One Programme, Fifteen Rulebooks: Why Cross-Border Delivery Breaks Where the Textbook Says It Won’t
Standardise the plumbing; vary the fixtures.
The seventh country
The programme was, by its own reporting, a success. Eight months in, a large financial-services group had a single customer-onboarding process — designed, documented, signed off. One target operating model to replace the fourteen that had grown up country by country over two decades. The business case rested on that arithmetic: fourteen down to one. The steering pack was green. The central team had, in fairness, done good work.
Then the design met the seventh country.
There, a customer relationship was not legally formed at the moment the central process assumed it was. A signature captured on a screen did not carry the weight the design took for granted. And a category of identity document the process treated as routine could not, under local law, be copied and moved to the shared operations centre at all. None of this was exotic. It was simply true there, and it had been true there long before the programme existed. What had changed was only this: a design conceived as universal had finally been asked to touch ground in a place that never agreed to be universal.
The country team was told, as country teams always are, to raise a local variation request. By the close of the programme there were more than two hundred of them.
I want to sit with that moment, because it is not the misfortune of one programme. It is the signature failure of cross-border delivery, and it recurs so reliably that it deserves to be understood as a pattern rather than absorbed, each time, as bad luck.
The variation request is the instrument through which a programme quietly concedes that its central design was never universal — while carrying on as though it were.
The model everyone reaches for
The dominant approach to multi-jurisdictional delivery is so familiar that it is rarely argued for; it is simply assumed. Design once, at the centre. Build a common core. Then roll out to each country, absorbing whatever is genuinely local as adaptation at the edge. The promise is usually rendered as a ratio — ninety per cent common, ten per cent local — and the whole apparatus of the programme is shaped to deliver it: a strong central design authority, a set of local implementation teams whose remit is configuration rather than design, and, more often than not, a vendor operating model that mirrors the same shape — global template, local parameters.
It is worth asking why this model recurs with such force, because its persistence is not stupidity. It recurs because it is how the business case is written: the saving is the collapse from many processes to one, so the single process must be treated as the prize and every departure from it as leakage. It recurs because it flatters the centre, where the programme’s power and budget sit. It recurs because it matches how we procure and staff — a core team that designs, local teams that receive. And it recurs because the last great cross-border exercise most of these organisations lived through, this year’s scramble to adopt international financial reporting standards, really could be run that way: accounting rules that have been deliberately converged look, from a distance, like every other kind of rule.
The trouble is that most rules have not been converged, and the ones that decide a programme’s fate least of all.
Why difference is architecture, not translation
Here is the distinction the ninety-ten promise elides. Translation happens at the end and changes nothing structural — a screen relabelled, a currency swapped, a document set in another language. Architecture is decided at the beginning and constrains everything that follows. And the jurisdictional differences that break cross-border programmes are almost never matters of translation. They are matters of architecture.
Consider what national rules actually dictate:
- Where a record may physically reside — which decides the topology of your systems and whether a shared service centre is even lawful for that country’s data.
- When a contract is legally formed — which decides the sequence of the process, not merely its wording.
- What evidence must be captured and retained, in what form and for how long — which decides your data model and your storage, everywhere.
- What may be centralised at all, and how quickly — which is governed by labour law and, across much of the Union, by works councils whose agreement is not the programme’s to grant.
None of these can be bolted on at the edge. If you discover in month nine that one country’s customer data may not cross its border, you are not localising anything. You are re-architecting — the systems topology, the operating model, the location strategy — after the concrete has set.
And the reason it is discovered late is not carelessness; it is structural, which is why the standing exhortation to “engage the countries earlier” so rarely fixes it. The central design is validated against the centre’s own assumptions. The countries are enrolled as stakeholders to be managed rather than as co-authors of the design. And the governance rewards keeping the core stable, so every signal from a country arrives dressed as a change — something to be minimised, resisted, deferred — rather than as intelligence about the true shape of the problem. The organisation has, without meaning to, built a machine that is structurally deaf to precisely the information it most needs to hear early.
“A rule you discover in month nine is not a variation. It is an architecture you chose not to ask about in month one.”
Measured honestly — at the level of process steps that actually touch a regulator, a tax authority, or an employment regime — the genuinely common proportion of a cross-border process is rarely the promised ninety per cent. On the programmes I have seen dissected candidly, it sits closer to sixty. And the remaining, supposedly minor, forty per cent does not consume forty per cent of the effort. It consumes the majority of the risk and very nearly all of the schedule slippage, because those are the steps the programme cannot actually decide. They are decided by someone outside it, who was never in the room.
The objection that must be answered
There is a serious case against everything I have just argued, and it deserves to be put at full strength rather than waved away.
