TL;DR
- Core decision: Treat "I am leaving for the money" as a starting position, not a finding, and design the question so the rehearsal falls apart.
- When doing nothing is right: When the leaver is senior, the relationship matters more than the data, and you've no infrastructure to act on what they tell you.
- What has to be true: Someone in HR has to be willing to hear a real answer, including one that indicts a manager or a band the company won't move.
- How the options split: Direct amount, offer framing, timeline reframing, retention counterfactual, last internal pay conversation, or the deliberate silence.
- Decision rule: Pick the question that gives the leaver the lowest cost to tell the truth and the highest cost to perform a story.
- Outcome to expect: A signal that's either a benchmarking gap, a band problem, or a manager problem, and the clarity to tell which one it's.
The conversation that gives you nothing
The HR lead closes the door, opens the form, and gets the answer she has heard seven times this quarter. The senior product manager is leaving. He says the new package is the reason. He is careful not to give a number. He doesn't need to. She has the new role on LinkedIn already, and her own recruiter has been pricing replacements for three weeks.
She writes "compensation" in the reason box and moves on. That's the entire interview, in essence. The form goes into the file. The replacement is approved at the band ceiling. The next senior product manager leaves eleven months later and gives the same answer. The cycle closes on itself.
But the interview is doing something she has not noticed. He is doing what most people do in the last hour of a job they no longer want to discuss in detail. He is selecting the safest possible reason, because the reason needs to do two jobs at once. It has to be plausible, and it has to leave every relationship intact. Money is the only reason that does both without effort. So money is what gets said.
Best tools for Employee Retention
And the real reason, whatever it's, walks out of the room behind him, intact and unspoken. The job isn't to challenge the answer, which fails. The job is to find out what pay was standing in for. The real issue isn't the number on the offer. It's the gap between what the company thinks it pays and what the leaver thinks it's worth to stay.
When you genuinely do not need to act yet
Not every organisation has a pay question worth asking in the exit interview. Some setups are running well enough that the cost of a more aggressive question outweighs the marginal signal. A quick read of your own position comes first.
Your current setup is genuinely fine. A 60-person firm where the founder runs every exit conversation personally, knows each person's salary and tenure by heart, and adjusts bands within a week of losing anyone isn't going to learn anything from a structured exit interview that the founder doesn't already know. The founder is the data. In that kind of firm the interview is a relationship ritual, and treating it as a research instrument is a category error. Leave the form out of the room.
Mild friction, watch the trend. A 400-person company where pay exits have ticked up but where the leavers are spread across teams, levels and tenures has a signal but no pattern yet. At this scale the exit interview is still load-bearing for retention learning, but only if the answers are aggregated somewhere and reviewed monthly. Without aggregation, you've a pile of anecdotes. The honest move here's to fix the review process before refining the question. A better question asked of an answer that goes nowhere changes nothing.
Real risk, time to change the question. A 1,500-person firm where three senior engineers have left in six months, all citing compensation, and where the engineering director insists the team is paid at the top of band is in a different place. The form is being used to confirm a position the company already holds. At this scale the question itself has to change, because the question currently in use is producing the answer the director needs to hear. The interview has become a piece of internal theatre and is costing you information that has a market price.
Edge case, the senior individual contributor. When the leaver is the head of a function, the founder's chief of staff, or anyone whose departure will be discussed by name in the leadership team for the next quarter, the interview is no longer an HR instrument. It's a relationship instrument. Even the appearance of a structured question can be read as a request for ammunition by someone with political weight in the building. In this case, the conversation is best held by the person who will still be working with them in some capacity, with no form in front of them and no obligation to be quoted.
Five questions this reader asks at 11pm
Will asking about pay make the leaver clam up entirely? It can, if the question is the first one asked, if it's asked by someone they don't trust, or if the answer feels like it will be used against them. The reasoning is that a leaver has already rehearsed the answer on the way to the meeting, and any question that lets them restate the rehearsal is wasted airtime. If the answer is yes, you'll get a tidy exit and no information.
