Two years of data, one question
Think of the manager whose team you would least like to rebuild. Ask what that team's voluntary attrition rate was last year against the company's, and then what it was the year before. Most reporting packs cannot answer that, because they were built to produce one rate for everybody.
That rate is a single number describing the whole organization. It is also the only turnover number most boards have ever been shown, and one year of it cannot distinguish a pattern from a bad quarter.
What twelve months hides
Twelve months of attrition data tells you what happened. Twenty-four months tells you whether it is a pattern or a bad quarter. That is the whole argument, and it is the reason the standard reporting window is the wrong one.
Inside one year, a reorganization, a funding round, a bad hire in a key management seat and an ordinary run of retirements all compress into the same line. The line moves and the explanation is chosen afterward, usually by whoever has the most confident account of the quarter. Two years separates the one-off from the persistent, because a one-off appears once.
The data already exists. Every HRIS sold in the last decade stores termination date, termination type, department, manager, tenure and level. The twenty-four-month export is a request, not a project.
The one cut
One question resolves more turnover arguments than any other: which group is leaving fastest, and by how many points?
Pick the grouping that matters in your organization — department, manager, level, tenure band, location. Calculate the voluntary attrition rate inside each group across twenty-four months. Then rank them and read the spread between the top group and the company average.
The spread is the finding. A company at eleven percent overall, with one function at nineteen and the rest at nine, does not have an eleven percent problem. It has a ten-point gap sitting inside a number that matched the benchmark, which is why the number never raised anything. You can price that gap on data you already hold with the differential turnover calculator.
Why not just read the exit interviews
Because the exit interview measures the exit interview. Work Institute's own exit-interview data, published in its 2026 Retention Report, puts Management at 9.2 percent of all stated reasons for leaving in 2025, behind Career, Work-Life Balance, Health and Family, and Relocation. Read at face value, that says management accounts for roughly one departure reason in eleven.
Sit with that for a second. The largest commercial exit-interview dataset in the United States, asked why people leave, returns management as a one-in-eleven answer. Anyone who has run an organization knows the honest figure is higher, and the gap between those two numbers is the measurement of the instrument rather than a measurement of why anyone left. The report is also a commercial publication rather than peer-reviewed work, it does not state its 2025 sample size anywhere, and "preventable" in it is the publisher's own classification of its own categories.
So the useful sentence is not that exit interviews lie. It is that an exit interview collects what a departing employee is willing to put on the record with someone who can affect a reference. The termination data collects what happened. Only one of those two sources has no incentive attached.
Run it yourself
- Export every termination for twenty-four months: date, type, department, manager, level, tenure, and whether it was voluntary.
- Calculate voluntary attrition inside each grouping, annualized, for each of the two years separately.
- Rank the groups and read the spread. One number.
- Check whether the top group held its position across both years. If it did, it is a pattern.
An analyst can do all four in an afternoon. The export request is the slow step, and it is slow for calendar reasons.
It is also the cut that names departments and managers, so decide before you run it who sees the output. A finding that arrives in the wrong inbox stops being a diagnostic and becomes an accusation, and the second one does not get run twice.
I'll be straight about where this stops. A third to a half of professional departures were never addressable: people move cities, follow a spouse, retire, take an offer no one was going to match. From the outside I cannot tell you which of yours sat in which half, and neither can anyone else. The spread narrows the question. It does not answer it.
Two years instead of one, and one spread instead of one rate. Most finance functions have never seen either, and the spread is the number that survives a budget meeting because it points at a population rather than at a mood.
What is your voluntary attrition rate inside your fastest-leaving department, across two years rather than one?
---
Forward this to whoever owns your HRIS reporting. The full twenty-four-month procedure is set out in this piece on reading HRIS turnover data, and the structural change that quietly produces some of the spread is this week's companion piece.
Sources
- Work Institute. (2026). 2026 Retention Report: Employee Turnover Insights and Trends in 2025, tenth edition. Management at 9.2 percent of stated reasons for leaving in 2025; reasons derived from the publisher's own exit-interview data. A commercial publication rather than peer-reviewed work; the 2025 sample size is not stated in the report. Work Institute, 2026 Retention Report
- Bureau of Labor Statistics. Job Openings and Labor Turnover Survey, August 2026 data, released 29 September 2026. Quits 3.1 million, a 1.9 percent rate; total separations 5.1 million. BLS, Job Openings and Labor Turnover Survey
- SHRM. Replacement cost bands: 30–50 percent of salary for entry-level roles, 100–150 percent for professional roles, up to 200 percent for specialized roles. SHRM
- Mann, P. A. (2026). The Perception Revolution. Fast Company Press. Chapter 2, on the exodus, exit interviews and what a departure costs.
Peter Allen Mann is the author of The Perception Revolution (Fast Company Press, 2026). He spent thirty-seven years in leadership, including four years as a Navy Surface Warfare Officer and two companies founded and run. He writes about late-recognized autism and perception at peterallenmann.com.