What to centralise in a group — and what to leave alone

Every multi-company group discovers the same arithmetic. Eight companies do not need eight finance departments, eight HR functions, or eight legal reviews of the same supplier contract. Consolidating those saves real money and removes real duplication.

Then the group over-applies the logic, centralises something that should have stayed local, and spends the next two years wondering why everything takes longer than it used to.

The savings from centralisation are easy to calculate and appear on a budget line. The costs are diffuse — a slower decision here, a customer commitment nobody could authorise there — and rarely get traced back to the reorganisation that caused them. That asymmetry is why groups tend to centralise too much rather than too little.

The test

One question separates the two categories reliably:

Does doing this well require knowing the customer, or does it require knowing the rules?

Functions that depend on knowing the rules — accounting standards, employment law, tax treatment, security policy, contract terms — get better when consolidated. The rules are broadly the same regardless of which company applies them, expertise is expensive and hard to hire, and consistency is a genuine benefit. One person who knows Kuwaiti labour law properly serves the whole group better than five people who each know a bit.

Functions that depend on knowing the customer — pricing, service design, delivery commitments, hiring for operating roles — get worse when consolidated. The knowledge is specific, changes quickly, and lives with the people who face the customer daily. Moving those decisions away from that context reliably makes them worse, however senior the person receiving them.

The test is not perfect, but it resolves most cases without argument, which is more than most operating-model debates achieve.

What usually belongs at group level

Finance and treasury. Consolidated reporting, banking relationships, audit, and cash management. The rules are identical across companies and the expertise is genuinely scarce. Note the boundary: consolidated reporting belongs at the centre; a company's budget ownership should not leave the company.

Legal and compliance. Contract templates, corporate structure, regulatory filings, disputes. Nobody benefits from eight companies independently inventing a services agreement, and inconsistency here creates liability rather than differentiation.

Technology infrastructure and security. Identity, devices, network, backups, security policy. The rules are the same everywhere and the weakest link sets the group's actual security posture — which is the strongest argument for consolidation anywhere on this list.

Brand standards. Not marketing execution — standards. What the name may be attached to, and what quality bar it implies. A shared brand is only an asset if someone owns the standard behind it.

Capital allocation. The defining function of the holding layer and the one thing that cannot sit inside an operating company. A subsidiary cannot fairly evaluate its own claim on group funds against a sibling's.

What usually should not

Pricing. It depends on competitive position, customer relationship, and market conditions that differ by company and country. A central pricing committee is almost always slower and worse-informed than the person negotiating.

Service design. What a logistics service should include and what a software engagement should include have nothing in common. Standardising them produces a compromise that serves neither.

Operating hiring. The centre can own the process, the bands, and the standards. It should not own the judgement of whether a specific person is right for a specific team.

Customer commitments. If a client asks for something unusual and the answer requires three approvals, you have already lost the responsiveness that made the group worth buying from.

Local marketing execution. Standards travel; execution does not. What works for Egyptian e-commerce buyers is not what works for a Kuwaiti corporate procurement team.

The failure modes, named

The centre becomes a queue. Consolidation without capacity means the shared function is now a bottleneck for every company at once. The savings are real and immediate; the delay costs are real and delayed, which is exactly the wrong order for noticing a mistake.

The centre stops being accountable. When a shared service is the only option, it has no competitive pressure. Groups that avoid this give companies a stated service level and a real escalation path — the internal equivalent of being able to leave.

Centralisation as a control instrument. Sometimes consolidation is proposed for efficiency and adopted for control. It is worth being honest internally about which is happening, because a control decision dressed as an efficiency decision produces a badly-designed shared service and a cynical organisation.

The reverse: duplication defended as autonomy. The opposite failure is real too. Eight separate contract templates is not autonomy, it is eight chances to sign something bad.

A sequencing note

Groups usually get better results centralising in the order the pain arrives, rather than all at once.

Finance and legal are typically first, because inconsistency there creates risk immediately. Technology infrastructure follows, usually after the first security scare. Brand standards tend to come when the group's name starts appearing on things it did not vet. Capital allocation formalises last in many groups, though arguably it should be first — it is the function the holding layer exists to perform.

Attempting all of it in one reorganisation reliably produces the queue problem, because the centre has to build five capabilities simultaneously while every company waits.

How this works at PNM

The group runs shared functions — finance, legal, technology, talent, and brand — across its operating companies, while sector decisions stay with the companies that own them. A freight operation and a software company are not run the same way, and pretending otherwise would make both worse.

The ecosystem page describes how the companies connect in practice, and the operating principle behind the split is set out under integration across sectors: the companies operate independently, but share a common backbone. The holding layer's own job — of which shared services is one part — is covered in what a holding company actually does.

Three questions before centralising anything

Who is worse off if this moves, and have we asked them? The company losing the function usually knows the failure modes first, and usually is not consulted until after the decision.

What is the service level, and what happens when it is missed? A shared service without a stated commitment becomes a queue by default rather than by decision.

Are we doing this to save money or to gain control? Both can be legitimate. Confusing them is not, and the resulting design will show it.

Final thought

The instinct to consolidate is sound; groups exist partly to capture exactly this. The error is treating centralisation as a direction to keep moving in rather than a choice to make function by function.

The right shape is uneven — heavily centralised in some places, deliberately untouched in others, and specific about why in each case. A group that can explain the reasoning for every function on the list has an operating model. A group that centralised everything has a headquarters.