How to Retain Your Key People (Without Simply Paying More)
You have lost two people in five months and you have decided you are not going to lose the third. So you asked around. Your CA mentioned a key-man policy. A friend in Mumbai described a deferred bonus vesting over three years. Somebody forwarded a note about ESOPs with a lock-in.
Every one of those answers was about money arriving later. Not one touched the question you actually asked, which was how to keep the four people you cannot replace.
Search the question and you get the same answer again, more confidently. The first screen is banks and wealth advisers — Bank of America, Hancock Whitney, Westpac, a couple of financial planners — and each answers how to retain key employees with an instrument. Insurance. Deferred compensation. Golden handcuffs. It is a coherent answer to a different question, and we will come to which one.
The management answer is missing from that page, so here it is.
Retention is not a programme and it is not a payment. It is four conditions. Where all four hold, a person with options stays. Where one is missing, they leave eventually regardless of what you pay — it simply takes longer, and costs more when it arrives.
Gallup puts the cost of replacing someone at one-half to two times their annual salary — for a plant head, a number you would refuse on a purchase order, and one you pay anyway each time a condition quietly fails.
Why money is the fourth thing, not the first
Culture Amp analysed 175 teams of at least eight people each, across hundreds of fast-growing companies, and sorted the reasons people actually left. Development opportunities drove 52% of departures. Leadership drove 28%. The direct manager, 12%. Pay, 11%.
That ordering inverts the instinct. The thing you can buy sits last; the thing that costs almost nothing to build sits first, at roughly five times the weight of pay.
It is also why a decision taken at the resignation rarely holds. By then the person is at the end of a sequence that began nine to fourteen months earlier with something they asked for and never got an answer to — the mechanism in why your best people leave. What follows is the other side of it.
The Four Anchors
For any given person, each is present or absent. There is no partial credit, which is the point of using them.
A next thing that exists
Not a promise. Not we'll see how the year goes. A named next role or next scope, with a rough timeframe, that the person can repeat back to you in one sentence.
Most promoters believe they have done this because they have said encouraging things. Encouragement is not an anchor. "You are very important to this company" is a sentence with no date in it. A capable thirty-eight-year-old with a home loan is doing arithmetic about where he will be at forty-five, and vagueness reads as an answer — just not the one you intended.
How to install it. Construct the next step out loud, in three parts: what it is, roughly when, and what has to be true for it to happen. Where the honest answer is that no seat exists — the role is held, or held by family — say so and build something else. A P&L for a product line. A new geography. The plant they set up. A capable person can absorb "there is no seat above you for four years, so here is what we will build instead". What they cannot absorb is three years of implication followed by an outside hire.
How to check it. Can they say in one sentence what is next for them here? Can you? Both answers must exist and must match. If yours exists and theirs does not, you have thought about it and never said it, which for retention purposes is the same as not having thought about it.
Authority that is real and bounded
A stated set of decisions this person makes alone, and that you do not reverse.
The boundary matters more than its size. A purchase authority of ₹2 lakh that is genuinely theirs anchors better than ₹20 lakh that is theirs until you happen to disagree, because the first is a fact about the company and the second is a fact about your mood that day. People calibrate to the second within two months and stop deciding anything.
How to install it. Write the boundary down, numeric or categorical. Spend up to this. Hire at this level without me. Approve a discount to this floor. Then hold it the first time they use it in a way you would not have — which is when the anchor is set or destroyed. Review a decision afterwards if you must; reverse it in front of others and nothing remains.
How to check it. Name the largest decision this person made last quarter that you did not review. If nothing comes to mind, the authority is notional whatever the delegation matrix says, and they know it even if the matrix does not. What that does to a whole team is in why your employees don't take ownership — the same structure, seen from the other side.
Consequence that arrives
Their outcomes visibly attach to them, in both directions, close enough in time to be connected to the act.
The asymmetry in most Indian promoter-led businesses is nobody's design. The downside is fast, sharp and public — a mistake is known in the building by lunchtime, often mentioned in front of two other people. The upside is slow and diffuse — a good year becomes an increment in April, bundled with everybody else's, in a letter with no sentence about what this person did. That teaches a capable person not laziness but precision: the safest strategy here is to minimise visible error rather than produce a visible result. Your quietest senior people learned it best.
How to install it. Attach the outcome by name, close to the event. If the Chennai account was saved, say who saved it, to the room, that week. If the launch slipped, say so to the person who owned it, in private, that week. Same speed, both directions.
How to check it. Take the last significant thing that went well in this person's area and ask whether anyone outside your cabin could name who was responsible. If credit dissolves into "the team", the anchor is absent — and you see it first in your strongest people, for the reasons in why high performers leave.
