The politics of centralisation, and how to survive them
The first centralisation proposal I wrote was technically excellent. Consolidated media buying, one vendor panel, a single content engine, shared analytics, a group contact centre. It would have reduced duplicated spend and raised the floor on quality at every unit. It was also, from a unit head’s chair, a proposal to take away three of the few levers they still controlled over their own numbers.
It went nowhere for two quarters. Not through open opposition — nobody says no to a group function in a steering committee. Through something harder to fight: polite agreement followed by nothing happening. Data requests answered late. Local agencies quietly retained. A camp run last weekend that nobody mentioned.
Centralisation in a multi-unit hospital group is not an operating-model question. It is a question about who is accountable for a unit’s performance and what they are allowed to do about it. Until you answer that honestly, every structure chart you draw is a wish.
Start with who carries the number
In most Indian hospital groups, the unit head — whatever the title, COO, facility director, cluster head — carries occupancy, outpatient footfall, revenue and cost for that facility. They are reviewed on it monthly. Their increment depends on it.
You, in a group function, usually carry none of it. You carry programme delivery, spend efficiency and perhaps enquiry volume. That asymmetry is the whole politics.
When a unit head resists centralisation, the substance of their objection is almost always the same, whether or not they articulate it: you are taking control of an input to a number I am accountable for, and when it goes wrong I will still be the one in the review meeting. That is a legitimate position. Any argument you make that ignores it is heard as evasion.
So establish, early and explicitly, what you are accountable for. I learned to say it in one line in the first meeting: if the centralised function fails to deliver what the unit needs, that is my failure and I will say so in the same forum where you are reviewed. Saying it is cheap. The unit will check whether you do it, usually within a quarter.
The flagship and the small unit want opposite things
Group functions tend to treat the units as one constituency. They are at least two, with opposed interests, and your proposal reads completely differently to each.
The flagship in a Tier 1 city usually has its own marketing headcount, its own agency relationships, a strong local brand and a unit head with seniority and board visibility. Centralisation offers them very little. Whatever you provide centrally, they can already buy at a quality they consider acceptable, and a group standard is more likely to constrain them than to help them. Expect the real resistance here, and expect it to be competent.
The smaller unit, often in a Tier 2 city, has one marketing executive handling camps, collateral, the local paper, the front desk complaint and whatever the unit head asked for that morning. They are drowning. Centralised production, regional-language creative and analytics are a genuine gift.
The practical consequence: build your case with the smaller units, where the value is obvious and the gratitude is real, and let the flagship come last. If you lead with the flagship because it is the biggest number, you will spend your political capital at the least winnable point and arrive at the easy units with none left.
What genuinely should be centralised
The test I use: centralise where fragmentation creates either risk or a cost that no single unit can see.
- Anything carrying clinical or regulatory claim. Medical content, treatment descriptions, outcome language, consent wording on digital forms. A unit marketing executive in a Tier 2 city writing their own treatment page is an accreditation and liability exposure, not a creative difference.
- Identity and the truth of listings. Entity data, map listings, doctor profiles, timings, the phone numbers published anywhere. This is where local autonomy does the most damage and the damage is invisible at unit level.
- Technology and data architecture. One website platform, one CRM, one analytics definition, one integration path into the hospital information system. Ten units with ten stacks is not flexibility, it is ten vendor relationships and no group-level view.
- Media buying mechanics and the vendor panel. Rate negotiation, platform access, fraud controls, consolidated billing. Note the word mechanics — not the decision about what to promote.
- Specialist capability nobody can afford alone. Analytics, performance marketing expertise, design craft, regional-language production at volume.
What should stay local, and you should defend it staying local
This is the part group functions get wrong, usually out of tidiness.
- The catchment. Which colonies, which feeder towns, which competitor is pulling orthopaedic cases this quarter. The unit knows this and you do not.
- Doctor referral relationships. The general practitioners and nursing homes that send cases. These are personal, built over years, and centralising them destroys them.
- Camps, corporate tie-ups and community activity. Standardise the format, the collateral and the lead capture. Do not decide from head office which housing society gets a camp on which Sunday.
- Local language nuance and festival context. Translate centrally if you must; approve locally, always.
- Front-of-house patient experience. Queue management, signage placement, who walks a nervous patient to radiology. Central dashboards do not fix this.
Announcing what you will not centralise is one of the most effective political moves available to you. It converts your proposal from a land grab into a division of labour, and it costs you nothing you actually wanted.
