Centralise to Save, Fragment to Fail: Why the Shared Services Promise Keeps Half-Failing
They are the same act, described first by its intention and then by its result.
Executive Summary
Consolidating the scattered back office into a single shared centre is one of the most durable ideas in modern administration, and one of the most reliably disappointing. Its logic is almost unarguable: gather the duplicated work spread across a dozen business units — the same invoice keyed a dozen different ways, payroll run on four systems that do not speak to one another — into one place, and let scale, standardisation and specialisation do their work. The first-year savings are real, and large enough that the business case very nearly writes itself.
Yet the pattern recurs with unsettling regularity. The consolidation books its headline savings, and then, quietly, the organisation begins to fragment again — not on the chart, where everything looks admirably tidy, but in the places the business case never measured. Shadow finance clerks reappear inside the divisions. The centre and the units it serves negotiate at arm’s length across service-level agreements that count activity rather than outcome. The judgement that once sat at the next desk now sits two floors away and one service request behind.
This essay argues that centralise to save and fragment to fail are not two problems but one, seen from two angles. Both flow from a single error: treating what is at heart an operating-model and relationship problem as though it were a matter of structure and efficiency. Redrawing the boxes is mistaken for changing how the work is done and governed. What follows traces why the promise is genuinely attractive, what forces keep the pattern alive despite its record, and the precise mechanism by which the savings leak back out — before turning to what the more durable implementations do differently, and to what the whole episode tells us about the gap between transformation as designed and transformation as lived.
The business case that always reads beautifully
Picture the steering group in the late autumn of the programme’s first year. The slides are good. The finance shared services centre stood up on schedule; three hundred and twenty desks in two locations replaced by a hundred and ninety in one; the run-rate saving, annualised, sits just north of the number promised in the original submission. There is a genuine, earned sense of achievement in the room, and it is not misplaced. The consolidation happened. The heads came out. The ledgers close.
Eighteen months later, a different meeting, a different mood. A divisional managing director wants to know why, having “given up” his finance team, he has quietly had to rebuild a small one — three analysts, a manager, sitting outside the centre — simply to get answers he used to get by walking down the corridor. The centre, for its part, reports that it is meeting every service level in its agreement: invoices processed within target, calls answered inside the threshold, the month-end timetable held. Both are telling the truth. That is the uncomfortable part. Nobody is failing at what they were asked to do; the thing is failing anyway.
The promise that got everyone into the room in the first place is worth stating in its strongest form, because it deserves respect. Across most large organisations, the same administrative processes are performed many times over, each in its own way, on its own systems, to its own local standard. The duplication is not a moral failing; it is the residue of how the organisation grew — by acquisition, by division, by the accretion of local fixes. The year-2000 remediation, only recently behind us, made the cost of that fragmentation impossible to ignore: the bills laid bare just how many near-identical systems each ran their own slightly different version of the same process. Consolidation answers all of this at once. One process instead of twelve. One platform to license, patch and support. A cadre of people who do the work often enough to become genuinely expert at it, with a career path that the isolated departmental clerk never had.
Why the promise is real
It is important not to caricature the idea in order to knock it down. The economics of the shared centre are not a delusion. Three effects are real, and they compound.
The first is scale. Transaction processing has a large fixed component — the system, the controls, the expertise — that is wasted when it is replicated in every unit. Pool the volume and the fixed cost is spread across far more transactions, and the cost per transaction falls. The second is standardisation. Twelve local variants of an accounts-payable process are not twelve considered designs; they are eleven accidents and one intention. Forcing a single standard removes not only the duplicated effort but a whole category of reconciliation work that existed only because the variants disagreed. The third is specialisation. A clerk who processes one difficult transaction a month never becomes good at it; a team that processes a hundred a day develops a competence — and a body of documented practice — that the dispersed model cannot match.
The case for consolidation is not wrong. That is precisely what makes the failure so hard to see coming: the savings the business case promises are genuinely there, and they genuinely arrive. The trouble is what the case leaves out.
