Centralize or Decentralize? The Operating Model Decision Every Multi-Business Owner Has to Get Right

Warren Buffett runs Berkshire Hathaway — a company with hundreds of thousands of employees across insurance, railroads, energy, manufacturing, and retail — from a headquarters in Omaha that has historically held around two dozen people. There is no central marketing department, no corporate strategy office dictating to the operating companies, no HR function reaching down into the businesses. The managers run their companies; the center allocates capital and otherwise stays out of the way.

Now hold that image next to Inditex, the parent of Zara. From a single complex in Arteixo, Galicia, the company centrally controls design, inventory, logistics, and the twice-weekly decisions about what ships to thousands of stores worldwide. A store manager in Tokyo does not decide what to stock; the center decides, and the center is astonishingly good at it.

Two of the most admired companies of the last fifty years, run on opposite philosophies. That is the whole point. The question is never "should I centralize or decentralize?" in the abstract — asked that way, it has no answer. The real question is far more specific: **which functions belong at the center, which belong at the edge, and how do I keep that boundary from drifting in the wrong direction as the group grows?** Running eight businesses as different as wine, real estate, hospitality, mineral water, electricity, and marine forces you to answer it deliberately, because the default — drift toward the center — is almost always wrong.

## Why the Default Drift Is Toward Over-Centralization

Left alone, a holding company centralizes. Not by decision, but by gravity. Every time something goes wrong in a business unit, the instinct at the top is to add a control: a new approval, a new report, a corporate policy to make sure "that never happens again." Each control is individually reasonable. Collectively, they suffocate.

I have watched this happen and I have done it myself. A distributor relationship sours, so you require head-office sign-off on distributor contracts. A hire goes badly, so you centralize senior recruiting. A unit overspends, so every expense above a threshold now routes upward. Within a few years the people closest to the customer cannot move without permission, and you — the founder — have become the bottleneck you swore you would never be.

The cost of over-centralization is invisible on a spreadsheet, which is why it wins. You can measure the deal that lost money because a local manager had too much rope. You cannot easily measure the three good deals that never happened because that manager stopped bothering to bring ideas to a center that says no. The decentralized model fails loudly and occasionally; the centralized model fails quietly and constantly. Founders systematically overweight the loud failures.

## The Test: Is This Function a Source of Edge or a Source of Risk?

The framework I use is a single question applied function by function: **does centralizing this create competitive advantage, or does it only reduce variance?** Centralize the first kind aggressively. Leave the second kind at the edge.

Some functions genuinely get better when shared — they have economies of scale, scarce expertise, or compounding learning:

- **Capital allocation.** This is the one function that must live at the center, always. The whole logic of a diversified group is moving cash from businesses that generate more than they can reinvest to businesses that can earn a higher return. A unit manager cannot make that call — they only see their own business. I decide where the group's euros go.

- **Cost of capital and financing.** A group can borrow more cheaply and negotiate better terms than eight standalone businesses. Treasury, banking relationships, and the balance sheet belong at the center.

- **Brand architecture and reputation.** When businesses share a name or a heritage, no single unit can be allowed to spend down the trust that took generations to build. Standards for what the name stands for are set centrally.

- **Senior talent and succession.** The decision of who runs each business is the highest-leverage decision the center makes. You can decentralize almost everything except the choice of who you decentralize it to.

And some functions get worse when centralized, because they depend on local knowledge, speed, and ownership:

- **Customer and commercial decisions.** Pricing nuance, client relationships, the read on a local market — these die in a head-office queue. The person who loses if it goes wrong should be the person who decides.

- **Day-to-day operations.** How a winery schedules its harvest, how a hotel staffs a busy weekend, how a water plant runs its line — the center has no useful opinion and should not pretend to.

- **Product and creative choices inside each vertical.** A real-estate development and a marine business and a hospitality venue do not share a playbook. Forcing one is how you get bland, identical, soulless units.

The discipline is to be ruthless about the difference. Danaher — one of the most successful industrial groups in history — is sometimes described as decentralized, but that is only half true. It pushes operating decisions down hard while imposing one thing centrally and without exception: the Danaher Business System, a shared toolkit and management language every unit must use. Strong center on the few things that compound; strong edge on everything else. That combination, not pure decentralization or pure control, is what actually scales.

