Nobody looks up turnover benchmarks out of curiosity. You looked this up because you have a number — 18%, 31%, 46% — and you need to know whether to panic. So let's answer that first, then explain why the answer is more complicated than a single figure.
The short version: US turnover averages somewhere in the high 30s to mid 40s percent annually across all industries, but that average is close to meaningless. A 40% rate is a crisis in finance and a good year in fast food. Industry is the single biggest determinant, and it swings the "normal" range by a factor of four.
Turnover Rates by Industry
The figures below are approximate annual total separations rates drawn from the US Bureau of Labor Statistics' JOLTS program, which is the closest thing to an authoritative public source. Treat them as ranges rather than precise values — BLS revises them, and they moved substantially through the 2020s.
| Industry | Approx. annual total separations | What drives it |
|---|---|---|
| Accommodation & food service | 70–80% | Seasonal, part-time, low switching cost |
| Arts, entertainment & recreation | 55–65% | Seasonal and event-driven staffing |
| Retail trade | 45–60% | Holiday cycles, scheduling volatility |
| Construction | 50–60% | Project-based; completion counts as separation |
| Professional & business services | 45–55% | Includes staffing agencies, which inflate the category |
| Transportation & warehousing | 40–50% | Driver shortages, physically demanding roles |
| Health care & social assistance | 35–45% | Burnout, shift work, acute nursing pressure |
| Information / tech | 30–40% | High demand, high poaching, layoff cycles |
| Manufacturing | 30–40% | Shift work; durable goods steadier than non-durable |
| Educational services | 25–35% | Academic-year contracts |
| Finance & insurance | 25–30% | Credentialing, deferred comp, higher switching cost |
| Government | 18–25% | Pensions, tenure, structural stability |
Read the Definition Before You Read the Number
Most benchmark comparisons are wrong because the two numbers being compared measure different things. Three distinctions matter:
Total separations vs. voluntary quits. The table above is total separations — quits, layoffs, discharges, retirements, deaths. Your internal "turnover rate" is very often just voluntary quits, which typically runs roughly half to two-thirds of the total. Comparing your quits rate to a total separations benchmark will make you look far better than you are.
Regrettable vs. non-regrettable. Losing a low performer you were managing out is not the same event as losing your best engineer, though both increment the same counter. Any turnover number that doesn't separate these is a poor management tool.
First-year turnover. This one is worth tracking separately because it diagnoses something specific. Heavy first-year attrition points at hiring and onboarding, not at culture or compensation — a genuinely different problem with different fixes. Our post on the first 90 days covers the recognition side of that.
How to Calculate It Correctly
The standard formula:
Turnover rate = (Separations during the period ÷ Average headcount during the period) × 100
Average headcount means (starting headcount + ending headcount) ÷ 2, not the headcount on December 31st. Growing companies that use ending headcount systematically understate turnover — sometimes dramatically. The full walkthrough, including monthly and annualized variants, is in how to calculate employee turnover rate.
So Is Your Number Bad?
Four questions, in order:
- How does it compare to your industry band above, using the same definition?
- What's the trend? A stable 35% is a very different situation from a 22% that became 35% in eighteen months. Direction beats level.
- Where is it concentrated? Company-wide rates hide everything interesting. Turnover almost always clusters in specific teams, specific managers, or a specific tenure band. A "company turnover problem" is usually three managers.
- How much of it is regrettable? If most departures are people you'd have fought to keep, the number is worse than it looks — even if it's below benchmark.
The Segments Worth Tracking Separately
A single company-wide rate is the least useful number you can produce. Four cuts turn it into something you can act on:
By tenure band. Split at 0–6 months, 6–18 months, 18 months–3 years, and 3 years plus. Each band fails for a different reason. Heavy 0–6 month attrition is a hiring or onboarding defect. A spike at 18–24 months is usually a career-path problem — people who like the job and can't see the next rung. Losing 3-year-plus people is the most expensive and the most under-diagnosed, because those departures take institutional knowledge that never appears in a cost model.
By manager. Uncomfortable, and the most predictive cut available to you. If two managers out of twelve account for half your departures, you don't have a culture problem — you have a management problem wearing a culture problem's clothes, and it's a solvable one.
By role family. Engineering, sales, and support have genuinely different market dynamics. Blending them produces a number that describes nobody. Sales turnover of 35% may be structurally normal; the same figure in finance is an emergency.
By regrettable vs. non-regrettable. Tag every departure at exit. It takes thirty seconds and turns your turnover metric from a headcount statistic into a business one.
What It Costs
Benchmarks are abstract; the bill isn't. Replacement cost estimates cluster in a wide but consistent band: roughly 33% of annual salary as a conservative floor (Work Institute), 50–60% as a typical figure (SHRM), and up to 150% or more for specialized and senior roles (Gallup's range).
Run it on a mid-size company: 100 employees, $65,000 average salary, 15% annual turnover, at a middle-of-the-road 50% replacement cost. That's 15 departures × $32,500 = $487,500 per year. For a company of that size, turnover is frequently a larger line item than the entire HR budget — and it never appears as one.
Our employee turnover cost calculator runs this on your actual headcount, salaries, and rate, including the three multipliers above. If you want the components broken out by role, what turnover really costs and the cost to replace an employee go through recruiting, onboarding, lost productivity, and the ramp period separately.
The Part Benchmarks Won't Tell You
Here's the thing about the industry table: it describes structural forces. Seasonality, shift work, credentialing, project cycles. You mostly cannot change those. What you can change is your position within your band — and the gap between the top and bottom quartile of the same industry is enormous.
What separates them is rarely compensation, because compensation is roughly market-set within an industry. The Work Institute's finding is the useful one: roughly three out of four voluntary departures are preventable, and the reasons cluster around career development, manager relationship, and feeling unvalued — not pay.
Recognition sits directly on that third one, and the effect sizes are large enough to be worth taking seriously. Employees who feel inadequately recognized are roughly twice as likely to say they'll quit within a year (Gallup/Workhuman). Organizations with strong recognition cultures see up to 31% lower voluntary turnover (Deloitte). Applied to the example above, a 31% reduction is about $151,000 a year — considerably more than any recognition program costs.
The manager point compounds it: roughly 70% of the variance in team engagement traces to the manager (Gallup). Which is why company-wide turnover numbers mislead so badly. You don't have a 35% turnover rate. You have a 12% rate on eight teams and a 60% rate on two, and the average is a fiction that prevents you from looking at the two.
Full disclosure of bias — this is the problem Propsly was built around. It runs in Slack: /props to send recognition with a message and a hashtag, posted to a public feed. Free for unlimited users with 200 props per person per month, leaderboards, and the feed. The Pro tier ($50/month flat) adds the analytics that make the concentration visible — which teams are going quiet, who hasn't been recognized in sixty days, where recognition is pooling. That's the early-warning layer sitting underneath the turnover number, and we wrote about how to read it in recognition as an early-warning system for attrition.
Benchmark yourself, absolutely. Then stop looking at the company average and go find the two teams generating most of it.