Set against those gains, the argument for keeping the work dispersed can look sentimental — a preference for the comfort of a familiar face in the finance office over the discipline of a properly run operation. And in the first year, the numbers appear to settle the argument. The saving is banked. The case is vindicated. This is the moment at which the seeds of the second act are already in the ground.
The forces that keep the pattern alive
If the pattern half-fails so predictably, why does it recur? Not because practitioners are foolish, but because a set of structural forces makes this particular shape of programme almost irresistible, and makes its characteristic blind spot almost invisible.
- The business case is built from what is easy to count. Headcount, floor space, system licences — these are visible, attributable and bankable, and they all point one way: consolidate. The costs that consolidation creates — coordination effort, lost local knowledge, the friction of working through a service desk instead of a colleague — are diffuse, deferred and hard to attribute. A case built honestly from the countable will always favour the centre, because the countervailing costs never make it onto the page.
- The savings and the costs land on different people. The saving is central, immediate and owned by the sponsor of the programme. The cost is local, later, and borne by the divisions. When the party that captures the benefit is not the party that absorbs the cost, the appraisal is structurally optimistic, and the people who could see the trouble coming have no standing in the decision.
- The moment rewards it. In the downturn that followed the collapse of the dot-com boom, “do more with less” stopped being a slogan and became a survival instruction. A programme that promises a hard, early, defensible cost reduction is exactly what a board under pressure wants to approve. Shared services fits the mood of the moment, and the mood of the moment does not ask searching questions about year three.
- The template travels. The shared-services programme is one of the most heavily templated things an adviser can bring through the door: a reference operating model, a migration plan, a benefits schedule, all pre-built. Templates are efficient, but they carry the same silence everywhere they go. The one thing the template is weakest on is the very thing that determines whether the model holds — the relationship between the centre and the business, and the governance that has to hold it together.
Put these together and you have a programme that is over-recommended, over-approved and under-scrutinised on precisely the dimension that decides its fate. The forces do not push towards doing it badly; they push towards seeing only half the problem.
How the savings leak back out
The fragmentation is not a metaphor. It has a mechanism, and the mechanism runs in a recognisable sequence.
It begins with what is usually called migrating the work “as is”: the work is moved to the centre broadly unchanged, because moving it and redesigning it at the same time is judged too risky. So the duplication is consolidated but the underlying process is not truly reengineered; the centre inherits the variants it was meant to eliminate, and absorbs them as “exceptions.” Then the relationship is formalised. Because the centre no longer reports to the business it serves, the connection between them has to be written down, and it is written down as a service-level agreement: a list of activities, volumes and response times. The agreement measures what the centre does, not what the business needs. Processing an invoice within two days is in the agreement; making sure the invoice was the right one to pay is not.
Now the divisions discover that the standard service does not quite fit the non-standard reality, and they do the entirely rational thing: they rebuild, locally and quietly, just enough capability to bridge the gap — a business partner here, a couple of analysts there, a spreadsheet that shadows the centre’s ledger because the report the centre produces does not answer the question the division actually has. The saved cost reappears, off the programme’s books, dressed in different job titles. And because these roles are scattered and small, no one adds them up.
A composite is worth working through in numbers, because the shape of it matters more than any single figure.
| Line | Business case | What actually settled |
|---|---|---|
| Roles removed from the divisions | 130 | 130 (this part is real) |
| Roles in the centre (from 320) | 190 | 190 |
| Net roles removed | 130 | 130 on paper |
| Coordination roles quietly rebuilt in divisions | 0 | ~35 |
| Rework on non-standard transactions (extra centre effort) | 0 | ~15 roles’ worth |
| Effective net reduction | 130 | ~80 |
The headline saving was never fictitious. Roughly a hundred and thirty roles genuinely left the divisions; the centre genuinely runs on a hundred and ninety rather than three hundred and twenty. But some thirty-five coordination roles reappear in the units to manage the new distance, and the exceptions the centre absorbed consume perhaps another fifteen roles’ worth of effort that the standard cost model never anticipated. The realised saving is not the hundred and thirty on the slide; it is closer to eighty — respectable, but a long way from the case, and achieved at a cost in responsiveness that appears in no ledger at all.
