Attrition Rate: How to Calculate, Analyse & Reduce It

Attrition rate is the most quoted and least understood number in HR. Leadership asks for it in every review, founders compare it over coffee with other founders, and yet ask three people in the same company to calculate the attrition rate and you will often get three different numbers — because they used different formulas, different headcount bases, and different definitions of who "left". This guide fixes that. It explains how to calculate attrition rate correctly, how to slice it so it actually tells you something, how to estimate what attrition costs, and how to build a retention plan that moves the number instead of just reporting it.

It is written for HR managers, founders, and people-analytics beginners at Indian SMBs, but the math and methods are universal. You will find worked examples with illustrative numbers throughout — treat those as arithmetic demonstrations, not industry benchmarks.

What attrition rate actually measures

Attrition rate measures the proportion of your workforce that left the organisation over a period, relative to the average size of that workforce in the same period. It answers one question: how fast are we losing people?

Before the formula, settle the vocabulary, because sloppy definitions are the root of most attrition confusion:

• Attrition vs turnover. In Indian usage the terms are largely interchangeable; some organisations use "attrition" for all exits and "turnover" to include the churn-and-replace cycle. Pick one convention, define it in writing, and use it consistently.
• Voluntary attrition: the employee chose to leave — resignation for another job, higher studies, relocation, personal reasons. This is the number most retention work targets.
• Involuntary attrition: the organisation initiated the exit — termination for performance or conduct, redundancy, non-confirmation after probation, end of contract.
• Regretted vs non-regretted: among voluntary exits, would you have fought to keep this person? A resignation you quietly welcomed is analytically different from losing a top performer. This tag requires honest manager input, and it transforms the metric's usefulness.
• Early attrition: exits within the first 90 days (some companies use 180). These are usually hiring or onboarding failures rather than engagement failures, and they deserve their own line.

An exit's date should be the last working day, not the resignation date — you are measuring when the organisation actually lost the person. Count interns and fixed-term contract completions separately (a completed internship is not attrition), and decide explicitly how you treat absconding cases and internal transfers between entities.

The core formula

The standard attrition rate formula for a period:

Attrition rate (%) = (Number of exits during the period ÷ Average headcount during the period) × 100

Average headcount for a month is typically:

(Headcount at start of month + Headcount at end of month) ÷ 2

For longer periods, average the monthly averages rather than just the endpoint months — it smooths hiring spikes.

Worked example 1: monthly attrition

Suppose a company starts June with 118 employees, hires 10, and 6 people leave, ending the month at 122.

• Average headcount = (118 + 122) ÷ 2 = 120
• Attrition rate for June = (6 ÷ 120) × 100 = 5.0% for the month

Worked example 2: quarterly attrition

Exits in April, May, June are 4, 3, and 6; monthly average headcounts are 112, 116, and 120.

• Total exits = 13
• Average headcount for the quarter = (112 + 116 + 120) ÷ 3 = 116
• Quarterly attrition = (13 ÷ 116) × 100 ≈ 11.2% for the quarter

Annualising: handle with care

Executives usually want an annual number. Two legitimate approaches:

• Trailing twelve months (TTM): total exits in the last 12 months ÷ average monthly headcount over those 12 months × 100. This is the most honest single number; it uses only real data.
• Simple annualisation of a shorter period: monthly rate × 12, or quarterly rate × 4. Fast, but it amplifies one unusual month into a scary annual claim. If one month had a team-level layoff, annualising it fabricates a crisis.

A note on compounding: some analysts prefer a compounded annualisation for high monthly rates. For SMB reporting, TTM is simpler and harder to argue with; use annualised monthly numbers only as an early-warning proxy, clearly labelled.

Common calculation mistakes

• Using end-of-period headcount as the base. A shrinking company dividing exits by a shrunken base inflates the rate; a fast-hiring company deflates it. Always use average headcount.
• Mixing voluntary and involuntary silently. A quarter with a planned restructuring is not comparable to a normal quarter. Report total attrition and voluntary attrition as separate lines.
• Counting internal movement as exits. Transfers between departments or group entities are not attrition (though a department head may track "team attrition" separately for their unit — label it as such).
• Including interns or contract completions. Planned endings are not attrition.
• Changing definitions between reports. The single worst sin: any trend becomes meaningless. Freeze definitions in a one-page metrics dictionary.

