The exit conversation almost never tells you. People leaving are managing a reference and a last two weeks, so you get a version: better opportunity, bit more money, closer to home. Some of that's true. It's rarely the reason.
So don't start with the exits. Start with the pattern.
Look at where they leave from, not why
Map the last several departures by seat, not by person.
If they cluster under one manager, you have a management problem and you almost certainly already knew which manager. If they cluster in one function, the job design is wrong there. If they cluster at a tenure mark, eight months, eighteen months, that's a ceiling, and the ceiling is usually real.
If they're spread evenly across everything, then it's the company, which most often means it's you. That's the least comfortable finding and the most actionable.
The four reasons, in the order they actually occur
The job wasn't the job. They accepted a role and got the pile. This is the number one cause in owner-led companies and it's almost always accidental. The posting described the work you needed most that week. The actual work was everything nobody else would pick up. Nobody lied. But the person did the math at month four and concluded the job they interviewed for doesn't exist.
There's nowhere to go and nobody said so. A capable person will stay a long time in a role with no promotion available if you're honest about it and pay them accordingly. They will not stay if they're waiting for something you're implying and can't deliver. Vagueness about the future reads as a promise.
Their decisions don't stick. They make a call, it gets overruled, and after the fourth time they understand their authority is provisional. Strong people leave over this faster than over money, and they never cite it on the way out, because there's no polite phrasing. See why your managers won't decide without you.
They were managed by someone with no time to manage. Usually you. Not from indifference, from load. A manager who's also carrying a full delivery job manages in the gaps, and the gaps go to whatever is loudest. Good performers get the least attention, because they're not the fire.
The signal that shows up months early
Watch for the shift from asking to informing.
Early on, a good employee asks. What do you think about this, should we do it this way, is that the right call. They're building a model of how you think.
Then at some point it changes. They tell you what they did. Then they tell you less. Then they stop bringing problems, and the problems arrive later, from someone else, already expensive.
That progression isn't disengagement in the way owners read it. It's a person concluding that the asking isn't worth what it costs. By the time it completes, they're interviewing, and you have maybe a month.
The retention question that works
Skip the annual review. Ask this, and ask it when nothing is going wrong, because that's the only time you'll get a real answer.
What's the most frustrating part of your week that I probably don't know about?
Then say nothing. Let it be awkward. The first answer will be polite, and the second answer, the one that comes after the silence, is the useful one.
Most owners have never asked this of anyone, and the reason is straightforward: the answer usually implicates something the owner is doing, and everyone in the room knows it. Which is also why asking it counts for so much. See why your team won't tell you what's wrong.
What turnover is really costing you
Owners price a departure at recruiting plus a few weeks of lost output. That number is far too low.
The real cost is the ramp, and the ramp is long precisely because the work isn't specified. A new person takes months to be useful, and during those months your strongest remaining people absorb the gap, which is how one departure becomes the cause of the next one. That's the part that compounds.
If your onboarding takes five months, turnover is nearly the most expensive thing happening in your business, and the fix isn't retention tactics. It's writing the job down so the ramp is weeks. See signs your business has outgrown its systems.
Where this sits in the data
People is the constraint owners name most often in the Platform One Index, 46 percent, more than marketing and finance combined. It also scores worst of the five areas measured, 32.9 out of 100.
What the owner sees
- Approvals move same day
- Everyone knows who owns what
- The process is documented
- People speak up when something's off
the gap
What the team operates
- Sits a day in drafts while we word it right
- Two of us both think it's ours, so neither starts
- Nobody's opened that doc since onboarding
- We stopped raising it in 2024
Both accounts are honest. Neither one is complete.
The recurring shape in what owners write in their own words is worth sitting with: they know the answer is to hand work off, and they either can't afford to or don't trust anyone with it. A company in that position generates turnover structurally, because the work stays undefined and the authority stays centralized, and those are the two conditions capable people leave over.
Start with the measurement
Twenty questions, about ten minutes, free.
You'll get scored in five areas and see which one your answers point at. If you've lost more than one person this year, ten minutes is a cheap way to find out whether you're looking at a people problem or the thing underneath it.