Attrition in India: What the Numbers Actually Say

Sir, we are at fourteen percent. The national average is sixteen.

Somebody said this to you in a review meeting, or you read it off a benchmark table forwarded into a WhatsApp group of promoters, and you allowed yourself to relax slightly. The number was accurate. Nothing in that sentence was false.

It was also, as a guide to whether your business has a retention problem, close to worthless — and not because the benchmark is wrong. Because the two numbers being compared are made of different things.

The claim this page is built on

Every page on this subject prints roughly the same table. National figure, a column of sector figures, a line about the market cooling. What none of them does is separate the departures you chose from the departures that were chosen for you, and that separation is the entire story.

Start with the headline. Aon's India survey put attrition at 16.2% in 2025, down from 17.7% in 2024 and 18.7% in 2023. Voluntary attrition inside that came to 12%, down from 13.4%. The market has cooled.

Now the number that changes what those figures mean. As Amit Otwani of Aon puts it: "Nearly 75% of attrition remains voluntary, far higher than 50-66% in major global markets."

In most large economies, between a third and half of all departures are the company's decision — restructuring, performance exits, redundancy, the end of a project. Those are planned and in some sense wanted. In India, three out of four people who leave your business decided to leave. Nobody asked them to go.

This has a consequence that no benchmark table can show you. A falling headline attrition rate can sit comfortably alongside a worsening retention problem, because the part that fell may be the involuntary part. A business that stops letting people go — because growth slowed, because a restructuring got postponed, because nobody wants to run a difficult conversation this quarter — will report improving attrition while losing exactly the same people it was losing before.

So when a promoter compares his 14% against a national average, he is not comparing his business to the country. He is comparing one composition to a different one. The pillar piece on why your best people leave sets out the sequence that produces a voluntary exit — five stages over nine to fourteen months. This article is about counting them properly.

What the benchmark tables get wrong

Take two businesses. Neither exists; the numbers below are chosen to make a point and are not reported from anywhere.

Both employ 150 people. Both lost 21 people last year. Both report attrition of 14%, and both sit below the national figure.

In the first, sixteen of those twenty-one were resignations the promoter did not want and could not easily replace — a plant manager, two branch heads, a cluster of experienced sales people, the woman who held the entire receivables process in her head. Three were departures he was relieved by. Two he initiated.

In the second, ten were company-initiated: an underperforming branch was closed and a layer that had stopped earning its cost was removed. Six more were resignations that solved a problem he had been avoiding. Only five were people he wanted to keep.

Same headline rate, same position against the benchmark. One business is bleeding capability and calling it a good year. The other is doing housekeeping. No table will ever tell them apart.

The Voluntary Share

The correction is not complicated and needs no software. Sort your leavers into three piles instead of counting them into one.

The three categories

Unwanted voluntary. People who resigned and whose departure genuinely damaged the business. The test is not seniority or salary — it is whether you tried to keep them, or would have if you had known in time, and whether refilling the role was hard, slow or expensive.

Neutral voluntary. People who resigned and, when you were honest with yourself a week later, you were relieved. The performance conversation you had postponed twice resolved itself. These are not a retention failure. They are usually a delayed management decision arriving by another route.

Company-initiated. Exits you decided on. Performance, restructuring, closure, misconduct, the end of a contract. Whatever you feel about them, they are evidence of a decision, not of a retention problem.

The calculation

Take the number in the first pile — unwanted voluntary exits over the last twelve months. Divide it by your average headcount across the same twelve months. Multiply by a hundred.

That figure is your real attrition rate. It will be a good deal lower than the number you have been quoting, which is not comforting but clarifying: it tells you how many events you are actually trying to prevent, and it gives you a list of names rather than a percentage.

Then do it a second way. Divide the same first pile by your total exits. That percentage is your Voluntary Share. If it runs near or above the national three-quarters, your attrition is being driven by other people's decisions. If it is well below, much of your headline number is you managing your own business.

Why almost nobody has done this

Because attrition reaches you as a single figure from a person whose job is to report a single figure, and because the split requires an admission. To mark an exit as neutral is to say out loud that you kept somebody eleven months longer than you should have. To mark one as unwanted is to accept it was preventable, and Gallup's finding that 52% of voluntarily exiting employees say their manager or organisation could have done something to prevent them from leaving makes that an uncomfortable column to total.

Promoters running this split for the first time are usually startled, and rarely by the size of the number. They are startled by the names.

The sector figures, and how to use them

Aon's sector data puts e-commerce at 25–28%, IT services at 13–15%, and global capability centres at 12.6% — the last at historic lows. Two rules for reading them.

First, a benchmark is only informative against your own sector. A manufacturing promoter reading an IT services figure is reading fiction. That number describes short tenure norms, campus intake, published salary bands and a resignation process that is nearly administrative. None of it maps onto a business where the plant maintenance head is the only person who understands a fifteen-year-old machine.

