Turnover Rate

Taken one at a time, every departure gets explained as a personal circumstance. One person is going to graduate school, one is moving back home, one got an offer from a bigger company. Nothing the company could have done appears in any individual case, so each one ends with refilling the seat.

The problem is that while the conversation stays individual, the pattern stays invisible. Only after a year does it surface that a twenty-five person team replaced more than ten people, and by then nobody remembers the specific cases. The exchange goes: "a lot of people left this year," answered with "it feels that way, but is it more than last year?" — and with no number to compare, the conversation ends.

The turnover rate fixes that pattern as one figure. The goal isn't to erase the individual stories but to make it possible to ask whether this year is worse than last and which teams are worse than others. Only once that number exists can you decide what to spend on employee retention — the whole set of activities that keep the people you already have. Without knowing where the leak is, there's no basis for allocating the budget.

Fix the period, then divide the number of people who left by the average headcount over it.

1. Fix the period. A year is the default. Measured quarterly, the figure swings wildly on one or two people; at forty-five employees, a single person is more than 2 percent.
2. Count the numerator: everyone who left the company during the period. Decide edge cases in advance — someone who left and returned, headcount inherited through an acquisition.
3. Work out the denominator. Use the average headcount, commonly the starting and ending counts added and divided by two. In a growing company, using the ending count alone inflates the denominator and understates the rate.
4. Divide. Eleven people leaving a company that averaged forty-four gives 11 over 44, or 25 percent.
5. Separate voluntary from involuntary. Combine people who chose to leave with people the company let go and a year with a layoff looks identical to a year when people simply moved on — same 25 percent, opposite causes.
6. Write down the population. Whether people who left during probation, interns and contractors are counted shifts the same year's figure by several points. Without the definition recorded, whoever recalculates next year produces something that can't be compared. One paragraph next to the number is enough, and it saves an argument every single year.

The baseline varies by industry. At software companies, voluntary turnover of 10 to 15 percent a year is treated as ordinary and anything above 20 gets investigated. Applying that benchmark directly to your own company is harder than it looks, though. At forty-five people, 20 percent is nine people, and if those nine are concentrated in one team, the problem belongs to that team rather than to the company.

So two things matter more than the absolute figure. One is the trend: going from 12 percent last year to 25 this year means something changed, whatever the benchmark says. The other is the breakdown. Split by team, by tenure, and by manager, the same 25 percent tells several different stories.

Splitting by tenure is especially informative. If more than half of the leavers had under a year, either hiring set the wrong expectations or the settling-in period is empty. If the departures cluster at two to three years, it reads closer to people not seeing a path to the next step. Same headline number, completely different places to intervene. Splitting by manager is the version companies hesitate over, and also the one that most often explains the cluster.

It's also worth remembering that this metric only moves after the fact. By the time you know this year was 25 percent, those people are gone. So it gets paired with signals that move earlier: an engagement survey to find where things are bad, and six months of exit interviews read together to find the reasons that repeat. Three people from three different teams giving the same reason is not a personal circumstance.

"Turnover was 25 percent last year and half of them had under a year. That's hiring or onboarding."

Breaking the figure down to narrow where the cause sits.

"Counting the layoff makes this incomparable to last year. Send me voluntary only."

Mixing two different causes makes the trend unreadable.

"Turnover is an outcome, not something we can act on today. Let's start with the engagement survey."

You can't build a response on a lagging indicator alone.

Pagedeck has forty-five employees and eleven people left last year. Dividing by forty-four — the average of forty-two at the start and forty-five at the end — gave 25 percent, of which exactly one departure was the company's decision.

Grace broke that number down three ways. Six of the eleven had under a year of tenure, and four of those six were in the same team. Reading that team's exit interview notes together, one thing repeated: weeks had passed in the first month without anyone settling what the new person owned. Looking only at the company-wide 25 percent, the response would have been compensation or culture. What actually needed fixing was one team's handling of the first month.

The most common assumption is that closer to zero is better, which follows naturally from treating every departure as a loss. But an unusually low figure has things to read in it too. Almost nobody new arriving, ways of working set hard in place, or low performers staying put all produce a low number. What matters is not the percentage but who left. That's why some companies separately track the share of leavers they had wanted to keep, a figure that can move in the opposite direction from the headline rate.

Judging the situation from the single number is nearly as common, since a percentage looks immediately comparable across companies. But an undecomposed turnover rate doesn't say what to fix. The same 25 percent concentrated in one team is that team's operating problem; spread evenly across every team, it points at compensation or culture. Pick a response without the breakdown and you repair something that was fine while the actual leak stays open. The breakdown costs an afternoon, which is cheap next to a quarter spent on the wrong fix.