It runs like this. If every country is allowed to design its own version of the process, you have abdicated. You will end up with fifteen systems and no leverage — a federation of local baronies that share a logo and nothing else. You will have re-created, at ruinous cost, the very fragmentation the programme existed to remove, and you will pay to maintain fifteen bespoke builds for a decade. The discipline of a common core is exactly what prevents this. A “thin core” is not pragmatism; it is surrender with better manners.
This objection is right about the danger. Uncontrolled localisation is a real and expensive failure mode; I have watched a programme dissolve into fifteen bespoke builds that shared branding and little else, and the maintenance invoice arrived, punctually, for years afterwards. Anyone who has seen it does not need persuading that standardisation matters.
But the objection smuggles in a false choice — centralise everything, or localise everything — and the real question lives in the space it skips over. The question is not whether to draw a line between common and local. It is where that line goes, and, above all, how it is decided. The default model draws the line by central fiat and surface plausibility: surely onboarding is basically the same everywhere? The inverted model draws it by empirical jurisdictional analysis, done first.
And the economies of scale the objection rightly prizes are real — but they do not live where instinct places them. They do not live in the surface process that touches the regulator; that is exactly the layer where difference is irreducible and where forcing commonality does the most damage. They live one and two layers down: in the shared data model, in the integration contracts between components, in common reference data, in the shared infrastructure and shared services that never touch a national rule at all.
Standardise the plumbing; vary the fixtures. You can run one data platform beneath fifteen legitimately different contract-formation sequences and still capture the great majority of the leverage — while conceding the difference that was never yours to remove.
The inversion that actually delivers
The programmes that deliver across borders invert the default posture. Three moves define them.
- They hold a deliberately thin core. The core is the data model, the integration contracts, the shared services, and the genuinely universal controls — the handful of things a group regulator or auditor truly requires everywhere, whether that is a Sarbanes-Oxley control over financial reporting, a group risk standard, or a Basel II data requirement now bearing down on every large bank. Everything else is presumed local until proven otherwise — the reverse of the usual presumption.
- They push real design authority out to the jurisdictions for anything that touches a national rule, and they do it early, as co-design, not late, as variation. The country is treated as the author of its own layer, not the consumer of the centre’s.
- They govern the seams, not the centre. The programme’s real work becomes the interfaces — the contracts between the thin core and the thick edges, the hand-offs, the data that crosses a border — because that, not the centre, is where cross-border programmes actually break.
The most consequential change is one of sequence. Jurisdictional analysis moves to the very front — before the target operating model is drawn. For each country the programme maps the small number of architecturally decisive rules: data residency, contract formation, signing and identity, retention, and the labour and works-council constraints on centralisation. That map is cheap relative to the programme, and it is the highest-return activity on the whole plan, because it reveals the true shape of the common core before a penny has been spent building the wrong one.
| Dimension | Default: central design, local adaptation | Inverted: thin core, governed seams |
|---|---|---|
| Source of truth | The centre; countries are deviations | The seam; each country co-authors its layer |
| When difference surfaces | Late, as variation requests | First, as jurisdictional requirements |
| Where scale is sought | The surface process | The data model, contracts, shared services |
| The centre’s job | Own and defend the design | Govern the seams and the universal controls |
| Characteristic failure | Death by two hundred variations | Slack discipline at the interfaces |
It changes the governance, too. The local variation request — that revealing little instrument — should be retired as the primary mechanism, because it encodes the wrong belief: that country reality is an exception to a norm. In its place, jurisdictional requirements are owned by the countries and fed into the design as first-class inputs, standing alongside the centre’s. The centre’s authority does not vanish; it moves — from owning the design to owning the seams and the small set of controls that genuinely must hold everywhere.
None of this is free, and it would be dishonest to pretend otherwise. A thin core demands more discipline at the seams than a thick one demands at its centre, because interfaces are harder to govern than edicts. It demands a centre secure enough to relinquish surface control — which is a matter of temperament as much as of design. And it demands that the decisive jurisdictional analysis be done early and done well, by people who understand both the law and the architecture and who are, predictably, scarce. The inverted model asks more of the programme’s leadership than the default does. It simply asks it at the start, when the asking is cheap, rather than at month nine, when it is not.
What the seventh country was really telling us
The rule that stopped that programme was always there. It did not arrive; it did not change; no one sprang it as a surprise. The only real variable was whether the programme would meet it in month one, as intelligence, or in month nine, as damage — and that was settled not by the country but by the shape of the programme that went to meet it.
The competence that separates the teams who deliver across borders from those who stall is not superior central design. It is the humility to treat each jurisdiction as an author of the design rather than a recipient of it, joined to the architectural judgement to know which few rules are decisive and which are merely noise. Get that pairing right and the borders stop being the place where your programme goes to fail. They become, instead, simply part of its architecture — which is what they were all along.