Leavers divided by average headcount, times 100. That is the formula, it takes ten seconds, and it is almost never where the difficulty is.
The difficulty is that two people in the same company, using the same data and the same formula, will produce different numbers because they made different choices about the denominator, about who counts as an employee, and about which leavers to include. In a growing company those choices are worth several percentage points, which is more than most of the changes anyone is trying to measure.
The formula, and the version people actually need
| Metric | Formula | Note |
|---|---|---|
| Attrition rate | Leavers in period divided by average headcount, times 100 | The average is the mean of monthly headcounts where you have them |
| Annualised attrition | Leavers divided by average headcount, times 12 divided by months in the period | Only meaningful when the period is representative |
| Voluntary attrition | Resignations only, over average headcount | The number most decisions should be based on |
| Regretted attrition | Leavers you wanted to keep, over average headcount | Requires a decision recorded at exit, not reconstructed later |
| First-year attrition | Leavers with under 12 months tenure, over hires in the period | The denominator is hires, not headcount |
| Retention rate | Employees present at both start and end, over headcount at start | Not the complement of attrition |
Most reporting stops at the first row. The useful reporting is rows three, four and five, and the last row is the one that gets confused with the first.
Define who counts before you count
None of these decisions has a universally correct answer. All of them need to be made once, written down, and then left alone, because the value of the metric is entirely in its comparability across periods.
- Are contractors, consultants and agency staff in the population? Usually out, but then they must be out of both the numerator and the denominator
- Are interns and fixed-term staff in? Usually out, because a fixed-term contract ending is not attrition in any useful sense
- Do internal transfers count as exits? No, and a system that treats a move between business units as a termination will quietly inflate your number
- Do people on long leave count in headcount? Usually yes, since they are still employed
- Is a retirement attrition? Technically yes, and it belongs in involuntary or its own category rather than in voluntary
- Is a death in service in the number? Include it in total attrition and exclude it from voluntary, and be careful how that report is circulated
Write the definition into the same document as the formula. When someone asks in eighteen months why the number jumped, the first thing you will want to rule out is that the definition changed.
The denominator decides the answer
This is the part that produces most of the confusion, and it is easiest to see with numbers.
Take a company that begins the year with 200 employees, hires 120 during the year, loses 30, and ends with 290. One company, one set of facts, three defensible-sounding calculations.
| Denominator used | Calculation | Reported attrition |
|---|---|---|
| Opening headcount | 30 divided by 200 | 15.0% |
| Simple average of opening and closing | 30 divided by 245 | 12.2% |
| Closing headcount | 30 divided by 290 | 10.3% |
Nearly five percentage points of difference, with nothing changing except the denominator. In a fast-growing company, opening headcount flatters nobody and closing headcount flatters everybody, which is why the closing figure has a way of appearing in board packs during a hiring surge.
Use average headcount, and where your system allows it, calculate that average from monthly headcounts rather than from opening plus closing divided by two. If the company grew steadily the two are close. If most of the hiring happened in one quarter, they are not, and the monthly average is the honest one.
Whichever denominator you pick, the rule that matters more is that you keep picking it. A number computed one way this year and another way next year is not a trend, it is two unrelated facts.
Annualising a monthly or quarterly figure
For a quarter, divide leavers by the average headcount for that quarter and multiply by four. Eight leavers against an average headcount of 245 is 3.27 per cent for the quarter, or roughly 13.1 per cent annualised.
Two cautions. Annualising a single month is usually noise, particularly below a hundred people, where two extra resignations swing the annualised figure by several points. And attrition is seasonal in ways that make naive annualisation misleading: in India the months after appraisal and increment letters carry a predictable spike, so annualising a post-appraisal quarter will overstate the year and annualising a quiet one will understate it.
Report the actual period figure alongside the annualised one. The annualised number is for comparison, the raw number is what happened.
Voluntary, involuntary and regretted
A blended attrition number cannot answer the only question worth asking, which is whether you are losing people you wanted to keep.
Voluntary means the employee chose to leave. Involuntary covers performance exits, redundancy and end of contract. A company at 20 per cent total attrition where half of it is a deliberate performance clean-up is in a completely different position from one at 20 per cent where all of it is resignations, and the blended figure describes both identically.
Regretted attrition is the sharper version: of the people who left voluntarily, how many did you want to keep. It requires a judgement recorded at the time of exit, ideally by the manager and reviewed by someone else, because a rating reconstructed six months later tends to be generous to everyone. This is the number that deserves a threshold and an escalation, rather than headline attrition.
First-year attrition, and its denominator trap
First-year or early attrition counts people who left within twelve months of joining. It is one of the most diagnostic numbers in HR, because it points at hiring quality, onboarding and the honesty of the role description rather than at anything that happened later.