Edge cases your formula must survive

Real companies are messier than textbook examples. Decide these in advance and write them into your metrics dictionary:

• Rehires. An employee who left in January and returned in July is one exit and one hire; their tenure clock restarts unless you deliberately track "boomerang" tenure separately. Tag boomerangs — a healthy boomerang rate is a good employer-brand signal worth reporting on its own.
• Entity or payroll transfers. Movement between group companies, or from contractor payroll to employee payroll, is not attrition if the person kept working for you. Count conversions as internal movement, and keep a consistent rule for which entity's headcount they occupy each month.
• Mergers and acquisitions. An acquired team joining mid-year distorts both the base and, often, next-year exits. Report acquired-cohort attrition separately for at least four quarters so integration issues are visible instead of blended away.
• Seasonal and project workforce. If you staff up for seasons with fixed-term roles, planned completions are not attrition; mid-season quits are. Two lines, two stories.
• Probation non-confirmations. These are involuntary exits and also early-attrition data points — count them in both cuts. A rising non-confirmation rate is a hiring-quality finding.
• Long unpaid absence. Someone on extended unpaid leave is usually still headcount until formally separated; define the trigger (e.g., separation letter date) so months don't silently disagree.
• Death or medical separation. Count separately with dignity in reporting; these should never inflate a "voluntary" line or trigger cause-analysis debates.

None of these choices is universally right; consistency is what makes trends real. The test of a good metrics dictionary is that a new HR analyst reproduces last quarter's number exactly, from raw data, without asking anyone.

Slicing the number: where insight actually lives

A single company-wide attrition rate is a headline, not a diagnosis. The value appears when you segment:

By tenure cohort

Break exits by tenure at exit: 0–3 months, 3–12 months, 1–3 years, 3+ years. Each band has a different story:

• 0–3 months: hiring promises vs reality, onboarding quality, role clarity. High numbers here are recruitment-process findings, and they are expensive — you paid full acquisition cost for near-zero productive output.
• 3–12 months: manager quality, integration, early growth signals.
• 1–3 years: career-path stagnation, compensation drift vs market, the "learned everything, see no next step" exit.
• 3+ years: senior stagnation, leadership changes, or perfectly healthy alumni movement.

By team, manager, and location

Attrition by manager is the most uncomfortable and most useful cut. One manager losing people at twice the company rate over several quarters is a signal no engagement survey states as plainly. Insist on minimum cohort sizes (don't compute rates on teams of three) and use rolling periods to avoid noise.

By performance band

Losing your bottom band at a modest rate may be healthy; losing your top band at any meaningful rate is a fire. Combine performance ratings with the regretted flag: regretted attrition among high performers is the single most important people-risk metric an SMB can track.

By recruitment source and recruiter

If exits within the first year cluster around a particular source (one agency, one job board, one referral pool) or a particular interview panel, your funnel is importing attrition. Feed this back into sourcing spend.

By demographic dimensions

Cut by gender and life-stage where cohorts are large enough to be meaningful and anonymous. For instance, elevated attrition among women returning from maternity leave points directly at return-to-work support, flexibility, and manager behaviour — fixable things.

Worked example 3: reading one company's numbers

Put the cuts together for a fictional 150-person software services firm, "Meridian Tech", reviewing its trailing twelve months.

The headline: 34 exits against an average headcount of 152 — TTM attrition of 22.4%. The COO's first reaction is alarm; a board member has been quoting someone else's "industry average". Then HR presents the segmentation.

Of the 34 exits, 6 were involuntary (4 non-confirmations, 2 performance terminations), leaving voluntary attrition at 18.4%. Of the 28 voluntary exits, managers flagged 11 as regretted — a regretted rate of 7.2%. Eleven of the voluntary exits, though, occurred within the first year, and 6 of those within 90 days. Early attrition among the last year's 38 hires is 15.8% — the ugliest number on the page.

