can be checked. Evidence can be assigned. Evidence can be turned into a decision. A finding that names a Tuesday in March when the redesign was paused, with the person who paused it and the reason given, is a finding an organisation can act on. A finding that says the redesign process was unclear isn't.
Where it genuinely falls short. It requires an interviewer who can hold space for a long answer without leading, paraphrasing, or rescuing the leaver from silence. Most line managers can't. Most HR generalists can, with training. The other failure mode is identification. The specific moment will, often, identify a person the leaver didn't mean to name. The interviewer has to be ready to receive that, to protect the leaver from unintended consequences, and to write the note in a way that conveys the signal without indicting the individual by name in a document the leaver will later read.
The counterfactual question
What it's. What would have had to be different for you to have stayed. If you could change one thing about how this team worked, what would it be. What did your manager do in the last month that worked, and what didn't.
Why it earns a place. It bypasses the rehearsed answer. The leaver who has spent two weeks constructing a graceful exit narrative has not, in most cases, constructed a counterfactual. The counterfactual is closer to the unedited truth, and it's closer still to a finding you can act on, because the leaver has just done part of the work of recommending the action.
Where it genuinely falls short. The counterfactual is still the leaver's view. A leaver who wanted promotion and didn't get it will say the promotion policy was unfair, which is one reading and not the only one. The interviewer has to hold the counterfactual as evidence without treating it as a verdict. Used alone, the counterfactual question produces a list of wishes. Used alongside the specific-moment question, it produces a list of wishes that can be tested against incidents.
The what-should-we-ask-that-we-have-not question
What it's. The last question in a well-run interview. Anything we've not covered that you want on the record. Anything you wish someone had asked you.
Why it earns a place. It returns control to the leaver. It gives them one chance to put on the record the thing they have been carrying through the interview and didn't say. That thing is, often, the most useful sentence in the entire transcript. It's also the sentence that signals to the leaver that the interview was a conversation, not an interrogation, which is the difference between a leaver who tells their friends the exit interview was worth doing and one who tells their friends it was a tick-box.
Where it genuinely falls short. It only works if the interview up to that point has been conducted well. If the prior questions have been leading, or have named the manager, or have hinted at the answer the interviewer wants, the leaver will use the open prompt to give the safe answer, because the safe answer is the one that matches the frame the interviewer has already set. The question is also the easiest to drop under time pressure, and the one whose absence the leaver notices most.
The Decision Table
| Situation | Scale | Setup | Primary Pain | Recommended Starting Point |
|---|---|---|---|---|
| Small team, low leaver volume, current process lands | Under 50 people, fewer than 5 leavers a year | HR lead runs the conversation, summary read at quarterly meeting | None of significance | Keep the current set, document the owner of each finding |
| Medium team, notes exist but do not change decisions | 50 to 250 people, steady leaver flow | Conversations run by line manager, notes filed in a folder | Findings have nowhere to land | Move the conversation outside the reporting line and assign a finding owner with a deadline |
| Large team, structured process, scales but does not diagnose | 250+ people, formal programme with template | HR runs it, rating scales dominate | Numbers in decks, no operational signal | Replace at least two rating items with specific-moment questions |
| Sector or jurisdiction with consultation rights around leavers | Any size, regulated environment | Notes must be defensible if read by the leaver or a representative | Notes can become evidence | Write as if the leaver will read the note next week, take local advice |
| Heavy redundancies, leaver cohort is structurally different | Any size, redundancy programme | The leaver's narrative is shaped by the programme, not the work | Findings about the work are masked by feelings about the process | Run a separate post-programme review, do not collapse redundancy exit conversations into the standing exit interview |
| Early-tenure leavers dominate the cohort | Any size, short average tenure | The conversation happens before the leaver has enough to evaluate | Findings are about onboarding, not about the job | Shift the question set toward the first ninety days and treat it as an onboarding review with an exit timestamp |
| Manager is the subject of a recurring finding across cohorts | Any size, multiple leavers flag the same individual | Findings cluster around one name | The conversation has become a personnel matter | Hand the interview to someone outside the function, redact the manager's name in the shared summary, treat the pattern as a separate workstream |
| Exit interview is being added for the first time | Any size, no current process | Starting from a blank page | Temptation to build a fifty-question template | Start with eight questions, three of them specific-moment, run it for a quarter, then extend |
A Question Set You Can Run Tomorrow
The set below is ordered. The first question earns the leaver's trust. The middle questions pull specific evidence. The last question returns control to the leaver. Run the questions in this order. If you only have fifteen minutes, drop from the bottom. Never drop from the top.
