Two records exist for the same departure.
The first is an exit interview. A senior engineer, with four years on the job, talks to HR. She says she’s found a role with more scope. She's warm about it. She thanks the company. The form is filed as voluntary and non-regretted. The reason listed is career growth.
The second record is in your HRIS. It shows she was the fourth person to leave that manager's team in fourteen months, out of eleven. It shows the previous three all had above-median tenure. The group she belonged to has been separating at almost twice the rate of the highest-retained group in the company for two years.
Both records describe the same event. Only one of them is true, and it isn't the one that anybody reads.
The instrument is asking the wrong person at the wrong time
Exit interviews fail due to structural issues, not procedural ones. That’s why better questions won’t solve the problem.
You're asking someone to explain why they're leaving on a form with their name. This happens when they're trying to move on. Every incentive suggests a clear, positive direction: they are seeking new opportunities. Better compensation. Time for a change.
I've written those. When you leave a place that’s worn you down, don’t write it down on your way out. You are polite, you shake hands, and you leave with your reference intact. Everyone in the room understands the transaction, including the person conducting the interview.
This is not an HR failure. The people running exit interviews are using the instrument they received. The instrument needs to gather information that only one person has. That person has good reasons to keep it hidden.
There is a harder version of the point. If you first learn about a problem during the exit interview, the problem was really never hidden. Everyone below a certain level saw it for months. You created an organization where no one could speak up.
What the HRIS Records Instead
Your HRIS has no incentive problem. It doesn’t rely on anyone's honesty. It’s not gathered when people feel most defensive. Instead, it shows what happened, not what someone chose to share.
It also has data that most organizations have never queried in the way that matters. Separations by date, by function, by level, by tenure band, by manager, by demographic group. Most HRIS systems have the necessary dimensions for this analysis, though they may vary.
People often ignore turnover reports. They focus on board questions, not daily tasks. The board asks whether attrition is in line with the industry. Twelve percent, in line; next slide. That number is an average, and an average is a device for making a distribution disappear.
Cut by Manager First
Start with the cut that produces the longest silence in the room.
Company-wide averages can mask individual managers. For instance, a 3.8 average psychological safety score might cover a 2.6 for one team and a 4.2 for another. Run separations by manager over two years, adjusted for team size. You'll usually find that a few managers cause most voluntary departures. Surprisingly, no one has seen this on paper.
Then cut by group. Function, level, tenure band, and whatever demographic dimensions your data supports. What you're looking for is not the average. It's the spread between your highest retained group and your lowest.
That spread is the finding. In many organizations, retention rates may be about eight percentage points higher. One part of your company has seen better results due to your compensation, systems, and leadership. It isn't a hypothetical target. It's a result you're already producing somewhere.
Everything below the top group is the gap between what you're getting and what you've proven you can get.
Why Twelve Months Isn't Long Enough
Most turnover reporting runs on a rolling twelve-month window. That window is too short for this analysis for three reasons.
Small groups don't produce readable numbers in a year. A forty-person function might have five departures in twelve months. The difference between five and eight is noise at that scale. In twenty-four months, you can spot a pattern from a bad quarter.
One reorganization swamps the signal. Restructuring, closing a site, or changing leadership can overshadow a year's data. This makes the underlying rate hard to read. Over two years, the one-time events become visible as one-time events rather than as the trend.
A manager's effect takes longer than a year to appear. A person who starts under a new manager and leaves because of them usually doesn't quit in the sixth month. The sequence lasts longer than that. It starts with enthusiasm, then fades. The employee is labeled as difficult and disappears around month eighteen. A twelve-month window halves that sequence. It links the departure to events at the end.
A manager effect and a group effect can be separated from normal churn in at least two years.
Translating It into a Number Your CFO Will Engage With
The analysis matters only if it creates a figure that grabs attention alongside other finance topics. It does if done properly.
Take a replacement cost. Estimates suggest that entry-level roles cost 30 to 50 percent of salary. For professional roles, it’s 100 to 150 percent. Specialized positions can be even higher. Gallup says the range is half to twice the annual salary. They call this conservative. Pick a figure your finance team will defend, and use your own salary data rather than a benchmark.
For a professional workforce, two times a $90,000 salary equals about $180,000 per departure. That number is an assumption, not a measurement, and you should replace it with your own.
Now multiply. In a company of a thousand, a ten-point spread in retention can lead to 25 extra departures each year. Compare this as if all groups retained employees at the same rate as your best group. At $220,000, that means $5.5 million each year for replacement costs. This doesn't even include the knowledge that walks out your door when you lose good people.
That figure is not on any report you currently receive. An aggregate absorbs it, reading as normal, and it recurs every year until something changes. Or someone is willing to dig deeper.
The scale is consistent with what the external research finds. MIT Sloan's 2022 study of 34 million employee profiles found that toxic workplace culture predicts turnover 10.4 times better than pay. SHRM's 2019 work put US organizational spend on turnover at up to $223 billion over five years. Those are national figures and they don't tell you anything about your company. Your HRIS does.
What to Do This Week
You don't need the full analysis to start. You need one query run against data you already own.
What is the gap between my highest and lowest retained groups over the last twenty-four months?
If it's two points, your problem is smaller than most companies. So, verify the data before trusting it. If it's eight, you have the standard version, and you now know approximately what it costs.
You’ll learn something that your turnover rate can’t show you. Plus, you’ll have this insight before your next board meeting, not after another resignation.
Run your differential turnover number → calculator
Enter your headcount, your replacement cost assumption, and your spread. It returns the annual figure and a one-page summary for your CFO.
If you'd rather start with the measurement side, the six metrics that make this visible covers what to instrument and how to read it.