Tenure cuts show the 1–3 year band otherwise stable. Team cuts show one delivery team of 18 contributing 9 exits — half the regretted list — under a manager whose engagement scores sit in the bottom decile. Exit-interview categories for that team cluster on "workload" and "manager support", while company-wide the top category is "career growth". Source analysis shows 5 of the 6 ninety-day exits came from a single hiring agency's candidates, all into the same team.

Now the story writes itself, and it is not "we have an attrition crisis". It is three specific findings: an onboarding/hiring-quality problem concentrated in one agency pipeline, one manager whose team is bleeding, and a broader growth-path itch. The actions: pause the agency, install structured week-2/week-6 onboarding check-ins, a direct intervention with the delivery manager (coaching, workload rebalance, and a follow-up review in one quarter), and an accelerated career-framework rollout communicated honestly.

Two quarters later, the TTM number has barely moved — trailing metrics move slowly — but 90-day attrition has fallen to a single case, the delivery team has had zero regretted exits, and stay interviews confirm the framework announcement bought patience. The board conversation changes from "why is the number high?" to "the leading indicators turned; the trailing number follows". That is what analytically mature attrition management looks like: the same 34 exits, but three actionable causes instead of one scary percentage.

What is a "good" attrition rate?

The honest answer: it depends on industry, role mix, city, growth stage, and the year's market — and published benchmark numbers vary so much in methodology that borrowing them is risky. Rather than chasing an external number, use three internal reference points:

1. Your own trend. Is TTM voluntary attrition rising or falling over four quarters? Direction beats absolute level.
2. Your regretted rate. Total attrition of 20% with near-zero regretted exits can be healthier than 10% where every exit hurt.
3. Your replacement reality. If time-to-fill and ramp-up time are long for your roles, the same attrition rate does more damage; your tolerable level is lower.

Where leadership insists on external comparison, source current-year data for your specific industry and city from reputable surveys, note the methodology, and still anchor decisions on trend and regretted mix.

There is also such a thing as too-low attrition: no one leaving for years can indicate below-market pay locking people in, stagnant roles, or accumulating performance debt no one addresses. Healthy organisations have a pulse — modest, mostly non-regretted outflow with strong internal mobility.

The cost of attrition: a framework you can defend

You do not need a precise rupee figure to make decisions, but a defensible framework earns retention its budget. Cost per exit typically includes:

For a mid-level role, running this exercise honestly usually lands total cost between several months' and a year's salary of the departed employee — the range most practitioners use as a rule of thumb. Compute it once for two or three representative roles in your company; multiply by regretted exits per year; present that annual figure when discussing retention investments. Suddenly a manager-training budget or a compensation correction cycle looks cheap.

Diagnosing why people leave

The attrition number tells you speed; diagnosis tells you cause. Combine four inputs:

Exit interviews that actually collect data

• Conduct them in the last week, by someone outside the direct reporting line (HR or a founder), after F&F anxieties are settled in writing so answers are candid.
• Use a consistent short questionnaire — primary reason (one choice from a fixed list), contributing factors (multi-select), destination type, "would you return?", "would you recommend us?" — plus open conversation.
• Fixed lists matter: free-text-only exit interviews produce anecdotes, not analysable data. Track category shares over rolling periods.
• Discount politeness bias: "better opportunity" is often the socially safe wrapper around "my manager" or "my pay". Triangulate.

Stay interviews with the people still here

Interviewing only leavers is studying the crash after ignoring the dashboard. Structured stay conversations with your regretted-risk population — top performers, critical roles, people at tenure inflection points — surface fixable issues while they are still fixable. See our stay interview playbook for scripts and cadence.

Engagement and pulse data

Correlate survey dimensions (manager support, growth, recognition, workload) with subsequent team-level attrition. Even simple eyeballing — teams in the bottom quartile of manager scores versus their next-two-quarter attrition — usually shows the pattern that justifies manager investment.

Compensation position

Attrition clustered in roles where your pay has drifted below market is a compensation finding wearing an engagement costume. Refresh salary benchmarking for high-attrition roles before concluding anything cultural.

Designing your exit-reason taxonomy

Analysable exit data starts with a disciplined reason-code list. Principles first: one primary reason per exit, chosen from a fixed list; unlimited contributing factors; codes assigned after the exit interview, not copied from the resignation email; and a quarterly HR calibration pass so two HRBPs code the same story the same way.