A conversation that happens before the resignation
Twice a year, conducted by you and not by HR, with two questions.
What have you asked me for that has not happened? And what is next for you here?
Those two, in that order, and nothing else in that hour.
The first finds the unanswered ask while it is still cheap, because people stop repeating asks long before they stop having them. The second forces the development question open while you can still act on it. Both come from the promoter, because in a company this size he is the only person whose answer is a commitment.
Do not do this at scale with an engagement survey instead. A survey asks for an opinion and produces no consequence: the person says something, nothing visibly happens, and the exact pattern that caused the withdrawal repeats, this time with your signature on it. Anonymity makes it worse, because nobody can act on an anonymous ask. The same defect at the other end of the timeline is why exit interviews tell you nothing.
Gallup found that 52% of voluntarily exiting employees say their manager or organisation could have done something to prevent them from leaving. This conversation is the cheapest way to learn what that something was, about a year before it reaches you as a number on an offer letter.
What to do about money, honestly
Pay matters. Anybody who tells a promoter otherwise is selling a workshop.
Gallup found 44% of employees would consider a job elsewhere for a raise of 20% or less — 54% among the actively disengaged, 37% among the engaged. Even among committed people, more than a third are movable at a fifth more money. Pay is rarely the reason they go, but it is often the door they leave through — see do employees really leave for money.
Two rules follow, and they are not the ones you expect.
Fix genuine market gaps proactively, without being asked. The most avoidable loss in an Indian mid-market business is the long-tenured person whose increments compounded slower than the market. Eight percent a year on a base set in 2016 does not track a role priced fresh in 2026. He has not complained, because he is loyal and asking feels like a threat. Then a lateral hire mentions a number in the canteen and a fifteen-year employee learns in ninety seconds what his loyalty cost him. Run that check yourself, once a year.
Then stop using money as the retention lever. Once the market gap is closed, more money buys months, because it touches none of the four anchors. And paying a retention premium after a resignation does worse than fail: it teaches the organisation, accurately, to produce offer letters. You will meet that lesson again across the same desk — the arithmetic is in should you counter-offer.
Why the instruments on that search page answer a different question
They are not frauds. They answer a question you did not ask.
Key-person insurance protects the company from the loss. If your technical head dies or is disabled, the policy pays the business for the disruption. That is useful, and if one person carries a disproportionate share of your revenue you probably should hold it. It retains nobody. It is a hedge against losing them, not a reason to stay.
Deferred compensation and golden handcuffs retain the body, not the investment. A vesting date holds someone in the building. It does not restore the ask you never answered or create the next step that does not exist. Someone locked in by a vesting date is often already at stage two of the Exit Clock: still delivering, no longer investing beyond the role, waiting for a date. You have bought eighteen months of compliance, and you get the resignation anyway, the week after the tranche clears.
These instruments belong in succession planning, shareholder agreements, valuation and continuity. They have no place in a retention conversation, and that page is full of them because the firms writing it sell them.
Who to apply this to
Not everyone. Name the three to five people whose departure would genuinely hurt — the ones who take a customer relationship, a plant, a capability or a decade of institutional memory with them. Write the names down and run the Four Anchors on those individuals, by name.
A retention strategy applied to all 140 people is a policy, and policies do not retain individuals. They produce a wellness initiative, a revised leave structure and a poster, while your four names each have a different missing anchor. Include your second line even where they look settled — managers leave more quietly and cost a promoter-led business most.
How to test it
Twenty minutes, this week, in writing.
Take your three to five names. For each, score the four anchors present or absent. No "partly", no seven out of ten. The middle rating is where promoters hide, and anything you want to score as "somewhat present" is absent — the test is whether the person would call it present, not whether you would.
Is there a named next thing, with a rough timeframe, they could state in one sentence?
Is there a written decision boundary they own, that you have not reversed?
Did a consequence arrive for them last quarter, attached to their name?
Have you had the two-question conversation with them in the last six months?
Any person with a blank is at risk regardless of how content they appear. Contentment is not evidence here. It is frequently the visible form of stage two, where a person has stopped investing beyond the role and become easier to manage. The behavioural signs are in employee flight risk indicators.
Redo the grid every six months. The blanks move.
Where to start
Take the name at the top of your list and fix the anchor that is blank. Not all four, and not for all five people. One anchor, one person, this month. If two are blank, start with the next thing: it is the largest driver in the evidence and the cheapest of the four to build.
If you want the structural version — which roles have no next step attached to them at all, where authority is written down but not honoured, which senior people are priced below their role — the Business Pulse diagnostic maps it. Fix that version before it becomes four minutes in your cabin on a Tuesday.