Shared Services Save Nothing When the Work Stays Fragmented
Centralisation changes where work sits; standardisation changes what the work is.
The Bargain on Paper
At the end of the first steering meeting, the arithmetic looks irresistible. Fourteen business units employ 430 people in accounts payable, payroll administration, purchasing support and general accounting. A shared service centre is expected to do the same work with 280. The property footprint will shrink, seven finance systems will eventually become one, and annual savings of £7.5 million will repay the move within two years.
Eighteen months later, the centre employs 348 people. Another 67 remain in the businesses under titles such as finance liaison, commercial support and local administration. Invoice backlogs have doubled, divisional finance directors have created small teams to chase the centre, and the promised system consolidation has been deferred because the transfer itself consumed the available management attention. The organisation has centralised the visible clerks while reproducing the old fragmentation at the boundary.
This is not simply poor execution. It reveals a contradiction inside the shared-services promise now attracting so much attention. The promise says that common work can be pooled, standardised and managed as a service. The typical programme, however, begins by moving people and only later asks whether the work is common. It treats location as the cause of cost when location is often only where historical differences have accumulated.
Centralisation changes where work sits; standardisation changes what the work is. Confusing the two produces a centre that inherits every exception, every disputed ownership line and every weak piece of data, then carries the blame for exposing them.
Why the Idea Has Such Force
The appeal of shared services at the beginning of this decade is understandable. Large organisations have spent heavily on Year 2000 remediation and enterprise systems. The collapse in technology valuations has made boards more sceptical of investment stories built on novelty alone. Capital is tighter, administrative cost is visible, and finance directors are being asked to demonstrate returns from systems already purchased.
Shared services appears to answer several anxieties at once:
- It offers economies of scale without requiring the businesses themselves to merge.
- It promises common controls after years of local system growth.
- It creates a single management point for routine administration.
- It supplies a practical route towards common data and enterprise-wide reporting.
- It converts dispersed overhead into a programme with a business case, milestones and a named executive.
There is real merit in each proposition. Repetitive transaction work often does benefit from scale. A team handling 600,000 invoices can justify specialist workflow, document imaging, supplier help desks and disciplined queue management in a way that fourteen small teams cannot. Common payroll or accounting rules can also make control more reliable. The serious mistake is not belief in shared services. It is the belief that the service becomes shared at the moment the organisation chart changes.
The word shared conceals two quite different ideas. One is common ownership: several businesses receive work from the same centre. The other is common design: those businesses agree that the work will follow the same rules, use the same data and accept the same service choices. The first can be decreed. The second must be negotiated, built and governed.
A Centre Built Around Exceptions
Most early centres are designed from head-count schedules. Each division identifies staff performing an activity, a percentage of those posts is declared transferable, and a productivity factor is applied to the total. The resulting number becomes both the staffing plan and the savings case.
The method is quick, but it measures boxes on an organisation chart rather than demand for a service. Two payroll clerks may have the same title while one processes a stable monthly payroll and the other spends half the week correcting time records from sites with different shift arrangements. Five accounts-payable teams may all “process invoices”, yet accept different authorisations, tolerate different purchase-order practices and maintain separate supplier codes. When these people are moved together, their work does not become uniform. The diversity has simply been concentrated.
A composite centre formed from nine operating companies illustrates the mechanism. Before transfer, the programme counted 312 finance-administration posts and assumed that scale would remove 25 per cent. Process mapping, conducted only after the property lease was signed, found:
- 41 invoice approval routes, including 13 that depended on named individuals.
- Six charts of accounts with incompatible cost-centre conventions.
- Nine supplier files containing 74,000 records, of which roughly 11,000 appeared to be duplicates.
- Four definitions of an “on-time” payment.
- 22 local spreadsheets used to bridge gaps between purchasing, receipts and accounts.
- More than 300 standing exceptions, many of them undocumented.