Should I ask what they're being paid at the new place? In some places the answer is straightforward, in others asking puts you inside pay-history rules you may not have meant to enter. The reasoning is that the law treats the new employer's offer as the leaver's own information to share, not yours to collect, and the obligation to disclose what you'll do with the number lands unevenly across jurisdictions. If you ask without thinking about where the data goes, you've created a record of your own collection that you may have to defend later.
What if the real reason is the manager and pay is the cover? Then asking about pay directly will produce a confident answer, and the question has done nothing. The reasoning is that no one under pressure names a manager in a setting where the manager is in the next building. Pay is the reason that lets the leaver leave the relationship intact. If you don't design for this, the interview will confirm the manager's view of themselves.
What if several people in the same team cite pay at the same time? That's a pattern, and a pattern is more valuable than any single answer. The reasoning is that one person may genuinely be leaving for money, two may be coincidence, three is a signal that something in the team or the band is off. If you treat each exit as independent, you'll keep re-litigating the band instead of seeing the team.
Is there a point where the answer changes what I do? Only if the answer is connected to a process that can act on it. The reasoning is that an interview that produces a finding the organisation won't move on isn't a learning instrument, it's a ritual, and rituals corrode trust. If nothing downstream changes, don't ask.
Three honest categories the approaches split into
Direct quantitative questions. This is asking the new salary, the new package, the headline figure, or the percentage uplift. The strength is that it produces a number you can put into a model. If your problem is that you're systematically below the market and you need to know by how much, the number is the answer. The failure mode is that you don't get a usable one. Leavers are coached by their new employer, by their own instincts and by their friends to be vague about money. You will get "north of" and "a significant uplift" more often than you get a number. The number is also collected in a context where, in some places, the new employer's offer is treated as the leaver's own information. You have created a record of asking, with no usable answer to show for it. This approach is right when you've a benchmarking problem and a process to compare the answer against your own bands. It fails when you don't.
Indirect framing questions. This is asking when they started looking, what they would have needed to stay, what the last pay conversation with their manager felt like, or what the new role offers that this one doesn't, with pay as one of several dimensions. The strength is that the leaver can't answer these in one sentence, and the rehearsal breaks down. They will end up talking about the new offer's flexibility, the new manager's reputation, the new company's stock plan, the commute, and somewhere in the middle, the pay. The failure mode is that you've now collected several data points from a single conversation and may overweight whichever one you wanted to hear. This approach is right when the issue is likely the manager or the band shape, and you've time in the interview for a real conversation. It fails when the interviewer is rushed, junior, or has a position to defend.
Deliberate silence on pay. This is asking about the work, the manager, the team, the growth path, the decision process, and not asking about pay at all. The strength is that you've removed the most rehearsed reason, which forces the leaver into the next-best answer. That answer is usually closer to the truth, because it isn't the one they prepared. The failure mode is that you'll miss the genuine pay departures entirely, the ones where the leaver would have told you straightforwardly, and you'll mistake those for something else. This approach is right when your biggest risk is a manager or culture problem you've not been able to see, and your interview has a long enough runway to let the conversation find its own shape. It fails when your retention problem is at the band level, where you have to ask to find out.
Five diagnostic questions to run against your own organisation
How many exit interviews did your last twelve leavers actually do? The way to answer this is to count the completed forms against the total leavers in the period, and to do it without rounding. If fewer than half of leavers are doing the interview, the interview isn't a research instrument, it's a sample. A sample can't tell you whether pay is the reason people leave. It can tell you whether the people who agree to talk to you cite pay. Those are different statements.
Where does the form go after it's filled in? The way to answer this is to follow the artefact, not the policy. Find a real exit interview from six months ago and see who has opened it since. If no one has, the data is being collected for compliance, not for learning, and refining the question is wasted effort until the loop is closed.