Second, even inside your sector the comparison only holds if the composition holds. A global capability centre at 12.6% and a family-owned engineering business at 12.6% are describing different events. The centre has a talent function, a calibration cycle and a real involuntary component. You may have none of those, in which case your 12.6% is almost entirely people walking out on their own.

Sector figures are useful for one purpose: knowing what the market around you is doing to the availability of the people you need. They are not a scoreboard, and a business that beats its sector average while losing its four most difficult-to-replace people has beaten nothing.

On direction, Kapil Joshi of Quess IT expects attrition to ease further, to around 13–14% from 2026 onwards, and Roopank Chaudhary of Aon has described organisations as benefiting from a more stable workforce. Both are probably right about the aggregate. Neither tells you anything about the six people whose departure would cost you a year.

Why India's voluntary share is structurally higher

This part is observation rather than research, and it should be read that way. What follows are hypotheses the Aon figure is consistent with, not findings anybody has demonstrated.

• A young workforce. Early-career people move more, everywhere. A country with a large proportion of them generates more voluntary movement for reasons that have nothing to do with any individual employer.
• Faster switching norms. In many Indian sectors, changing employers every two to three years is unremarkable and carries no stigma. Staying eight years invites the question of why nobody else wanted you.
• Raises priced by offer letter. The most reliable way to obtain a significant increase here remains bringing in a competing offer. Where the internal increment cycle cannot match what an external move produces, resignation becomes a compensation mechanism rather than a verdict on the job — a pattern we take apart in do employees really leave for money.
• Indian firms performance-exit less readily. For cultural and practical reasons — notice periods, relationships, the discomfort of the conversation, family proximity in the leadership layer — Indian businesses part with underperformers more slowly than American ones. That alone shrinks the involuntary side of the ratio without anything changing in how people are managed.

None of these is established. They are the explanations that fit the shape of the number, and each is a different problem with a different response.

What it costs, and how to compute your own figure

Gallup puts the cost of replacing an employee at one-half to two times their annual salary — recruitment, the notice period gap, the ramp, the mistakes in month three, the customer who noticed.

Apply it to the first pile only. Add up the annual cost-to-company of your unwanted voluntary leavers over the last twelve months, halve that total for your floor and double it for your ceiling. Your real number sits between the two, and for a senior person nearer the top, because the ramp is longer and the relationships they held do not transfer.

Do not apply the multiple to your total exits. That produces a frightening figure that includes people you were glad to see go — and a frightening figure you do not believe is a figure you will ignore.

The larger number that is not in any attrition report

Your voluntary attrition figure counts the people who left. Gallup's India data counts the people who stayed: 59% of Indian employees are not engaged — present, compliant, doing the defined task, investing nothing beyond it.

Set the two side by side. Aon's voluntary attrition figure of 12% describes the people who acted. The 59% describes the people who made the same internal decision and stayed — because the market is uncertain, or a home loan is running, or the commute suits them. They cost you continuously and appear in no report you receive, which is the subject of quiet quitting in India.

This is also why attrition falling is not automatically good news. A cooling market keeps disengaged people in their chairs. The number improves and the problem does not.

What "good" looks like

Not a percentage. A composition.

Good is: your unwanted voluntary exits in roles you cannot easily refill at zero or close to it, over a rolling twelve months — with your total attrition possibly unchanged, or even higher, because you have started making decisions you were postponing.

A promoter whose attrition rises from 12% to 15% while his unwanted losses fall from seven to one has had an excellent year, and will be criticised for it by anyone reading the headline figure. That trade is unavailable to you until you keep the two numbers separately.

The layer most often miscounted is your second line, because a manager's resignation gets filed as neutral when it was anything but — the role gets covered, the reporting lines get redrawn, and nobody records what was lost. Why your managers leave deals with that layer specifically.

How to test it

An hour, with your accounts person and the full-and-final settlement records.

List every person who left in the last twenty-four months. Name, role, month, annual cost-to-company. Do not summarise; the rows are the point.

Against each name, mark one letter. U for unwanted: you tried to keep them, or would have. N for neutral: you were relieved. C for company-initiated: you decided.

Then recompute. Count the U rows, divide by average headcount, and that is your attrition rate. Count the U rows against total exits, and that is your Voluntary Share. Add the CTC of the U rows and apply the half-to-double range.

Finally, read the U list as people rather than a number, and ask what each of them last asked you for. That question is where the answer lives — and the behavioural signals that precede those exits are set out in employee flight risk indicators.

Where to start

Run the twenty-four-month split this week. It is one hour, and it produces the only attrition figure that should ever influence a decision you take.

Then stop reporting the headline number internally. Report the U count and the names behind it, monthly. What actually holds the people on that list is a small and specific set of things, set out in how to retain your key people.

If the U list is long and concentrated in one layer or one function, the cause is structural rather than personal, and the Business Pulse diagnostic maps where in your organisation people stop being able to grow — which is where unwanted resignations are manufactured, twelve months before they are counted.