Ask three people what it costs to replace an employee and you will get three answers an order of magnitude apart. The number most often quoted in board decks — one-half to two times annual salary — is real, well-sourced, and almost always presented without the context that makes it usable. Used carelessly, it produces a headline figure so large that finance discounts it entirely, which is the opposite of what the HR team wanted.
This piece does the unglamorous version. It sets out what the published estimates actually say and why they disagree, decomposes the cost into components you can defend line by line, identifies the three things that make the India number structurally different from the US benchmarks everyone cites, and works a full example for a 400-person company. There is a calculator at the end where the assumptions are yours to set rather than ours.
| India attrition, 2025 | 17.1% (Aon, 1,060+ companies) |
| Direction of travel | Down from 18.7% in 2023 |
| Published replacement cost | 20% – 200% of annual salary |
| Most-cited range | 50% – 200% (Gallup) |
| 2026 salary increase budget, India | 9.1% projected |
| 400 people at 12 lakh average CTC | ₹1.6 cr – ₹8.2 cr a year |
Start with the range, not the headline
There is no single credible figure for the cost of replacing an employee, and any article that gives you one is either selling something or has not read its own sources. The published estimates span a full order of magnitude. That spread is not sloppiness — it is the direct consequence of different studies counting different things.
| Source | Figure | What it counts |
|---|---|---|
| Gallup | 50% – 200% of annual salary | The most widely cited range; varies by role and seniority |
| SHRM (commonly cited) | Six to nine months of salary | Roughly 50% – 75%, per role |
| Review of 30 case studies | ~20% of annual salary | An explicit challenge to the higher estimates |
| SHRM 2025 benchmarking | $5,475 non-executive; $35,879 executive | Cost per hire only — direct recruiting spend, nothing else |
| Industry surveys (upper bound) | Up to 400% of salary | Senior, specialist and revenue-owning roles |
Sources: Gallup (via multiple secondary compilations); SHRM 2025 Benchmarking Report; a peer-reviewed review of 30 case studies arguing the true figure is nearer 20%; Applauz 2025 research for the upper bound. All figures are US-derived. See the caveat on transferability below.
What the number is actually made of
A defensible model separates costs you could point to in the general ledger from costs that are real but estimated. The split below is an illustrative allocation, not a benchmark: the percentages are ours, chosen so the components sum to roughly the mid-point of the published range for a mid-level professional role. Your own figures should replace them.
| Component | Illustrative % of annual CTC | Type | Where it shows up |
|---|---|---|---|
| Separation and settlement admin | 2% – 5% | Hard | HR and payroll time on exit formalities, full and final settlement, knowledge-transfer coordination |
| Recruiting spend | 8% – 20% | Hard | Job boards, agency fees (commonly 15% – 25% of first-year salary), assessments, background verification, referral bonuses |
| Internal hiring time | 3% – 6% | Soft (real) | Recruiter, hiring manager and panel hours — rarely tracked, always spent |
| Vacancy and coverage | 5% – 25% | Soft | Output not produced, or produced by colleagues at the expense of their own work |
| Ramp to full productivity | 10% – 40% | Soft | The gap between what you pay a new hire and what they deliver in months one to nine |
| Knowledge and relationship loss | Highly variable | Soft | Client relationships, undocumented process knowledge, institutional memory |
Two practical notes. First, only the top two lines will ever appear in your accounting system, which is why a CFO's instinctive estimate of turnover cost is usually far too low. Second, the ramp component dominates for skilled roles and is close to zero for roles where a trained replacement reaches full output in a fortnight — which is exactly why a single company-wide multiplier misleads.
Three things that make the India number different
Almost every published replacement-cost figure is derived from US data. Three structural features of the Indian market push the vacancy and ramp components in a direction those benchmarks do not capture.