A workable SMB taxonomy:

• Compensation — pay or total rewards below expectation or market
• Career growth — role stagnation, no promotion path, learning plateau
• Manager relationship — support, fairness, communication with direct manager
• Work content — role mismatch, monotony, technology or domain mismatch
• Workload and wellbeing — sustained overload, burnout, work-life strain
• Work model — location, commute, flexibility, return-to-office friction
• Culture and environment — values mismatch, team climate, conduct concerns
• Personal circumstances — relocation, family care, health, higher studies
• Better external opportunity (unspecified) — use sparingly; push for the underlying driver
• Involuntary — performance, conduct, redundancy, non-confirmation (sub-coded)

Resist two temptations. Don't let "better opportunity" absorb half your exits — it is a destination, not a reason; ask what made the person open to the call. And don't grow the list past a dozen primary codes; long taxonomies fragment data until every category is too small to act on. When a new pattern emerges (say, relocation for a spouse's job becoming common), add a contributing factor first and promote it to a primary code only if it sustains.

Close the loop: reason-code shares belong on the quarterly dashboard beside the rates, because the taxonomy is what converts attrition measurement into retention strategy.

Leading indicators: seeing exits before the letter

Resignations are lagging indicators. Watch the earlier signals:

• Sustained drop in engagement or pulse scores for an individual or team
• Declined promotions or transfers, or repeated "passed over" cycles
• Leave patterns shifting (sudden clustering, or zero leave followed by long leave)
• Skill-building bursts on external platforms coupled with disengagement at work
• Manager change: the first two quarters under a new manager are an elevated-risk window
• Compensation review outcomes below expectation, especially two cycles running
• A teammate's exit: departures cluster; treat each regretted exit as a risk event for the adjacent three or four people

None of these alone predicts an exit, and monitoring must respect privacy and dignity — the point is managerial attentiveness, not surveillance. A monthly 30-minute "retention risk review" where HR and team leads walk the regretted-risk list against these signals is the lightest-weight early-warning system that actually works.

From number to plan: retention levers that move attrition

Match the lever to the diagnosed cause — generic "engagement initiatives" spread thin move nothing.

Sequence matters: correct compensation and manager issues first — no growth framework survives an underpaying manager people distrust — then build the longer-arc career and culture work.

The 90-day retention sprint

For an SMB seeing worrying numbers, a focused sprint beats a grand strategy:

1. Weeks 1–2: Freeze definitions, rebuild the last 8 quarters of attrition by tenure/team/regretted flag, run compensation checks on the three highest-attrition roles.
2. Weeks 3–4: Stay interviews with the top-20 regretted-risk list; manager-level attrition review with minimum-cohort rules.
3. Weeks 5–8: Execute quick wins — pay corrections where indicated, onboarding fixes, one manager intervention; announce career-framework work honestly.
4. Weeks 9–12: Stand up the monthly retention review, the exit-interview taxonomy, and the dashboard below; set next-two-quarter targets for regretted attrition specifically.

Building the attrition dashboard

A monthly one-pager, generated from your HRMS rather than assembled by hand:

• Headcount: opening, hires, exits, closing; average headcount
• Attrition: monthly rate, TTM rate; voluntary vs involuntary; regretted vs non-regretted
• Early attrition: 90-day exit rate for cohorts hired in the last two quarters
• Segments: attrition by department and location (rolling 6 or 12 months); manager-level view for leadership eyes with cohort-size rules
• Exit reasons: category shares, rolling 12 months
• Watchlist: open regretted-risk cases and actions
• Cost line: regretted exits YTD × modelled cost per exit

In CozyHR, exits, tenure, and reason codes flow from the offboarding module into HR analytics automatically, so this dashboard is a report you open, not a spreadsheet you rebuild — and trends by team, tenure, and manager are a filter away.