The trap is the denominator. Dividing early leavers by total headcount understates it badly, because most of your headcount was never at risk of leaving in their first year during that period. Divide by the number of hires in the relevant period instead. Twelve early leavers out of 120 hires is 10 per cent, and that is the figure worth watching, not twelve out of 245.
Retention rate is not 100 minus attrition
This one appears in a lot of dashboards and it is wrong more often than not.
Retention is usually measured against a fixed starting population: of the people employed at the start of the period, what proportion are still here at the end. Attrition is measured against average headcount and counts every leaver, including people who joined and left inside the same period, who were never in the retention cohort at all.
With the earlier numbers: if 180 of the original 200 are still present, retention is 90 per cent, while attrition on average headcount is 12.2 per cent. Those do not sum to 100 and they are not supposed to. Report whichever you find more useful, define it, and do not present one as the inverse of the other.
The company number hides the answer
A single company-wide attrition figure is almost never actionable. The same 18 per cent can be a healthy business with normal churn in one function, or three teams under one manager falling apart while everything else is stable.
- By manager, which is usually the most revealing cut and the most uncomfortable to circulate
- By tenure band, since departures at three months, at eighteen months and at four years have entirely different causes
- By location and by function, which in a multi-city company often explains most of the variance on its own
- By performance rating, because losing your strongest performers and losing your weakest are opposite problems reported by the same number
- By gender and other groups where you have the data, since attrition patterns are one of the earlier places inequity becomes visible
One caution on small numbers. In a team of four, one leaver is 25 per cent, and reporting that beside a company figure of 12 per cent invites a conclusion the data cannot support. Below roughly twenty people, report counts rather than percentages, or use a longer window.
What attrition costs
The rate on its own does not persuade anyone to act. The cost usually does, and it is straightforward to assemble from numbers you already have.
Take the cost per hire for the role family, add the cost of the seat being empty until someone starts, then add the ramp period during which the new person is not yet producing at full output. For a specialist role with a two-month search and a three-month ramp, the total is normally a multiple of the recruiting cost rather than the recruiting cost itself.
A word of caution about the widely quoted figures that put replacement cost at one or two times annual salary. They vary enormously by role and the underlying studies are not always comparable. Build the number from your own cost per hire, vacancy cost and ramp assumptions, and it will survive a finance review, which the borrowed multiple usually does not.
Where the calculation goes wrong
- Changing the denominator between periods, which turns a definitional change into an apparent trend
- Counting internal transfers as exits, which inflates attrition and penalises the managers who develop people
- Mixing headcount and full-time equivalent between the numerator and the denominator
- Including contractors in one and not the other
- Comparing your figure to an industry benchmark computed on a different definition, which is most of them
- Reporting percentages for teams too small to support them
- Annualising a single volatile month and presenting it as a run rate
- Setting attrition as a target, which reliably produces managers blocking internal moves and delaying exits across a reporting boundary
That last one is worth resisting explicitly. Attrition is a diagnostic, not a goal. The moment it becomes a number someone is measured on, the cheapest way to improve it is to stop people leaving their team rather than to stop them leaving the company.
Frequently asked questions
What is the attrition rate formula?
Leavers in the period divided by average headcount for the same period, multiplied by 100. Average headcount is best taken as the mean of monthly headcounts rather than opening plus closing divided by two.
Should attrition use opening, average or closing headcount?
Average headcount. In a company that grew from 200 to 290 while losing 30 people, opening headcount gives 15 per cent, average gives 12.2 per cent and closing gives 10.3 per cent, from identical facts.
How do you annualise a monthly or quarterly attrition rate?
Divide leavers by average headcount for the period, then multiply by twelve divided by the number of months. Avoid annualising a single month in a small company, and remember that post-appraisal quarters carry a predictable spike.
What is the difference between attrition and turnover?
They are used interchangeably in most companies. Where a distinction is drawn, turnover covers all departures including those you replace, and attrition sometimes refers to positions left unfilled. Define whichever term you use rather than assuming a shared meaning.
Is retention rate the same as 100 minus attrition?
No. Retention is measured against a fixed starting population, while attrition uses average headcount and includes people who joined and left within the same period. They will not sum to 100.
What is regretted attrition?
The share of leavers you wanted to keep. It needs a judgement recorded at the time of exit rather than reconstructed later, and it is a better candidate for a threshold than headline attrition.
How do you calculate first-year attrition?
Divide leavers with under twelve months of tenure by the number of hires in the relevant period, not by total headcount. Dividing by headcount understates it substantially.
What is a good attrition rate?
It varies too much by industry, role and location for a single figure to be meaningful, and published benchmarks are usually computed on different definitions. Compare against your own trend, segmented by manager and tenure, and set the threshold on regretted attrition.
Fix the denominator, write down who counts, and split the number into voluntary, involuntary and regretted. That takes an afternoon and it is the difference between a figure people argue about and one they act on. Everything after that is segmentation, which is where the actual answer usually is.
