Attrition is one of the first numbers leadership asks HR for, and one of the easiest to report inconsistently. Two HR teams in the same company can produce different figures from the same data simply by making different choices about the denominator, the time period or who counts as a leaver. None of those choices is wrong in itself. The problem comes when nobody states which one was used.
The basic formula
The most widely used version is:
Attrition rate = (number of separations in the period ÷ average headcount in the period) × 100
Average headcount is commonly taken as (opening headcount + closing headcount) ÷ 2. For a company growing or shrinking quickly, a better average is the mean of the month-end headcounts across the period, because two snapshots can hide large swings in between.
A worked example
An illustrative IT services unit starts the financial year with 400 employees and ends it with 460. During the year, 92 people left.
- Average headcount = (400 + 460) ÷ 2 = 430
- Annual attrition = 92 ÷ 430 × 100 ≈ 21.4%
Had someone divided by the closing headcount of 460 instead, the figure would be 20%. Divided by the opening headcount of 400, it would be 23%. That two- to three-point spread is exactly the kind of inconsistency that causes confusion when numbers from different reports end up on the same slide.
Monthly versus annualised figures
Many dashboards show monthly attrition. A month with 8 exits on an average headcount of 430 gives about 1.9%. To compare with an annual figure, people often multiply by twelve, giving roughly 22%. That is a reasonable shortcut when months are similar, but it can mislead in a month with a seasonal spike, such as the weeks after appraisal letters or bonus payouts, when exits tend to bunch up. A rolling twelve-month figure smooths this out and is usually the better number to track over time.
Not all exits are the same
A single attrition number mixes very different events. Splitting it is where the insight usually is:
- Voluntary: resignations. This is what most people mean when they worry about attrition.
- Involuntary: terminations for performance or conduct, and layoffs. A rise here says something about hiring quality or business conditions, not about engagement.
- Regrettable versus non-regrettable: losing a strong performer in a scarce skill hurts far more than losing someone already on an improvement plan. Some organisations track regretted attrition as its own metric.
- Other exits: retirement, contract end, death. These are usually excluded from voluntary attrition but should be stated.
Early attrition: the number that points at hiring and onboarding
People who leave within their first 90 days or first six months usually leave for reasons connected to how they were hired and welcomed: the role did not match what was described, the manager relationship never formed, or onboarding left them unproductive and isolated. Tracking early attrition separately, by recruiter, by source and by hiring manager, often reveals patterns that a company-wide figure hides completely. In high-volume hiring environments such as BPOs, this is often the single most useful attrition cut.
Retention rate is not simply the opposite
It is tempting to say “21% attrition means 79% retention”, but retention is usually defined differently: of the people employed at the start of the period, what share are still here at the end? New joiners who arrived and left within the year affect attrition but not retention measured this way. Both are valid, they just answer different questions. Report which one you are using.
Turning the number into action
An attrition rate on its own rarely changes anything. It becomes useful when it is broken down by team, manager, tenure band, performance rating and location, and read alongside exit interview themes and engagement survey results. A company-wide 18% might hide one team at 45% under a particular manager and others below 10%. That kind of finding leads to a specific conversation, which is the point of measuring in the first place.
Written by the RP Strativa Team. Headcount figures in the example are illustrative.
Leave a Reply