Three denominators give three different attrition rates from the same leaver list. Here is which to use and why it matters.
Someone reports attrition at 12%. Someone else, in the same month, from the same HR system, reports 17%.
Both are honest. Both are working from the same leaver list. They are dividing by different things, nobody in the room knows that, and so the next forty minutes are spent arguing about whether people are leaving instead of about what to do.
Attrition is a fraction. The arguments are almost always about the top of it. The error is almost always in the bottom.
Take a business that starts the year with 100 people, ends with 90, and had 15 leavers.
| Denominator | Calculation | Rate |
|---|---|---|
| Opening headcount | 15 ÷ 100 | 15.0% |
| Closing headcount | 15 ÷ 90 | 16.7% |
| Average headcount | 15 ÷ 95 | 15.8% |
Nearly two points of spread, and not one of them is wrong. They are answers to different questions.
Average headcount is the right default for a rolling rate. It is the only one of the three that accounts for the business being a different size in March than it was in December, which matters enormously in a company that grew or shrank during the period. Use the average of the monthly closing figures, not the average of opening and closing — in a year with a mid-year restructure those are not the same number.
Opening headcount answers "what proportion of the people who were here at the start have gone?" That is a useful question, but it is a retention question about a cohort, not an attrition rate, and it systematically under-reports in a growing business because the people who joined and left inside the period are in the numerator but never in the denominator.
Closing headcount has no real defence. It flatters a growing business and punishes a shrinking one, and it is used mostly because it is the number that happened to be on the screen.
The point is less which one you pick than that you pick, write the choice down where the number is reported, and use the same one everywhere.
Whatever denominator you choose, it should reconcile. Every month:
opening + joiners − leavers = closing
If that does not tie, something is wrong, and it is usually one of three things:
Put the reconciliation in as a visible column that says OK or does not, next to the numbers, every month. Not as a calculation someone does once at year end when it is far too late to go and ask.
When a rate is disputed, this is what settles it. Not a better argument — a visible denominator that adds up.
Once the denominator is sound, one rate still is not enough to act on.
Record "regretted" when the person leaves, not six months later. Reconstruct it afterwards and you are no longer measuring who you wanted to keep — you are measuring how the replacement worked out.
Reporting only total attrition makes a restructure look like a business in trouble. Reporting only the headline hides the opposite case: a stable total rate concealing rising regretted attrition in the team that matters most.
A company-wide rate is an average of populations that have nothing to do with each other. It is almost never the number that leads to an action.
But when you split by team, put the counts next to the rates, and read them first. In a team of ten, one leaver is 10% and two are 20%, and the difference between those two percentages is one person who moved city. A team table sorted by rate will put your smallest teams at the top every single time, and you will spend the meeting on noise.
The real finding is usually a large team with a mid-range rate and a lot of people behind it — seven leavers at 12% is a bigger problem than three leavers at 23%, and only one of those two is visible in a sorted rate table.
The [Attrition & Retention Dashboard](/products/attrition-and-retention-dashboard.html) does all of this — the reconciling denominator, the three-way split, teams with counts beside rates, and the tenure curve — with a complete worked example so you can see what the analysis looks like finished. £29.