Who is allowed to change a band as a result of what is heard? The way to answer this is to ask your head of total rewards, in writing, what they would do if the next three senior leavers all said the same thing about the band for their role. If the answer is "we would review it", you've a process. If the answer is "we would need to see a business case", you've a process with a gate. If the answer is "we would need a board paper", you've a process with a gate and a queue. Only the first shape can act on what the interview produces.
What does your recruiter say the market is paying for your hardest-to-fill role? The way to answer this is to ask them, in the same conversation, what they would tell a hiring manager who said the budget was fixed. The answer will tell you whether your recruiter sees a gap or sees you as already at market. If they see a gap, your exit interviews are more likely to surface it. If they see you as at market, your exit interviews will tell you something else, and you should know which before you sit down with the next leaver.
When did your head of people last read an exit interview that changed a decision? The way to answer this is to ask them, and to listen for whether the answer is recent. If it isn't, the interview is a record of a conversation, not a driver of one. You can still ask, but you should ask knowing that the answer will land in a drawer.
Six Ways to Ask, and What Each One Surfaces
Asking the direct amount
This is the question most HR teams reach for first. It comes in several flavours: the headline number, the base, the on-target earnings, the package, or the percentage uplift. It's the most measurable question in HR, and that's part of the problem. It puts the leaver into a position where they're being asked to share a figure their new employer has marked as theirs to share or not. It also produces the answer the leaver has rehearsed, which is a version of the truth that lets them leave with no relationship to repair.
Where it earns a place is when your real problem is benchmarking and you've a defensible band the leaver is sitting inside. The number, if you get one, slots into a model. Where it falls short is when the issue is the band itself, or the manager, or the career path. A direct question about pay won't surface any of those, and will produce an answer that prevents them being surfaced by the next question.
Asking about the offer rather than the salary
This is the question that asks what the new package looks like in the round: base, bonus, equity, benefits, flexibility, location, title, reporting line. It treats pay as one element in a shape rather than as a number. The leaver will answer with a description, and the description will include pay and several other things. The other things are where the signal is.
Where it earns a place is when you suspect the gap isn't just a number. A person who can describe their offer in detail and has the new flexibility, the new manager, and the new title ready to discuss is rehearsed, and the interview won't break the rehearsal. A person who can describe the offer but keeps returning to one element, which is usually not the base, is telling you what the offer really represented for them.
Where it falls short is when your organisation can't act on anything except pay. If the only lever you can pull is the salary, a question that surfaces flexibility or title or commute will produce a finding you can't use. You will know more than you can do.
Asking when they started looking
This is a question about time, not money. It puts the leaver into a moment, which is harder to rehearse than a reason. The reason has been polished for weeks. The moment has a date, and the date usually lands in a window. The window is where the real signal is.
Where it earns a place is when your hypothesis is that something specific happened, a manager change, a reorg, a comp cycle, a promotion decision that didn't go the way the leaver wanted. The date will tell you whether your hypothesis is right. Where it falls short is when the leaver has been looking for a long time, in which case the date tells you nothing new. It also falls short when the interviewer is uncomfortable with silence, because this question is useless without a pause afterwards.
Asking what would have kept them
This is the question that asks the leaver to imagine the counterfactual. It's uncomfortable for both sides, because it invites the leaver to name a number that you might, in theory, match. In practice, the answer is almost never a number. It's a set of conditions, and the conditions are where the signal lives. Most leavers won't give a clean answer to this. The ones that do are either bluffing or telling you exactly what they need.
Where it earns a place is when you've not yet decided whether to make a counteroffer, and the conversation is happening before the resignation is final. Where it falls short is when the resignation is already accepted and the leaver has emotionally moved on. Asking the counterfactual question at that stage produces either a polite non-answer or a list of conditions that would have required a different organisation.
Asking about the last pay conversation they had internally
This is a question about process, not outcome. It asks how the leaver's last salary review, promotion discussion, or band conversation went. The answer is usually more honest than the answer to "why are you leaving", because the process question is harder to defend against. The leaver can't easily say the conversation went well if it didn't, without sounding evasive.