Presenting attrition to leadership and the board

The same data lands differently depending on framing. A few practices keep the conversation productive:

• Lead with the decomposition, not the headline. Open with total → voluntary → regretted → early, in four numbers. The headline alone invites benchmark debates; the decomposition invites action.
• Always show trend, minimum four quarters. A single-period number is a Rorschach test; a trendline is evidence. Annotate known events (restructuring, acquisition, market swing) directly on the chart so history doesn't get re-litigated.
• Name the concentration. "Attrition is 22%" frightens; "attrition is concentrated in one team and one hiring source, and here is the plan" reassures — and is more accurate.
• Pair every problem number with an owner and a date. Leadership reviews degenerate without this. The retention plan is a table: finding, action, owner, checkpoint.
• Report the cost line annually. Regretted exits × modelled cost per exit, with the model's assumptions stated once. This is the number that funds retention work.
• Protect individuals. Manager-level cuts go to the CEO/CHRO view with cohort-size rules, not to all-hands decks. Attrition analytics must never become public shaming, or managers will start gaming the regretted flag and your data dies.

One more habit separates mature teams: pre-commit the definitions. Share the metrics dictionary with the board once, get nods, and never again change a definition silently. The fastest way to lose a board's trust on people metrics is a number that improves because the formula moved.

Frequently asked questions

What is the difference between attrition rate and retention rate?

Retention rate measures the share of a starting population still employed at the end of a period; attrition rate measures exits against average headcount during the period. They are complementary but not simple mirror images once hiring during the period is involved. Cohort retention (e.g., "of the 40 people we employed on 1 April, 34 remain on 31 March — 85% retention") is especially useful for evaluating specific hiring classes.

Should notice-period employees count in headcount?

Yes — they are employed until the last working day, which is also the exit date you should use. Consistency is the point: the same person must not be an exit in one report and headcount in another for the same date.

How do we treat absconding employees in attrition?

Count them as voluntary exits dated per your absconding process's termination date, and tag them distinctly — a rising absconding share is its own signal about hiring quality and early experience. See our absconding process guide for the HR handling.

Is high attrition always an HR failure?

No. Market cycles, an acquisition, a relocation, or a deliberate performance reset all move the number for reasons outside classic engagement. That is precisely why you segment voluntary/involuntary and regretted/non-regretted before assigning meaning — the goal is honest diagnosis, not a single scorecard number to defend.

How often should we report attrition?

Compute monthly, review meaningfully quarterly. Monthly numbers in a small company are noisy — three exits in a 60-person firm is a 5% month — so pair monthly monitoring with rolling 6- or 12-month views for decisions.

What attrition rate should a startup target?

Resist adopting a borrowed number. Set targets on the two metrics you control interpretively: regretted attrition (drive toward a small single-digit annual figure) and 90-day attrition (drive toward near zero). Total attrition then lands wherever your strategy and market put it, and you can explain it.

Our attrition is concentrated in one team. Manager problem or role problem?

Compare with the same role in other teams and the same manager across roles, look at exit-interview categories for that team, and check the role's market pay movement. If exits cite growth and workload while other teams doing the same role are stable, it is likely the manager; if the same role bleeds everywhere and pay drifted, it is the role's deal. Often it is both — fix pay first, then coach or change the manager.

Can we predict attrition with AI?

Larger datasets support useful models, but for a few hundred employees, simple rules — tenure inflection points, manager change, comp-cycle outcomes, engagement dips — capture most of the signal without the opacity. If you adopt predictive tooling, govern it carefully: transparency, no automated adverse decisions, and human conversations as the only intervention. The point of prediction is a kinder, earlier conversation, not a label in a database.

Conclusion

Attrition rate is a thermometer, not a diagnosis. Compute it one way, every time, in writing; segment it by tenure, team, and the regretted flag; cost it once so leadership funds the cure; diagnose with exit and stay data; then apply the specific lever the diagnosis names. Companies that do this stop debating whose number is right and start having the only conversation that matters — who might leave next quarter that we would regret, and what are we doing about it this month.

The mechanical half of that discipline belongs to your systems. CozyHR captures exits with reasons and tenure automatically, computes attrition and cohort views in HR analytics, and keeps offboarding, engagement, and performance data in one place so segmentation is a filter, not a project. If your attrition reporting still lives in a spreadsheet with contested formulas, try CozyHR — measure cleanly, and spend your energy on keeping the people you would hate to lose.