Most leaders find out too late what their people already knew
I am an operator, writer, and I work with a small number of leadership teams a year. One quarter: I measure two things not currently measured, help redesign the practices producing them, and measure again at ninety days so you can see whether it moved.
Who this is for
- Founder-CEOs and COOs of 150 to 1,000 person manufacturers and engineering firms.
- Private equity operating partners, at the portfolio company level, usually in the first year after close.
It works where the leadership team is small enough to sit in one room, and where somebody at the top has already noticed something is wrong and cannot name it.
What gets measured
Two numbers. Both exist in every organization and neither appears on any report.
Differential turnover
Every company knows how many people left. Almost none know which ones.
Not your attrition rate. The spread between the group leaving fastest and everyone else, priced in dollars against your own replacement costs. Aggregate attrition averages that spread away, which is why a company can hit its target rate while quietly losing the same kind of person every time.
For scale: a thousand professional employees with a ten-point spread runs to roughly twenty-five departures a year above baseline. At a replacement cost of about $220,000 per professional employee, that is around $5.5 million a year, absorbed as ordinary attrition. The replacement figure is a working assumption built from the published SHRM and Gallup bands, not a measurement, and the calculator lets you substitute your own.
Signal loss
Someone already knows. The question is whether it changes anything.
Accurate concerns from the people closest to the work that never reach a decision. The launch that slipped. The cost cut that hollowed out a function. The manager hitting the numbers while the team stops speaking up. Measured as a chain, because a concern only has to fail at one point to never reach you.
The objection, before you make it
The usual answer to signal loss is that the person should have made the case better. Sometimes that is true, and the people who manage it do get heard. It is still the wrong fix, because it is an individual remedy for a structural problem: it does nothing for the next person, and it means your access to accurate information depends on each employee's ability to argue rather than on anything you built. Every organization already has the people who can argue well. The question is what happens to what everyone else knows.
How it runs
One quarter, start to finish. Measure first, we find the three or four practices producing both numbers, your team does the implementing, and then the same instruments run again on the same people at ninety days.
That last part is the key point and it is the part most work like this skips. A diagnosis with no second reading is an opinion. If a number has not moved, you will see that it has not moved.
Collection is anonymous, reported in aggregate with a minimum group size, and nothing produced measures or ranks an individual.
Why me
Four years in the Navy. I did not lose one person.
That is less impressive than it sounds until you know what I did not have. I could not pay anyone more. No equity, no title bump, no counteroffer, no retention bonus. The Navy set all of it and I had no say in any of it.
So the only variable I controlled was what it felt like to work there. The most junior sailor got the same respect as the Captain, and I was consistent about that in the places where consistency was inconvenient. One division, four years. I am not going to tell you it generalizes.
Thirty-seven years in leadership positions, including two Fortune 100 companies and two companies I founded and ran. Retention across these organizations near ninety-five percent.
I built these systems by accident, because I could not read implicit ones. Explicit systems turned out to work better for almost everyone. That is the argument of The Perception Revolution, and the second book, The Projection Fallacy, publishes in April 2027.
You have every lever I did not have, and most of them are aimed at a population you have not identified.
What this is not
- Not an engagement survey. Keep the one you have. It is built to read the average, and both of these numbers live in the spread.
- Not a culture program. Every fix is a written practice you can audit.
Starting a conversation
The first call is thirty minutes and is worth having even if nothing follows. Bring your last twelve months of separations by team and tenure, and you will leave knowing whether you have a differential turnover problem, which most companies have never checked.
Peter Allen Mann is the author of The Perception Revolution (Fast Company Press) and The Projection Fallacy, forthcoming April 2027. He spent thirty-seven years in leadership positions across the Navy, two Fortune 100 companies, and two companies he founded. He writes at peterallenmann.com.