Where it earns a place is when the manager is your suspected cause. The leaver may not name the manager. They will describe the conversation. Where it falls short is when the leaver's manager is also their interviewer. That's a setup error, not a question error.
Not asking about pay at all
This is the deliberate silence. The interview covers the work, the team, the growth, the decision process. Pay is never named. The leaver, asked nothing about money, will usually introduce it themselves, because they have rehearsed the answer and need to deliver it. The moment they deliver it, you can see how it sits with the rest of the story. If it sits on top, the leaver really is leaving for the money. If it sits underneath, the leaver is leaving for something else and using the money as the lid.
Where it earns a place is when your interview has the time and the interviewer to carry a long conversation. Where it falls short is when the interview is a fifteen-minute slot run by a junior HR generalist with a form to complete. In that setting the silence produces an awkward conversation and no usable finding.
The Decision Table
| Situation | Scale | Setup | Primary Pain | Recommended Starting Point |
|---|---|---|---|---|
| Founder-led, small team | Under 100 | Founder runs every exit personally | Ritual, not data | Leave the form out of the room |
| Pattern emerging, no process | 200 to 500 | Forms collected, never reviewed | Interview is decorative | Fix the review loop before refining the question |
| Senior leaver, politically sensitive | Any | Head of function, chief of staff, named leaver | Relationship matters more than signal | Hand off to the relationship owner, drop the form |
| Manager suspected, weak evidence | 500 to 1,500 | Three senior exits, all citing comp | Manager not named in any interview | Ask about the last internal pay conversation, not the new offer |
| Genuine band problem | 1,000 plus | Recruiter confirms gap, multiple roles affected | Band, not manager | Ask about the offer in the round, compare to your own band for that role |
| Counteroffer window still open | Any | Resignation fresh, no decision made | You do not yet know what would have kept them | Ask the counterfactual, in private, before the resignation is final |
| Exit volume high, leaver pool shallow | 1,500 plus | Monthly exits, mostly voluntary | Volume hides signal | Aggregate, do not interview; the unit of analysis is the month, not the person |
| Regulated pay-history environment | Any | Local rules on what you may ask | Asking creates a record you may have to defend | Describe the shape of the exposure, confirm locally before asking the direct amount |
Reading the Answer
A genuine pay departure has a shape. The leaver can name a number, or a range, and they can describe what the new role pays that this one doesn't, in language that sounds like they have done the comparison. They will tell you when they started the search. They will name a figure that's consistent with the role they're going to. They won't be defensive about pay. Pay is the reason they're giving you because it's the reason. There's no performance to put on. The interview lands cleanly.
Money-as-tiebreaker has a different shape. The leaver is vague about the number. They describe the new role in terms that exceed pay. They talk about the new manager, the new flexibility, the new title, the new location, the new mission. If you ask when they started looking, the date usually lands just before or just after a specific internal event: a reorg, a promotion that didn't happen, a manager change, a comp cycle that disappointed them. If you ask what would have kept them, they name conditions that would have required the company to be a different shape, not just to pay more. Pay is in the answer, but it isn't on top.
The signal below is the working version of how to tell which one you've.
| Signal | Likely meaning | What to check next |
|---|---|---|
| Clean number, single comparator role | Genuine pay departure | Pull your band for the role, compare to recruiter's read of the market |
| Vague number, several dimensions named | Tiebreaker, not primary | Ask about the last internal pay conversation and the date the search started |
| Date of search lines up with an internal event | Manager or process, not pay | Pull the manager's other team exits in the last eighteen months |
| Conditions named are not pay-related | Pay is the lid, not the floor | Cross-check with the manager's last three direct reports' exit reasons |
| Number is well above your band ceiling | Market shift, not a single decision | Confirm locally before recording the figure, band shape may need to move |
The next decision is which of three problems you're looking at.
A benchmarking problem is one you can fix. The data exists, the band exists, the recruiter knows the market. The question is whether the band moves and whether the business case for moving it can be made before the next exit.