The money question you cannot dodge
Sooner or later somebody asks where the budget sits, and the answer determines whether centralisation is real.
Three arrangements, in ascending order of difficulty and effectiveness.
Unit holds budget, group advises. Easy to agree, nearly useless. You will spend your time chasing units to follow recommendations they have no incentive to follow, and your group programmes will be funded out of whatever is left.
Group holds a central pool, units hold local spend. This is where I would start, and where I have seen it work. The pool funds the things no unit will fund alone — platform, content engine, analytics, brand. Local spend stays local and remains the unit head’s lever. Arguments then happen at the boundary, which is manageable.
Group holds all marketing spend with unit allocation. Most efficient, most politically expensive. Do not attempt it in year one. If you get there, get there because two unit heads asked for it, having found that central buying gave them more for the same money.
Whichever you choose, publish the rule. Ambiguity about who can commit money is what produces the quiet parallel spending that tells you centralisation has failed.
The phasing that avoids open revolt
Sequence matters more than design. The order that has worked for me is roughly this.
Phase one: take over what nobody wants. Listings hygiene, vendor payment follow-ups, reporting that units currently assemble by hand, responding to online reviews. You are absorbing work, not authority. Nobody defends a chore.
Phase two: become the cheapest option. Offer production — creatives, regional-language versions, landing pages, camp collateral — faster and at lower cost than the local agency, funded centrally. Units switch for their own reasons. You have now centralised production without a single mandate.
Phase three: standardise where you already deliver. Once most units use your production service, introduce the template, the approval flow, the brand rules. Standards imposed on work you are already doing for people are accepted. Standards imposed on work others are doing are resented.
Phase four: take the contested things. Technology platform consolidation, CRM, contact centre routing, media consolidation. By now you have a track record, and at least two unit heads who will speak for you unprompted. That is the asset you have been accumulating.
What I would do differently: I tried to do phase four work in phase one at one group, because the business case was obvious and the waste was real. Being right was not sufficient. The proposal became a thing to be negotiated rather than a service to be adopted, and negotiating took longer than simply earning it would have.
The service-level promise you must keep
Compliance is purchased with reliability. If you take something from a unit and then become a queue, you will lose it back and you will not get a second chance.
The promise needs to be written, specific and boring.
- A named person per unit, with a mobile number, who answers. Not a shared inbox.
- Stated turnaround times by request type — a creative amendment in two working days, a new landing page in a week, an emergency listing correction the same day.
- An escalation path with names, and a commitment that escalation carries no penalty for the unit that uses it.
- A weekly call per unit that you never cancel, even when there is nothing to discuss. Cancelling it twice tells the unit where they rank.
- A monthly published record of your own performance against these commitments, including the misses.
That last one is the one group functions skip. It is the one that works. A unit head who sees you report your own missed turnarounds stops assuming you will hide failures, and that assumption is the main reason they want to keep control.
How you lose it back
Centralisation decays quietly. The signals are consistent and worth watching for deliberately.
- A unit raises a vendor purchase order for something your team already provides.
- You learn about a camp or a local campaign after it happened.
- A unit marketing executive stops attending your weekly call and sends apologies.
- Your data requests start coming back late, then incomplete, then reformatted.
- A unit head copies your manager on an operational request instead of your team.
Each of these is a service failure on your side before it is a compliance problem on theirs. Treat them that way, out loud, and you can usually recover. Treat them as insubordination and you will be fighting the same battle for years.
If you’re starting this next quarter
- Map who carries which number at every unit, and how they are reviewed. Do this before drawing any structure.
- Write down, publicly, the list of things you will not centralise. Circulate it before your first proposal.
- Take over two chores in month one. Do them visibly better than they were being done.
- Stand up production as a free, fast internal service. Resist attaching conditions to it for at least a quarter.
- Publish the service-level promise and your performance against it monthly, misses included.
- Settle the budget rule in writing with finance — central pool versus local spend — and stop relitigating it.
- Find the two unit heads who are getting the most value and let them make the case for the contested phase. Your version of that case is worth less than theirs.
- Only after that, touch the technology stack and the contact centre.
No unit head has ever been won over by an operating model. They are won over by a quarter in which your team made their worst Monday easier.
Questions people ask
Consolidating the things that fragmentation makes risky or invisible — medical content, listings and identity data, the technology stack, media buying mechanics, and specialist skills no single unit can afford — while leaving the catchment, referral relationships, camps and local language nuance with the unit. It is not an operating-model question. It is a question about who is accountable for a unit’s number and what they are allowed to do about it.
