Somebody in your finance function has a multiplier. It is probably 1.5. It is probably applied to every departure in the model, and there is a reasonable chance no one in the building knows where it came from.
That number is doing real work. It sizes the retention budget, it decides whether a counteroffer gets approved, and it is the figure that appears on the slide when someone asks what attrition is costing. It is worth knowing what it is made of.
The published numbers disagree, and the disagreement is the useful part
Three sources get cited for this, and they do not agree.
The Center for American Progress reviewed thirty case studies drawn from eleven research papers published between 1992 and 2007. Across twenty-seven of them the median cost of replacing a worker was twenty-one percent of annual salary. Twenty percent for workers earning under $50,000. Sixteen percent at $30,000 or below. That is one fifth of salary. Not one and a half times.
Gallup puts the range at one-half to two times annual salary and describes that as a conservative estimate.
SHRM's bands, which are the ones most often quoted second-hand, run thirty to fifty percent of salary for entry-level roles, one hundred to one hundred and fifty percent for professionals, and up to two hundred percent for executive and specialized positions.
So: twenty-one percent, fifty to two hundred percent, and thirty to two hundred percent.
Three credible sources, a large spread.
They are not contradicting each other. They are counting different things.
It is worth asking why 1.5 in particular became the house number. It sits roughly at the midpoint of the professional band, which makes it the figure least likely to be challenged from either direction. Nobody has to defend it, so nobody does, and it propagates through spreadsheets by inheritance rather than by measurement. A number that survives because it is unobjectionable is not the same thing as a number that survives because it is right.
What the multiplier is actually measuring
The CAP median counts what you can put on an invoice. Agency fees and job-board spend. Interview time at a loaded rate. Sign-on money. Temporary cover. The productivity gap while somebody new comes up to speed. Every item there is an expense line that already exists somewhere in your ledger, and when you total them for a role that can be filled from a normal candidate pool, one fifth of salary is about right.
The larger figures start counting things that never generated an invoice. Knowledge that was never written down. Relationships that belonged to a person rather than to an account. Judgment about which problems were going to become real. Those are genuine costs and they are not in the CAP number, because you cannot find them in a ledger.
Which means the multiplier is not a fact about employees at all. It is a record of which costs somebody decided to count, and the person who decided is usually not in the room when the number gets used.
That is also why the same figure fails in two directions at once. Applied to a role you can refill in six weeks from a warm pipeline, 1.5× overstates the cost and makes routine churn look like a crisis. Applied to the person who had been quietly preventing three problems a quarter for nine years, it understates it badly enough to make the retention conversation look optional.
Where my own number sits, and what it rests on
For the specific employee this site is about — the perceptive one, the one who sees the pattern early and says something about it — my estimate runs to two hundred to four hundred percent of salary. That is not a published benchmark. It is my accounting from thirty-seven years in leadership, and it includes the innovation pipeline that stalls, the pattern recognition that leaves with them, the client relationships that were personal, and the problems that stop being prevented.
That last category is the one that isn't priced, because the problems a perceptive employee prevented never happened and so never appeared anywhere. You do not get an invoice for the crisis you avoided. You get one after they leave.
Treat that range as a hypothesis about your own organization rather than as a benchmark. The point of naming it is to give you something to test, not something to adopt.
For scale, and only for scale: SHRM has put the five-year cost of turnover linked to workplace culture at $223 billion. That number is far too large to act on and I include it to establish that this is not a soft problem.
Run it yourself
Build two numbers instead of one. It takes an analyst an afternoon.
Your floor. Pick one job family and take last year's voluntary separations in it. Add up agency and advertising spend, interview hours multiplied by a loaded hourly rate, sign-on and relocation, temporary cover, and the ramp gap. This is weeks to full productivity, times the loaded rate, times the fraction of the job the new hire was not yet doing. Divide by the number of people who left. That is your floor, in your currency, from your own ledger. If your roles refill easily it will land closer to CAP's twenty-one percent than to anything else on the list.
Your exposure. Take the same list and mark every departure where the replacement took more than two attempts, or where something broke afterward that no one had connected to the departure at the time. Those are the ones the floor does not describe. If you want to know why each of them actually left, the exit interview file is not where that answer is. Count them. In most organizations they are a small fraction of separations and a large fraction of the cost.
Now you have a floor you can defend to a CFO and a concentration you can act on. One multiplier gives you neither.
What this changes about the attrition number
If the expensive departures were spread evenly across headcount, an average multiplier would work fine. They are not spread evenly. They cluster, usually by manager and usually by team, which is why the aggregate attrition rate can sit comfortably on benchmark while the cost underneath it moves.
Apply a single multiplier across every separation and you get a total that is arithmetically correct and operationally useless. It cannot tell you where to spend, because it has already averaged away the thing you would spend on.
Run the floor against the job families where departures are cheap and frequent. Run your own exposure figure against the ones where they are rare and expensive. The gap between those two populations is the number worth taking to a board.
What you could notice this week
Pull last year's separations and mark the ones you had to hire twice for.
Then ask your recruiting lead which roles they dread getting a resignation from. They know. That knowledge sits in one person's head, gets confirmed every time it happens, and appears in no system you own.
You do not need a model to do either of those, and between them they will point at the same place a model would have taken a quarter to reach.
Run this on your own headcount. The employee turnover cost calculator takes your figures and prices the spread. Substitute your own multiplier when you run it. That is what the field is there for.
Sources
- Boushey, H., & Glynn, S. J. (2012). There Are Significant Business Costs to Replacing Employees. Center for American Progress, 16 November 2012. americanprogress.org
- Gallup. This Fixable Problem Costs U.S. Businesses $1 Trillion. gallup.com
- Society for Human Resource Management, The High Cost of a Toxic Workplace Culture (five-year $223 billion figure) and SHRM's turnover cost guidance for the replacement-cost bands.
- Mann, P. A. (2026). The Perception Revolution, Chapter 2, for the 200–400 percent range and the categories behind it. Fast Company Press.