A band problem is harder. The band itself is the issue, not just where the leaver sat inside it. A band problem means the structure of the role doesn't match what the market now pays for it. Fixing it requires a re-grading, which has a queue.
A manager problem is the hardest. The data is anecdotal until it isn't, the manager will defend the position, and the organisation will need to decide whether to act on a pattern that's now visible. None of the three is a question for the exit interview to answer. All three are decisions for the people who own the structure.
What to Put in Writing
The conversation is the soft asset. The artefacts are the hard asset, and they're what turn a good decision into a defensible one. The list below covers the documents, records and notes this process needs to leave behind, who owns each one and what they prevent.
| Artefact | Who owns it | When it is written | What it prevents |
|---|---|---|---|
| Exit interview form or note | HR lead or interviewer | During or within 24 hours of the conversation | Loss of the leaver's own words before they are smoothed out by the summary |
| Aggregated exit data, monthly | People analytics or HR ops | End of each month, across all exits | Single anecdotes being mistaken for a pattern |
| Band-to-market comparison, by role | Head of total rewards | Quarterly, or after each cluster of exits | A genuine pay departure being misread as a manager problem |
| Re-grading business case, where relevant | Hiring manager with HR business partner | When the data supports one | A band problem being resolved by individual exceptions that erode the structure |
| Manager-pattern review, anonymised | HR lead with skip-level | When three or more exits cluster on one manager | A manager problem being visible only to the people closest to it |
| Local compliance check on what may be asked | HR lead with legal | Before the question template is finalised | An interview that creates a record the organisation cannot defend |
| Communication to hiring manager on reason for departure | HR business partner | Within one week of the conversation | The manager forming a private narrative that diverges from the leaver's |
| Decision log on what was acted on and what was not | HR lead | Quarterly, in the people committee paper | A process that collects data without closing the loop, which is the slow erosion of the instrument |
The single most skipped artefact on this list is the last one. The decision log is what tells the next HR lead what this one knew and what they did about it. Without it, the cycle restarts every time the function changes hands.
Questions to Ask Before You Commit
Process. Who in the business will see this finding and what are they authorised to do with it. A bad answer is "we will share it with the leadership team" with no description of what the leadership team can change. If the answer is "we will share it" without a closure mechanism, the interview is decoration.
Compliance. What does local practice allow me to ask about the new offer, and where does that practice sit in writing. A bad answer is "we've always asked". If no one in the room can name the source of that practice, you're operating on inheritance, not policy.
Data flow. Where does the completed form go, who opens it next, and what triggers a review. A bad answer is "it goes into the personnel file". If the personnel file is the final destination, no one is reading it, and the form is being kept for compliance, not learning.
Action. What is the smallest change the data could justify, and who has authority to make it. A bad answer is "we would need to see a pattern". A pattern is several exits, and by the time you've several exits the cost has already compounded.
Ownership. Who owns the interview programme, and is that person the same person who would need to act on what it produces. A bad answer is "HR" without a name. If the owner can't act, the data is going to a different room.
Volume. How many exits are we running through this process each quarter, and is the volume high enough to aggregate without over-fitting. A bad answer is "we've one or two a quarter". At that volume the unit of analysis is the individual, not the cohort, and the question has to be redesigned.
Risk. What would it look like if a leaver asked to see what we recorded about them, and what would we hand over. A bad answer is "we would have to check". If you don't know in advance, you'll find out at the moment of a request, which is the worst time to find out.
Tooling. What system holds these notes, who can search them, and how long are they kept. A bad answer is "our HRIS". The HRIS may be the right system, but the answer needs to be specific about retention, access, and the difference between the personnel file and the working file.
The cost of getting this wrong
The cost of a bad exit interview isn't the interview. It's what the interview prevents you from seeing, and what you don't see you can't fix. A leaver who leaves for the manager, citing pay, takes with them the only signal you were going to get. That signal, uncollected, doesn't return as a survey result or a stay-interview finding. It returns as the next resignation, from a different person, for the same reason.