1. The notice period changes the vacancy arithmetic
India's professional services and IT sectors run some of the longest notice periods in the world. Original research by Analytics India Magazine found that roughly one in three IT roles in India carries a full 90-day notice period, the highest share of any skilled job category in the country; the large services firms typically apply 90 days to experienced employees, while product companies and global capability centres tend to run 30 to 60 days.
This cuts both ways, and the direction depends on which side of the transaction you are on. As the losing employer, a long notice period is protective — you have three months of overlap to hire and hand over. As the hiring employer, it is the opposite: the distance between a role opening and a seat being filled can run past four months once a 30-to-60-day hiring process is followed by a 90-day notice. Companies that model attrition cost using US vacancy assumptions systematically understate the coverage period on the inbound side.
2. The replacement usually costs more than the person who left
Aon's Annual Salary Increase and Turnover Survey — now in its 31st year, covering over 1,060 companies across 45 industries — projects a 9.1% salary increase for India in 2026, following 8.9% actual growth in 2025 (an earlier release in the same survey cycle put the 2026 projection at 9.0%, revised upward in February 2026 — a small discrepancy worth knowing if you see both figures quoted). Internal increments run at that budgeted rate; external market offers for the same role frequently do not. The replacement premium is a genuine, recurring cost of attrition that a salary-multiplier model buries rather than surfaces.
3. The national attrition rate is close to useless for planning
The headline figure hides a spread wide enough to make it meaningless at company level. Secondary compilations of Aon and related survey data put global capability centres near 12.6%, IT services at roughly 13% to 15%, and e-commerce and fintech at 25% to 28%. We flag these sector splits as second-hand — they are reported by intermediaries rather than published directly in the primary release we could verify — so treat the direction as reliable and the decimals as indicative. The planning point stands regardless: benchmark against your sector, never against the national number.
A worked example: 400 people, 12 lakh average CTC
Take a mid-market company with 400 employees and an average cost to company of ₹12,00,000. Annual payroll is ₹48 crore. At the 2025 national attrition rate of 17%, that is 68 departures a year. What those departures cost depends entirely on the multiplier you consider defensible.
| Replacement cost assumption | Cost per departure | 68 departures | As % of payroll |
|---|---|---|---|
| 20% of CTC — hard costs only | ₹2,40,000 | ₹1.63 cr | 3.4% |
| 50% of CTC — hard costs plus partial ramp | ₹6,00,000 | ₹4.08 cr | 8.5% |
| 100% of CTC — full loaded cost | ₹12,00,000 | ₹8.16 cr | 17.0% |
| 200% of CTC — senior and specialist roles | ₹24,00,000 | ₹16.32 cr | 34.0% |
What a two-point improvement is worth
This is the calculation that belongs in a board pack, because it converts a retention initiative into a number finance can compare against its cost. Moving attrition from 17% to 15% in our 400-person company means eight fewer departures a year.
| Attrition rate | Departures | Annual cost at 50% multiplier | Saving vs 17% |
|---|---|---|---|
| 17% (2025 national rate) | 68 | ₹4.08 cr | — |
| 15% | 60 | ₹3.60 cr | ₹48 lakh |
| 13% | 52 | ₹3.12 cr | ₹96 lakh |
| 20% | 80 | ₹4.80 cr | ₹72 lakh worse |
Note what this framing does. It stops arguing about whether the absolute cost of attrition is ₹1.6 crore or ₹8.2 crore — a debate you cannot win — and asks instead what a marginal improvement is worth. The marginal number is far more robust to the choice of multiplier, because the same assumption sits on both sides of the comparison.
What this model does not capture
Being straight about the limits is what makes the rest credible.
Not all attrition is a loss. This is the single biggest omission in every turnover-cost model, including this one. A model that prices all departures identically implies the optimal attrition rate is zero, which is false. Separate regretted from unregretted attrition before you cost it, or you will build a business case for retaining people you should let go.
Replacement upgrades are not counted. Sometimes the new hire is materially better than the leaver. That is a real return which no replacement-cost model recognises.
