Attrition Rate

Attrition rate is the number of employees who left during a period, divided by the average headcount over that period, expressed as a percentage.

Average headcount matters more than it looks. Using the opening headcount understates attrition in a growing organisation, because the denominator excludes people who joined and left within the period. Using the closing headcount overstates it in a shrinking one. The average of monthly headcounts is the standard and the most defensible.

Annualising a shorter period requires care. A quarterly rate multiplied by four assumes exits are evenly distributed through the year, and they are not: they cluster after appraisal cycles, after bonus payment, and around the academic year in some sectors. A rolling twelve-month figure avoids the assumption entirely and is the more honest measure for a business with any seasonality.

None of these choices is wrong. What is wrong is not stating them, and then comparing your figure against an industry benchmark computed on different ones. Benchmark comparisons in attrition are frequently meaningless for exactly this reason.

The minimum discipline is to publish the definition alongside the number, and to keep it stable. A definition changed between years makes the trend, which is the only genuinely useful part, unreadable.

An organisation-wide rate tells you almost nothing, because attrition concentrates and the average conceals the concentration.

• By team and manager. This is the highest-yield cut by a distance. Attrition is frequently a small number of managers rather than a company-wide condition.
• By tenure band. Exits in the first six months point at hiring and onboarding. Exits at two to three years point at progression. Exits at the same point every year point at a cycle, usually the appraisal or bonus.
• Regretted against non-regretted. Losing people you wanted to keep is a different problem from losing people you did not, and mixing them makes both invisible.
• By role and grade. Attrition in scarce technical roles has a cost and a replacement lead time that a headline rate does not convey.
• By location. Where local markets differ, a national figure averages away the site that has a problem.

Two derived figures are worth more than the rate itself: the proportion of exits that are regretted, and the median tenure at exit. The second in particular says something the rate cannot, because two organisations with identical attrition can have entirely different retention profiles.

Attrition is a lagging indicator. By the time it moves, the causes have been operating for months and the people are gone.

Three cautions.

• Low attrition is not automatically good. In an organisation with no growth and no exits, nothing is moving, and the people who would improve it are not arriving. Extremely low figures in a competitive market usually mean either an unusually good workplace or an unusually immobile one.
• A spike is often a single event: a competitor opening a site, a team's manager leaving and taking people, a restructure. Reading a spike as a cultural problem sends the response in the wrong direction.
• Cost matters more than rate. Losing twenty people from a hundred-person support function and losing five specialist engineers are not comparable, and the rate treats them as if the first is four times worse.

The measure to pair it with is something forward-looking: engagement data read by team, skip-level conversations, or internal mobility rates. Attrition confirms a problem; it should not be the first place you learn of one.