So the second-order cost is structural. Each unanswered exit is a band or a manager that the company continues to fund. A band problem compounds at the speed of attrition. A manager problem compounds at the speed of that manager's team. Neither shows up on a single invoice. Both show up in the next quarter's retention numbers, which by then are someone else's problem to explain.
The reputational cost is slower and harder. An organisation that asks about pay and does nothing with the answer is teaching its remaining employees that the answer doesn't matter. Over time, the people who stay stop volunteering information in every internal channel, not just the exit interview. The instrument that was supposed to learn is now teaching the opposite lesson.
So the question isn't whether to ask about pay in the exit interview. It's whether you can act on what you hear, and whether the people who need to act are in the room when the data lands.
When you are ready to go further
If the shape of your exit interview programme has started to come into focus, the next move is to compare what you have to what other organisations in your size band have built. HROpsLab does independent comparison work across the HR operations stack, including exit interview tooling, total rewards benchmarking and HRIS configuration. We are a review publication, not a vendor, and we sell nothing. The comparison is the product, and it's free to read.
If your team would value a working session on the question design itself, the structure of the artefacts, or the hand-off between exit interview and band review, the contact route is below.
Frequently Asked Questions
Should I ask what they are being paid elsewhere?
It depends on where the data goes and what local practice allows. Asking creates a record of asking, and the obligation to handle the answer correctly lands unevenly across jurisdictions. In some places the question is uncontroversial. In others, the new employer's offer is treated as the leaver's own information. The shape of the exposure is the part you can describe. Confirm locally before you add the question to the template.
Is a counteroffer ever right?
Occasionally, when the leaver has named a specific gap that the band can close without distorting the rest of the structure, and when the leaver has not yet emotionally left. Most counteroffers fail because they address the stated reason, which is often not the real one, and the leaver has already decided. The instrument for testing whether a counteroffer is right is the counterfactual question, asked before the resignation is final, in private, by the person who will still be working with them.
How do I tell real pay departures from excuses?
A genuine pay departure has a clean number and a clean comparator role. The leaver isn't defensive about pay, because pay is the reason. A money-as-tiebreaker has a vague number, several dimensions named, and a search date that lines up with an internal event. The interview lands cleanly in the first case and lands with something still in the room in the second. Trust the shape, not the words.
What do I do when several people in the same team cite pay?
Stop interviewing and start aggregating. Three exits from one team citing pay isn't three independent findings. It's one finding with a team attached. Pull the manager's last eighteen months of exits, pull the team's last comp cycle, and pull the band the team sits in. The signal will be in the overlap.
Should I share the finding with finance?
Yes, but in a form finance can act on. "Three senior engineers have left citing pay" is a story. "Three senior engineers at band X have left for roles paying between Y and Z, against a band ceiling of W" is a case. Finance can act on the second. They will file the first under retention, which isn't their problem.
How do I ask without it becoming a negotiation?
By making clear, before the question, that the resignation is accepted and the conversation is for learning, not for retention. The moment the leaver thinks the answer can move their exit, the answer becomes a position. The interview has to be held after the decision is final, by someone who isn't the hiring manager, and with no downstream implication attached to the number.
What if the number is far above your band?
That's a market signal, not an individual complaint. Treat it as one. Pull the band, pull the recruiter's read of the market, and ask whether the band still describes the role or describes a role the company used to need. The answer will tell you whether you've a single leaver with an unusual offer or a band that has fallen behind. The first is a conversation. The second is a re-grading.
What if the leaver will not give a number?
Don't push. The number is the least informative part of the answer. The shape of the offer, the dimensions named, the date of the search, and the conditions that would have kept them are all more useful, and they don't require a figure. The figure is a benchmarking input. The shape is a diagnostic input. The diagnostic is what you need.
HROpsLab is an independent review publication. We don't sell software, payroll, or advice. We write so HR leaders can make better decisions, and we mean it.