The centre opened with 236 transferred staff, close to plan. Within six months it had added 38 temporary staff and each operating company had appointed at least one liaison officer. The unmatched-invoice rate rose from 9 per cent to 21 per cent, not because the centre’s clerks were less capable, but because local knowledge had formerly resolved incomplete orders and missing receipts without recording the intervention. Centralisation made the hidden work visible.
This is a recurring irony. The local process often appears efficient because exceptions are absorbed by people whose knowledge is not represented in the procedure. The centre removes proximity before it removes the need for that knowledge. A query that once crossed a desk now enters a queue, travels back through a liaison and waits for a business manager who regards it as the centre’s problem.
A shared service centre does not eliminate local complexity. It prices that complexity, queues it and makes its ownership impossible to ignore.
The Strong Case for Difference
It is easy to answer these failures with a call for absolute uniformity. That answer is as careless as the relocation-first programme.
Local variation is not always waste. Different businesses may face different tax rules, currencies, union agreements, customer terms, seasonal cycles or statutory reporting obligations. A high-volume distribution operation should not necessarily share every control with a project business that raises a small number of complex invoices. Divisional managers are also right to fear that a remote centre, judged mainly on unit cost, may protect its own efficiency by pushing difficult work back into operations.
The strongest argument against centralisation is therefore not sentimental attachment to local autonomy. It is that administrative processes frequently contain commercial judgement. Remove that judgement too aggressively and the business may save a clerk while losing a customer, delaying a shipment or weakening control. At a time when many enterprise-system programmes are already pressing organisations towards common processes, the risk of confusing consistency with competence is real.
Yet this objection does not justify inherited variety. It requires a more exact distinction between necessary difference and unexamined difference. Necessary difference can state its cause, owner and cost. It is designed deliberately and reviewed when the cause changes. Unexamined difference survives because nobody has had to defend it.
| Test | Necessary difference | Inherited fragmentation |
|---|---|---|
| Cause | Legal, commercial or operational requirement | History, preference or system limitation |
| Owner | Named manager accepts cost and control | Ownership shifts when questioned |
| Evidence | Consequence of removal can be described | Harm is asserted but not demonstrated |
| Design | Exception is explicit within the service | Workaround sits outside the documented process |
| Review | Reconsidered at an agreed date | Continues indefinitely |
This is the argument that shared-services programmes ought to force before they move work. The decision is not “standard or local”. It is which differences deserve to remain, how they will be served, and who will pay for them.
The Economics of Movement and the Economics of Change
Relocation and transformation have different economic shapes.
Movement creates visible, early costs: recruitment, redundancy, property, travel, training, temporary cover and disrupted service. Its benefits are easiest to express as posts removed and accommodation released. Change requires slower work on policies, data, systems, controls and managerial behaviour. Its benefits arrive through fewer exceptions, less rework, shorter cycle times and better use of information. These gains are harder to assign to one budget holder, so programmes often claim them but fund only the move.
That imbalance explains why the centre becomes the destination for unreformed processes. The property date is fixed. Consultation has begun. Managers have announced a savings number. Process decisions, by contrast, remain negotiable. When disagreement threatens the timetable, the variation is transferred “for later harmonisation”. Later rarely comes, because the centre must stabilise service while meeting a cost target based on the standardisation that was deferred.
The result is a peculiar double charge. The organisation pays once to move the work and again to operate the complexity in a more formal setting. It may even pay a third time when the businesses rebuild local capability to recover responsiveness.
The accounting can conceal this outcome. The centre reports lower cost per transaction because it counts only its own payroll, while the businesses report their liaison and correction work inside broader finance or operations budgets. Project savings are booked when posts leave the original cost centres, even if temporary labour, overtime and retained teams appear elsewhere. Service complaints are described as a transitional problem, and the transition has no agreed end.
A credible view of economics must follow the work across the boundary. If a central query takes ten minutes in the centre and thirty minutes in the business, it is not a ten-minute transaction. If a removed local post is replaced by a commercial analyst who spends half the week correcting service failures, the labour has changed its title, not disappeared.