Because they carry occupancy, footfall, revenue and cost for the facility, are reviewed on it monthly, and their increment depends on it. A group function usually carries none of that. Centralisation takes control of an input to a number they own while leaving them in the review meeting when it goes wrong. That is a legitimate objection. Resistance rarely comes as open opposition; it comes as polite agreement, late data and quietly retained local agencies.
The catchment knowledge — which colonies, which feeder towns, which competitor is pulling cases this quarter. Doctor referral relationships, which are personal and die when centralised. The decision about which housing society gets a camp on which Sunday. Local language approval and festival context. Front-of-house patient experience. Publishing this list before your first proposal converts it from a land grab into a division of labour, at no cost to anything you actually wanted.
Plan on four phases across roughly eighteen months to two years. Month one, take over chores nobody defends. The first two quarters, become the cheapest production option and let units switch for their own reasons. Then standardise the work you already do. Only after you have a track record and two unit heads who will speak for you unprompted should you touch the contested things — CRM, contact centre routing, media consolidation. Attempting phase four in phase one takes longer than earning it.
Start with a central pool for what no unit funds alone — platform, content engine, analytics, brand — and leave local spend as the unit head’s lever. “Unit holds, group advises” is easy to agree and nearly useless. “Group holds everything” is most efficient and most politically expensive; do not attempt it in year one. Whatever you choose, write the rule with finance and publish it. Ambiguity about who can commit money is what produces quiet parallel spending.
No, even though it is the biggest number. The Tier 1 flagship has its own headcount, agency relationships, a strong local brand and a unit head with board visibility; centralisation offers it little and the resistance will be competent. The Tier 2 unit with one marketing executive handling camps, collateral and front-desk complaints is drowning, and central production is a genuine gift. Build your case there and let the flagship come last.
A named person per unit with a mobile number who answers, not a shared inbox. Stated turnaround by request type — a creative amendment in two working days, a landing page in a week, an emergency listing correction the same day. An escalation path that carries no penalty. A weekly call you never cancel. And a monthly published record of your own performance against these, including the misses. That last one is the one that works.
Watch for five signals. A unit raises a purchase order for something you already provide. You learn about a camp after it happened. The unit’s marketing executive stops attending your weekly call. Data requests come back late, then incomplete, then reformatted. A unit head copies your manager on an operational request. Each is a service failure on your side before it is a compliance problem on theirs. Treat them that way, out loud, and you can usually recover.
They should be relieved, not removed. In most groups the unit marketing executive is one person handling everything from camps to complaints, and central production takes the collateral, regional-language versions and reporting off them. What remains is the local work only they can do: catchment intelligence, referral relationships, community activity. Redesign the role around that before centralising, and say so. A unit head who thinks headcount is the target will fight everything else.
One website platform, one CRM, one analytics definition and one integration path into the hospital information system. Ten units with ten stacks is not flexibility; it is ten vendor relationships and no group-level view. Identity and listing data — entity names, phone numbers, timings, doctor profiles — needs a single source of truth, because local autonomy does the most invisible damage there. This is late-phase work, though; earn it before you ask for it.
When you cannot keep the service promise. If you take something from a unit and become a queue, you lose it back and do not get a second chance. It is also wrong when the group has not settled who carries which number — centralising an input to an unowned outcome is theatre. And centralising referral relationships or catchment decisions is always wrong; those are the unit’s, and taking them destroys value rather than consolidating it.
Reduced duplicated spend, a floor on quality at every unit, and control of clinical and regulatory claims that are currently an accreditation and liability exposure at unit level. What it should not expect is speed. The business case is obvious; being right is not sufficient. Ask which phase the function is in, which two unit heads would speak for it unprompted, and whether the budget rule is written. If none has an answer, the structure chart is a wish.
Because a unit marketing executive in a Tier 2 city writing their own treatment page is an accreditation and liability exposure, not a creative difference. Treatment descriptions, outcome language and consent wording on digital forms carry clinical and regulatory claims, and the medical director needs one approval flow rather than eleven. This is the easiest centralisation to win with clinicians, since it protects them, and it should not wait for the contested phase.
Less structure, same principles. Centralise the things that carry risk — medical content, listings, the phone numbers published anywhere — and the technology, even at three units, because adding a fourth is far harder on three separate stacks. Leave the rest local and skip the phased politics; with three unit heads you can settle the budget rule in one meeting. The mistake small groups make is postponing the platform decision until the fifth unit.
