Why exit surveys never tell you why people left

If you want to know why people leave a company, the exit survey is the worst place to look.

That sounds like an exaggeration. I don't think it is. I've done a few hundred exit interviews, and the forms almost always say the same things. Family circumstances. A new industry. A bigger platform.

None of it is false. None of it is true either. The people filling in the form aren't lying. They're picking the answer that costs them nothing.

The numbers first

Japan's labor ministry asks job changers every year why they left their last employer. In the 2024 survey the top reason for men was other personal reasons, 20.2%. Same for women, 24.3%. Low pay came further down, 10.1% for men. Workplace relationships, 9.0% for men and 11.7% for women.

What are other personal reasons? The survey doesn't ask. It's a box that holds anything.

A firm that specializes in surveying leavers asked 574 people who had quit within the past year, in early 2026. 88% said there was a real reason they never told the company. The reason most often given to the company was family circumstances, 48%. The share who said that was the honest reason: 13%. A 35-point gap.

So the survey does measure something. It measures which reason felt safest to say on the way out.

Why they don't match: the survey asks one layer, people leave on another

I look at a person in four layers. The innermost is the core: what they want, what they'll never do. The second is patterns: who they become under pressure. The third is equipment: skills, experience, track record, the part you can develop. The outermost is context: whether this person plus this company, this boss, this team adds up or grinds.

Exit surveys ask what the outer two layers can answer. Pay, job content, hours, family. All of it fits on a form and offends nobody.

The reason someone actually leaves usually sits in the inner two layers, or on the one point in the context layer that's hardest to say out loud.

Leaving at the core layer means the job can't give them their top two drivers. Someone who needs to decide things, sitting in a seat where everything goes up for approval. Someone who wants to master one thing, pulled into firefighting every day. They won't write "my drivers weren't met" on the form. They'll write "want to try another industry."

Leaving at the patterns layer is harder to see. In good times the boss and the employee are both fine. Then the company gets nervous, the boss tightens control and starts checking everything, the employee's avoidance pattern kicks in and they go quiet. Two pressure patterns collided. The form says family circumstances.

Leaving at the context layer is the most common and the hardest to write down: they can't work with their direct manager. Saying so has a price. The manager stays at the company, and the employee stays in the industry. So 88% say nothing.

Why money never comes first

In the ministry's data, low pay is about one in ten. In leavers' own honest rankings, money isn't first either.

That matches my experience. I wrote about an engineer who resigned, was offered a 50% raise, said only "why didn't you do this earlier," and left anyway (A 50% raise, and he still left). Money lives in the equipment layer. It can buy a few more months of someone waiting to see, but it doesn't fill a gap in the core. For someone who has already left at the core layer, a raise just moves the resignation date.

It works the other way too. Pay was fine, the person left anyway, and the boss decides young people are never satisfied. That's an easy conclusion, and usually a wrong one. He's looking for the answer in the wrong layer.

So how should an exit interview go?

I rarely ask why someone is leaving. The answer is already on the form, and asking again gets you a more polite version.

I ask about the event. The day you finally decided, what happened? The answer is nearly always a specific scene. One sentence in a meeting. A proposal that got killed. A promise that wasn't kept. Events are more honest than reasons, because an event can't be repackaged as family circumstances.

One interview I remember clearly. The form said family circumstances. I asked what happened on the last day. There was a pause, and then a story about a clash with the manager. Going back, the same clash had happened several times, and each time the person had let it go. That day they didn't. The real reason was in the context layer and had nothing to do with family. It wasn't written down because the manager was still at the company.

Then I put the event back into the four layers. Did that sentence hit their drivers, or was it a pattern collision with the boss, or plain context mismatch? Which layer it lands on decides what the company should fix. A core-layer loss means the hire was wrong from the start, so fix hiring. A patterns-layer loss means looking at what that manager becomes under pressure, which is usually a person problem, not a policy problem. A context-layer loss can sometimes be solved by moving someone to another team.

There's one more method: line up the real reasons behind their last three job changes. What people say and how they vote with their feet reconcile once you look at the long run. Someone who left three times over a boss, and someone who moved three times for a raise, are probably leaving on the same line this time. It's the same approach I use when collecting evidence in an interview.

Japan adds one more layer: honne and tatemae

This is harder in Japan than elsewhere, because honne and tatemae, what people really think and what they say in public, are two separate registers to begin with. What's said in an exit interview is tatemae by default. I made the same point writing about reference checks: a referee on the phone won't tell you they'd never work with the person again.

So in a Japanese exit interview, don't expect to hear the honne. What you can do is ask about events and read the layers yourself.

When it gets harder still to say, some people skip the conversation entirely and use a resignation proxy. I looked at those numbers in Resignation proxies in Japan: is it 2% or nearly half?

Even when you learn the real reason, there isn't much the company can change. Context-layer and equipment-layer problems are fixable. Core-layer and patterns-layer problems mean changing a person, not a policy. A manager who tightens control under pressure is something HR can do very little about. Run the best exit interview in the world, and if the real reason ends up in a report nobody reads, it's no different from the family circumstances on the form.

The last time you left a job, was the reason you gave the company the real one?