| Number | Question | Purpose | Answer it surfaces |
|---|---|---|---|
| 1 | Tell me about your decision to move on. | Lets the leaver set the frame and produces the rehearsed answer. | The narrative the leaver has been telling. |
| 2 | Walk me through a specific moment in the last six months when you decided to start looking. | Forces an incident. | A date, a meeting, a triggering event. |
| 3 | What did your manager do in the last month that worked, and what did not. | Gets at management behaviour without asking the leaver to indict the manager. | Specific behaviour, not a rating. |
| 4 | What would have had to be different for you to have stayed. | Produces a counterfactual. | A testable wish, close to a recommendation. |
| 5 | Describe a meeting in the last quarter you remember clearly. It can be a good one or a bad one. | Pulls another incident, with the leaver choosing the valence. | A second piece of evidence, triangulated against question 2. |
| 6 | If you could change one thing about how this team works day to day, what would it be. | Pulls a recommendation. | An actionable suggestion. |
| 7 | What should we be asking leavers that we are not. | Returns control. | The sentence the leaver has been carrying. |
The point of the ordering is to move from general to specific, then back to general. The opening frame question gives the leaver a runway. The specific-moment questions do the work. The closing counterfactual and recommendation questions convert the evidence into a usable form. The final open question catches what the rest missed.
The Questions to Cut
These are the questions that feel important, that every template includes, and that cost you more than they return. Each one of them trades the leaver's goodwill for a piece of information you could not act on anyway. Cutting them is the single biggest improvement you can make to the conversation.
| Question | Why it stays on templates | Why it produces nothing |
|---|---|---|
| What did you like most about working here. | Feels kind. | Produces a compliment the leaver has already given in the leaving card. Cannot be actioned. |
| What did you like least about working here. | Feels balanced. | Forces the leaver to summarise a year into a single worst thing, then defends the summary against the manager who wants to argue with it. |
| Rate your manager from 1 to 10. | Produces a number. | The number cannot be verified, cannot be assigned, and cannot be defended against a manager with more context. |
| Did you feel valued here. | Sounds like a culture question. | Gets a yes or no shaped by how the leaver wants to be seen, not by what the leaver experienced. |
| Would you recommend us as a place to work. | Easy to aggregate. | A leaver who is angry says no and means it. A leaver who is not says yes and means almost nothing. The signal is mostly in the denominator, and the denominator is contaminated. |
| If you could do anything differently in your career here, what would it be. | Sounds reflective. | Pulls the conversation onto the leaver's choices, away from the organisation's. The interview is not the place for that. |
| Is there anything else. | Feels open. | Without the specific-moment questions earlier in the set, the leaver has nothing to add. The question earns its name only when the prior questions have earned their answers. |
The cost of these questions isn't just the time. It's the goodwill. Every question that doesn't produce evidence teaches the leaver that this is a ceremony. By the time you reach the specific-moment question, the leaver has already decided how much to give you. Cut the dead questions early, and the specific-moment questions land.