Revenue effects are excluded. For client-facing and revenue-owning roles the true cost can dwarf any salary multiple, because it is a function of accounts at risk rather than of compensation. Those roles deserve individual modelling, not a multiplier.
The benchmarks are US-derived. The 50% to 200% range comes from US research. India's lower absolute salary levels, longer notice periods and different agency-fee structures all shift the components. We use the range because nothing better exists for India — not because it transfers cleanly.
Survivor effects are unmodelled. Disengagement among the people who stayed is well documented and genuinely expensive, and we have not attempted to price it.
Using the number in a board conversation
Four rules that keep the discussion productive:
Lead with the assumption, not the total. Say "at 50% of CTC, which is the conservative end of the published range, attrition cost us ₹4 crore" — never just "attrition cost us ₹4 crore". The first invites scrutiny of a defensible assumption; the second invites dismissal of the whole exercise.
Show the range. Present three scenarios rather than one point estimate. A board that sees you have not cherry-picked will trust the mid-case.
Argue the delta, not the level. The value of a two-point improvement is a much stronger claim than the absolute cost.
Split regretted from unregretted before you present. If you cannot, say so explicitly — and make building that distinction the first ask.
The measurement problem underneath all of this
Most companies cannot produce this number reliably, and the reason is rarely analytical. Computing it requires exit dates and reasons from the HRMS, requisition-open and offer-accept dates from the recruiting system, cost to company from payroll, and settlement figures from the final payroll run — four facts that in a typical mid-market stack live in three systems and a spreadsheet, joined by hand each quarter by someone matching names across exports. The joins are where the number goes wrong: a leaver whose employee ID differs between systems simply drops out.
When hiring, HR and payroll records share a single database, the query is a query rather than a reconciliation project — which is the difference between a metric you can quote to a board with confidence and one you rebuild from scratch every time somebody asks. That is the architectural bet behind Helion, and it matters here mainly because a retention business case is only as good as the data lineage under it.
Frequently asked questions
What is a good attrition rate in India?
There is no universal answer, and the national figure of 17.1% for 2025 is a poor target because sector variation is so wide. Benchmark against your industry and, more usefully, against your own trailing twelve months split by regretted and unregretted exits.
Should I use 50% or 100% of salary as the replacement cost?
For a mixed workforce of professional roles, 50% is the more conservative and more defensible starting point, since it covers hard costs plus partial ramp. Use 100% or above only for senior, specialist or revenue-owning roles, and say so explicitly when you present it.
Why do the published estimates disagree so much?
Because they count different things. Estimates near 20% count invoiceable recruiting costs; estimates near 200% add vacancy productivity loss, ramp time and disengagement effects. Neither is wrong, but they answer different questions.
Does the 90-day notice period reduce our attrition cost?
On the outbound side it helps, by giving you overlap for handover. On the inbound side it hurts, by extending the period a role sits unfilled after you make an offer. Net effect depends on whether you are modelling exits or vacancies — most companies model only one and are surprised by the other.
Can I include gratuity and leave encashment in the cost of attrition?
Generally no. Gratuity and earned-leave liabilities accrue whether or not the employee leaves; departure changes the timing of the cash outflow, not the existence of the obligation. Including them double-counts a cost you have already provided for. The administrative effort of settling them is a legitimate inclusion; the settlement amount itself is not.
This article presents a cost model, not a benchmark. The component percentages in the decomposition table are illustrative allocations chosen by Helion to sum to the mid-point of the published range; they are not survey findings. Attrition and salary-increase figures are from Aon's Annual Salary Increase and Turnover Survey 2025-26 India; replacement-cost ranges are US-derived research reported by Gallup, SHRM and others, and their transferability to the Indian market is untested. Sector-level attrition splits are reported via secondary sources. Figures are current as at July 2026 and should be verified against the primary publications before use in a board or investor document. This is general information, not financial advice.