Employee Turnover Rate
Also called: turnover rate · attrition rate · employee attrition · staff turnover
Employee turnover rate is the percentage of employees who leave an organization over a defined period, typically annualized. The standard formula: (separations during period / average headcount during period) × 100. The metric is universal but the headline number is noisy — the useful decomposition is voluntary versus involuntary, regretted versus non- regretted, first-year versus tenured, and by manager, function, and location. High turnover is not automatically bad (some roles have structurally high turnover); low turnover is not automatically good (stagnation can hide as stability).
Why it matters
Turnover is expensive — replacement cost for a knowledge worker is typically 50-200% of annual salary when sourcing, interviewing, onboarding, ramp time, and productivity loss are totaled. Turnover is also a signal — rising voluntary turnover is usually the first measurable indicator that something is wrong (engagement, compensation, management, work design). Organizations that track turnover with discipline — decomposed, segmented, trended — catch problems early. Organizations that track a single headline number miss the signal because the number averages across segments with very different dynamics.
How it works
Take a 3,200-person technology company. The HR analytics team publishes a monthly turnover dashboard. Overall annualized rate: 14%. Decomposition: voluntary 11%, involuntary 3%. Within voluntary: regretted departures 8%, non-regretted 3%. By tenure: first-year 22% (high — onboarding investigation triggered); 1-3 years 13%; 3+ years 6%. By manager cohort: 10th percentile manager produces 28% voluntary regretted turnover, 90th percentile produces 4%. By function: engineering 9%, sales 19%, customer success 17%. Each decomposition points to a different intervention. The headline 14% number, used alone, would support no specific action.
Retention rate The inverse framing — percent of employees at the start of a period who are still present at the end. Turnover and retention rate are mathematical complements (roughly: retention = 100% - turnover in a stable organization), but retention framing often drives different conversations. Executives comfortable with "turnover is 14%" may find "we kept 86% of our people" feels different — and the framing can shape intervention priorities. Retention-focused organizations often invest differently than turnover-focused ones, even though the underlying numbers are the same.
The operator's truth
The biggest analytical mistake with turnover is confusing the average with the distribution. Company-wide turnover of 14% can mean "14% everywhere" or "6% in most of the company and 40% in three specific managers' teams." The intervention for the first is organization- wide; for the second it's three conversations. The second scenario is far more common. Mature HR functions also distinguish between regretted and non-regretted turnover — losing low performers is not the same as losing top performers, and a turnover rate that looks stable can mask a regretted-turnover spike that is the actual problem.
Industry lens
In retail and hospitality, turnover rates of 60-130% are structurally normal for hourly roles. The metric matters but the thresholds are different.
In professional services, turnover is a pyramid-structure feature — firms expect and manage significant annual attrition in the lower ranks.
In manufacturing, turnover in skilled trades is increasingly problematic as the skilled-trades labor pool shrinks.
In healthcare, RN turnover has been in crisis territory since 2020-2021; it is slowly normalizing but remains a board-level concern at many systems.
In tech, knowledge-worker turnover in 2025-2026 has moderated from the 2021-2022 peaks but remains elevated in certain hot specialties.
In public sector, turnover is historically low but rising, and often intersects with retirement waves (baby-boomer departures creating institutional-knowledge gaps).
In the AI era (2026+)
AI changes turnover analytics in 2026 by enabling more sophisticated cohort and attribution analysis — understanding not just "who left" but "what conditions preceded leaving" across performance trajectory, manager changes, role transitions, and engagement signals. Predictive turnover models are now standard in large HR analytics functions. The risk is using the models to manage individuals ("this person is at flight risk") rather than conditions (the manager pattern, the workload spike, the comp compression). The organizations getting this right use turnover prediction to diagnose systems; the ones getting it wrong use it to surveil individuals.
Common pitfalls
- Single-number reporting. "Our turnover is 14%" without decomposition tells leaders nothing actionable.
- Confusing voluntary with involuntary. These require different interventions. Report them separately.
- Not distinguishing regretted. Losing low performers is healthy; losing top performers is damage. Treat them differently.
- Ignoring first-year cohort. First-year turnover is an onboarding and selection signal. Separate it from tenured turnover.
- Comparing across different formulas. Benchmarks use different definitions (annualized vs point-in-time, excluding contractors or not). Mixed comparisons mislead.
- Chasing zero. The goal is not zero turnover; it is healthy turnover. Some turnover renews the organization.