What to Put in Writing
| Artefact | Who owns it | When it is written | What it prevents |
|---|---|---|---|
| The exit interview summary, in the leaver's own words | HR business partner or independent interviewer | Within five working days of the conversation | Editorial drift, post-hoc rewriting, dispute over what was actually said |
| The list of specific findings, each with a named owner | Whoever runs the exit programme for the function | Within ten working days of the conversation | Findings that sit in a folder and never reach the person who can act on them |
| The response deadline for each finding | The named owner | At the same time as the finding is logged | Findings that age into irrelevance because no one was watching |
| The anonymised cross-cohort summary | Head of people or function lead | Quarterly | A single leaver's view being treated as representative |
| The redaction note explaining any names withheld | The interviewer | At the point the note is written | Identification of individuals the leaver did not mean to name, and later dispute about it |
| The confirmation that the leaver was told the notes are disclosable on request | HR business partner | At the start of the conversation, recorded in the note | Later argument about whether the leaver understood the framing |
| The decision record for any finding that was acted on | The named owner | At the point of action | Repeat findings being raised in the next exit interview with no memory of the response |
| The decision record for any finding that was not acted on, with the reason | Head of people | Quarterly | Findings being quietly dropped without explanation, which is the single thing that poisons the next round of interviews |
The artefacts above aren't bureaucratic. They're what turn a good decision into a defensible one. The notes are personal data about a living individual, and in most jurisdictions the individual has rights of access. The interview itself is voluntary, and the leaver can decline, decline parts, or ask for the notes to be amended. None of that's optional, and none of it can be improvised at the point of request.
Questions to Ask Before You Commit
Interviewer identity. Who runs the conversation, and are they outside the leaver's reporting line. A bad answer sounds like "the line manager, because they know the work best". That answer puts the interviewer in the subject of the finding.
Question set source. Where did the question set come from, and when was it last reviewed against actual leaver responses. A bad answer sounds like "it was in the template when I joined". That answer means the questions haven't been tested against the answers they produce.
Specific-moment coverage. What proportion of the questions ask for an incident rather than a feeling. A bad answer sounds like "we've a mix" without a number. The right answer is a count.
Note ownership. Who owns the summary, who owns the findings list, and who owns the response deadline. Three different roles, three different people. A bad answer sounds like "HR owns it", which means no one owns it.
Disclosure readiness. Can the leaver read the note if they ask. Will the note survive being read. A bad answer sounds like "we would have to check". The note is the leaver's personal data in most jurisdictions, and the position should already be confirmed.
Redaction practice. How are names handled when a specific moment identifies someone the leaver didn't mean to name. A bad answer sounds like "we use our judgement". The right answer is a written rule.
Cohort review cadence. How often are exit findings read across leavers, and by whom. A bad answer sounds like "when there's a pattern". Patterns are visible only after the review has happened, so the right answer includes a defined cadence.
Action loop. What is the documented path from a finding to a decision. A bad answer sounds like "we discuss it in the leadership review", which is a venue, not a path. The right answer includes a named owner and a deadline.
Redundancy separation. Are leavers from a redundancy programme run through the same interview as voluntary leavers. A bad answer sounds like "we use the same template". The right answer is that they're separated, because the answers aren't comparable.
Local advice. Where the position touches works councils, employee representatives, or local consultation rights, who confirms the framing. A bad answer sounds like "we've not needed to". The right answer is the name of the adviser and the date of the last review.
The Cost of Getting This Wrong
The cost of a bad exit interview doesn't appear on an invoice. It appears three quarters later, when the next senior product designer hands in her notice and the manager who paused the redesign in March is still in role, still pausing, still asking why good people keep leaving. The interview produced a finding. The finding had no owner. The owner is the cost. Not the finding itself.
So the second-order cost is the pattern that quietly establishes itself inside the organisation. The first exit interview produces a finding that doesn't land. The leaver moves on. The next leaver sits in the same chair and hears, from someone who has been there longer, that exit interviews are a tick-box. The second finding doesn't land either, but the second leaver also tells three people on the way out that the conversation was theatre. By the third leaver, the interviewer is collecting, not learning. By the fifth, the people doing the work have stopped believing that the work sees them at all.