Distance Changes Control
Local administration is often controlled socially. People know whom to ask, which manager will approve late, which supplier uses an unusual reference and which month-end entry will reverse next week. This is fragile control, but it is still a form of control.
A centre replaces social memory with explicit rules, records and service commitments. That can be an enormous improvement, but only when the replacement is designed. Without it, distance does not create discipline; it creates correspondence. Forms multiply, calls are logged, spreadsheets track spreadsheets, and every party gains evidence that delay belongs to someone else.
The service-level agreement is particularly prone to misuse. It is valuable when it expresses a genuine choice: the time, quality and cost the businesses are prepared to fund. It is useless when written after the process has been transferred and used to convert unresolved design questions into response-time measures.
A centre can answer 95 per cent of calls within twenty seconds and still fail if callers need to telephone because invoice status cannot be seen. It can process 90 per cent of complete invoices within three days and still leave suppliers unpaid if “complete” excludes the cases created by weak purchasing discipline. Measures drawn only from the centre’s controllable portion reward the clean handling of work already clean.
The better control question is end to end: what event begins the service, what acceptable outcome ends it, and where does avoidable delay enter? For payables, the chain begins with a purchasing commitment, not with an invoice arriving at the centre. For employee administration, it begins with a management decision, not with a completed form. Shared services works when it creates joint accountability for that chain. It fails when the centre becomes the place to which every upstream defect is delivered.
What the Centre Reveals
There is a more useful way to interpret early disappointment. The centre is not merely a cost-saving machine that has underperformed. It is a diagnostic instrument.
By collecting work, it reveals incompatible policies, unreliable data, ambiguous authority and the true volume of exceptions. Queue data shows where incomplete requests originate. Supplier calls expose purchasing failures. Payroll corrections reveal where local managers submit changes late. The very problems that make the first year difficult can provide the evidence for the second year’s redesign.
But this learning occurs only if leaders resist two temptations. The first is to protect the original business case by classifying every shortfall as temporary. The second is to blame the centre for problems whose causes sit in the businesses. Both responses preserve fragmentation.
The centre manager needs authority to reject defective demand, but rejection alone is not enough. Business managers must receive measures showing the cost and service consequence of their exceptions. Finance must reconcile total administrative cost, including retained and liaison work. Process owners must be able to change a rule across organisational boundaries. Without these rights, the centre manages queues while nobody manages causes.
“The real unit of shared services is not the transaction. It is the rule that determines whether the transaction can flow without negotiation.”
That insight changes the sequence. Define the rule, expose the exceptions, decide which differences are justified, repair the data and only then set the capacity of the centre. Where immediate relocation is unavoidable, the organisation should at least separate the savings case for movement from the improvement case for redesign. Combining them produces a number nobody can later explain.
The Promise Worth Keeping
Shared services should not be rejected because the first generation of centres is proving difficult. The underlying idea remains sound: routine work should not be replicated simply because the organisation has several business units. Common expertise, stronger controls and investment in workflow can create value that dispersed teams cannot easily match.
But scale is an amplifier. It amplifies a clean process, and it amplifies a confused one. It makes good data more useful and bad data more damaging. It gives common rules reach, while turning each unresolved exception into a recurring cost.
The mature promise is therefore more demanding than “centralise to save”. It is to make administration explicit enough to be shared. That means treating policy, data, service choices and exception ownership as part of the design, not as matters to tidy up after the move. It means measuring the whole process rather than the centre’s portion. It means allowing legitimate difference while forcing each difference to declare its reason and cost.
The organisations most likely to succeed will not be those that move the most people fastest. They will be those prepared to learn what their administrative work actually consists of before they price its removal. They will accept a slower, less dramatic headline in exchange for an operation that can be understood and improved.
The centre is only the visible architecture. The shared service itself is an agreement: on the rules, on the evidence, on the choices and on who acts when the work does not flow. Without that agreement, centralisation does not cure fragmentation. It gives fragmentation a new address.