## Autonomy Has to Be Paired With Accountability

Decentralization is not the absence of control — it is the substitution of *output* control for *input* control. The center stops telling the unit how to operate and starts holding it rigorously accountable for results. Remove that pairing and "autonomy" just becomes a budget for unmonitored mistakes.

In practice this means a small number of non-negotiables for every business, no matter how independent:

- A clear owner who is unambiguously accountable for the unit's results — no committees diffusing responsibility.

- A short set of metrics the center actually watches, reported on a fixed cadence, the same way every period.

- Hard limits that protect the group: no leverage above an agreed line, no commitment that could threaten the whole if it fails, no spending of the shared brand's reputation.

- An honest, regular conversation about capital — what the unit is returning, what it is asking for, and why.

Inside those guardrails, the manager runs their business. Outside them, the center acts immediately. The guardrails are few enough to remember and firm enough to mean something. That is the difference between trust and abdication.

## Re-Drawing the Line as You Grow

The boundary is not set once. The right amount of centralization for a group of three businesses is wrong for a group of eight, and wrong again at twenty. Early on, the founder is the central function — you allocate the capital, you set the standards, you make the senior hires, because there is no one else. The danger is keeping that posture long after the group has outgrown it.

The signal that you have over-centralized is simple and behavioral: good people stop bringing you their best ideas, because experience has taught them the center will slow it down or say no. When that happens, the fix is not a better approval process. It is to give the decision away — to build the capability at the edge and then genuinely let go of it. A group that has founded businesses across more than 75 countries cannot be run from one desk, and pretending otherwise does not make you more in control. It makes you the constraint on your own enterprise.

## Key Takeaways

- "Centralize or decentralize?" is the wrong question. Berkshire (radically decentralized) and Inditex (tightly centralized) are both spectacularly successful — the real question is which functions belong at the center and which at the edge.

- Holding companies drift toward over-centralization by gravity: every problem tempts the center to add a control, and the controls compound into paralysis. The cost is invisible on a spreadsheet, which is why it wins.

- Apply one test per function: does centralizing it create competitive advantage, or only reduce variance? Centralize the first kind, leave the second at the edge.

- Always centralize capital allocation, financing, brand/reputation standards, and senior-talent decisions — these have scale or compounding effects no single unit can capture.

- Always decentralize customer and pricing decisions, daily operations, and product/creative choices — these depend on local knowledge and ownership that head office cannot replicate.

- Decentralization replaces input control with output control: pair autonomy with a clear owner, a few watched metrics, hard group-protecting limits, and an honest capital conversation. Without accountability, autonomy is just unmonitored risk.

- Re-draw the boundary as you scale. The warning sign of over-centralization is that good people stop bringing you their best ideas — the fix is to build capability at the edge and genuinely let go.

The hardest part of this is not the analysis; it is the temperament. Every fiber of a founder who built something wants to stay in the decisions. But the goal of an operating model is not to maximize your control — it is to maximize the number of good decisions made across the group, most of which you will never see. Get the line right, and the businesses run faster than you ever could from the center. Get it wrong, and you become the most expensive bottleneck in your own company.

*Meta description: Centralize or decentralize? The wrong question. Learn which functions a multi-business owner should centralize (capital, brand, talent) and which to push to the edge — and how to keep the line right as you scale.*

*SEO keywords: centralization vs decentralization, operating model, multi-business management, holding company strategy, capital allocation, organizational design, business unit autonomy, accountability framework*

*Meta description (ES): Centralizar o descentralizar? La pregunta equivocada. Descubre que funciones debe centralizar el dueno de varios negocios (capital, marca, talento) y cuales empujar al borde — y como mantener la linea correcta al crecer.*

*SEO keywords (ES): centralizacion vs descentralizacion, modelo operativo, gestion multinegocio, estrategia holding, asignacion de capital, diseno organizativo, autonomia unidades de negocio, rendicion de cuentas*