What does the next round of exit interviews cost when the leaver arrives already rehearsed in the interview's emptiness. What does the next hiring round cost when the candidates the leaver tells about the exit process are the candidates you most wanted. What does the next reorg cost when the manager who should have been moved two years ago is still in role because no individual finding was ever aggregated into a pattern. The line items are small. The compound isn't.
When You Are Ready to Go Further
If the question set above reads like the conversation you've been trying to run, the next question is usually not another template. It's a way to check what you've built against what others in your sector have built, and a way to pressure-test the role of the exit interview against the rest of the listening you already do. HROpsLab is a review publication. We don't sell software, payroll, or survey tools, and we don't run exit interviews on your behalf. What we do is independent comparison work, written by people who have sat in the seat you're sitting in. If that's the next thing you need, the work is there.
Frequently Asked Questions
How many questions should an exit interview contain.
Aim for seven to ten, of which at least three ask for a specific incident or a counterfactual. A short, well-run interview produces more usable signal than a long, polite one. The first question should let the leaver set the frame. The middle questions should pull specific evidence. The last question should return control to the leaver. If you can't fit the question set inside thirty minutes, the leaver has already told you everything they intend to, and the rest of the time is a cost you're paying in goodwill you'll need for the next exit interview down the hall.
Do leavers actually tell the truth in exit interviews.
Most leavers aren't lying. They're managing. They're protecting the reference, the relationships they still need, and the story they want future employers to hear. The job of the question set is to design questions that make the truthful version easier to say than the safe one. A specific-moment question lets the leaver describe what happened without indicting anyone by name. A feelings question forces them to either indict or soften, and most choose to soften. The interviewer's job is to hold space for the specific answer and to write the note in the leaver's own words rather than in a paraphrase that flattens the signal.
When in the notice period should the exit interview run.
Week one captures the moment of leaving and the relief of having decided. Week three captures the handover and any bitterness that built up during it. Both windows produce different answers. Pick one, write the choice down, and tell the leaver which window they're in. The single biggest mistake is running the conversation on the leaver's last day, when the leaver is mentally already gone, the handover is incomplete, and the interviewer is under pressure to close the loop before the desk is cleared. If you can only run one interview, week one is the right default for most voluntary leavers.
Should the exit interview be recorded.
In most cases, no. A written summary in the leaver's own words, reviewed with the leaver before they leave, is more defensible and less likely to suppress candour. Recording changes the dynamic, often in ways the interviewer can't see. If recording is necessary for a specific reason, the leaver must consent, the recording must be stored with the same care as the notes, and the leaver must be told that the recording is personal data about them and is disclosable on request. Confirm the position locally before recording anything.
Should every leaver get the same questions.
Yes, with two exceptions. Leavers from a redundancy programme should not be folded into the standing exit interview, because their answers are about the programme, not the work, and the two signals contaminate each other. And if a leaver raises something outside the standard set that needs a follow-up, the follow-up question is allowed, but it should be recorded as such. Consistency is what lets you compare one cohort against another, and comparison is what lets you know whether the intervention you made after the last batch actually moved anything.
What do you do if a leaver declines the exit interview.
Thank them, confirm there's no consequence, and leave the door open. Exit interviews are voluntary in most organisations. The decision to decline is itself a finding, and it should be recorded, but it should not be reported in a way that lets anyone identify the leaver. Don't chase, don't reschedule three times, and don't ask the line manager to ask on your behalf. The leaver who declines this one is often the leaver whose friends will tell you whether the next round of exit interviews is worth taking seriously.
How do you ask about the line manager without leading the leaver.
Ask about specific behaviour rather than about the person. "What did your manager do in the last month that worked, and what didn't" pulls the conversation onto incidents. "What did you think of your manager" pulls it onto character. The specific-moment question also protects the leaver, because they can describe what happened without naming anyone, and the interviewer can decide whether to record the name in the note or to redact it. If a pattern emerges across leavers, that's a separate workstream, and it should not be run as part of any single exit conversation.