That last cost is the one that corrodes. The clerk who sat in the division carried knowledge that was never written down: which supplier’s invoices are always wrong, which manager will dispute the recharge, which quarter-end will bring the awkward accrual. Consolidation treats that tacit knowledge as waste to be standardised away, and then spends years rediscovering, through error and escalation, that it was in fact load-bearing.
The same mistake, twice
Here is the essay’s central claim. Centralising to save and fragmenting to fail are not a success followed by a failure. They are the same act, described first by its intention and then by its result.
The error is to treat the organisation chart as if it were the operating model. Consolidation is a change to the chart: it says where the people sit and whom they report to. The operating model is something else entirely — it is how work actually flows, how decisions are made, how the centre and the business hold each other to account, how knowledge moves, how exceptions are handled, how the two sides resolve the inevitable disputes about what “good” means. The programme delivers the chart and assumes the operating model will follow. It does not follow. It has to be designed, and it rarely is, because the countable business case gave no one a reason to design it.
Moving the boxes is not the transformation. It is the part of the transformation that is easy to draw, easy to cost and easy to declare complete — which is exactly why it is so often mistaken for the whole.
Seen this way, the shadow teams are not a betrayal of the model. They are the organisation supplying, at its own expense and after the fact, the operating model the programme failed to build. Fragmentation is the system healing the wound the chart-change opened — clumsily, expensively, and invisibly to everyone whose attention moved on the day the centre went live.
The strongest case for the defence
The fair objection to all of this is simply that shared services demonstrably works. There are organisations running large, mature centres that deliver real and sustained savings, year after year, without the divisions rebuilding in the shadows. If the model were as self-defeating as the argument so far implies, those centres could not exist. The objection is correct, and it is the most useful thing in this essay, because the difference between the centres that endure and the ones that quietly unravel is not scale, sector or system. It is a difference of conception.
The implementations that hold up stop treating the centre as the output of a one-off restructuring and start treating it as a service business that happens to have internal customers. That change of conception has concrete consequences. Such centres invest in demand management, not only supply: they work to reduce and standardise what the business asks for, rather than merely processing whatever arrives. They measure themselves on outcomes the business recognises — the accuracy of the recharge, the usefulness of the report, the time to resolve a dispute — and not merely on the internal activity counts that fill a conventional agreement. They govern the relationship as a live thing, with real forums where the standard is renegotiated as the business changes, rather than freezing it in a document signed at go-live and never reopened. And, crucially, they treat the tacit knowledge that lived in the divisions as an asset to be transferred and retained, not as waste to be eliminated.
None of this is exotic. But notice what it is: it is the operating-model work that the countable business case omitted, done deliberately and paid for on purpose. The centres that succeed are the ones that did the expensive, uncountable half of the job anyway — usually because someone insisted on it against the grain of the appraisal. The model is not the problem. The habit of buying only the half of it that shows up in the spreadsheet is the problem.
Intent and reality
There is a general lesson here, and it reaches well beyond the back office. The gap between transformation as it is designed and transformation as it is lived is, more often than not, the gap between what a change programme finds easy to measure and what actually determines the outcome. We are drawn, structurally and almost irresistibly, to the countable: the heads, the boxes, the licences, the run-rate. We design the change around what we can put on the slide, declare victory when the slide is delivered, and then spend the following years absorbing, off the books, the costs of everything the slide left out.
“The organisation chart is where transformation is declared complete; the operating model is where it is actually decided.”
The shared services promise is not a bad idea that keeps fooling sensible people. It is a good idea that keeps arriving stripped of the half that makes it work, because that half is the half we find hardest to count. The remedy is not to abandon consolidation — the economics are real, and in a cost-pressured moment they are not going away. The remedy is to be honest, at the point of decision, that the boxes are the beginning and not the end; to fund the operating model and the relationship as deliberately as we fund the migration; and to treat the reappearance of the shadow teams not as an anomaly to be stamped out but as a message to be read. They are telling us, in the only language an organisation ever really speaks — where it actually spends its effort — that the transformation we declared finished had barely begun.