Interposing a profit motive between an enterprise and one of its functions stops the work nobody specified — and the unspecified work is frequently the reason the function was worth having.
Every enterprise faces the same tension about its own boundary. Holding a function inside costs visible money, carries fixed overhead, and ties up management in work the enterprise is not for. Buying it converts fixed cost to variable, brings a specialist's scale, and frees leadership for what only it can do. Both sides have force, and enterprises that never test the boundary calcify around functions they no longer need.
What makes it treacherous is that only one side of the comparison is written down. The function's cost is documented to the dollar. Its output is documented only as far as somebody once had reason to describe it — a service standard, a schedule, a set of volumes. Everything nobody described enters the comparison at zero.
A contract is a complete statement of what will be paid for, and on the supplier's side of the boundary anything outside it is cost without revenue. That is not bad faith; it is what a margin is for. A well-run supplier does what it is paid to do, does it efficiently, and does not fund from its own margin work the buyer did not think worth naming.
Inside the enterprise the same work behaves differently. It is done by someone already in the building, on the way to something else, at no incremental cost. Because it costs nothing extra it never enters a ledger; because it never enters a ledger it is never valued; because it is never valued it is never specified; and because it is never specified, it stops the day the boundary moves. Every step is unremarkable and the total is not.
The Strategic Context
This is a business-model decision wearing procurement clothing. It changes what the enterprise is made of, and is usually taken inside a transformation programme whose benefits case closes long before the consequences surface. A project can be judged when it ends and a programme when its capability lands; a boundary change is judged years later, when the people who made it have moved and the comparison case is gone.
The teaching material handles both halves of the mechanism and never joins them. It treats a transferred function as a specification — scope, service level, price — and organises everything around getting that right. Its risk discussion concedes in passing that transferring work erodes the buyer's own skills and the understanding it needs to manage and audit what it has bought. Its treatment of privatisation states the tension plainly: a private operator is driven to maximise profit while the entity it replaced was obliged to serve everyone, and unprofitable segments will not be served unless funded or compelled.
Those are the same mechanism at two scales, and between them sits a question the material never asks and gives no instrument to answer: what is the function doing now that no specification will carry across, and what is that worth? The asymmetry underneath states in a line: the enterprise pays for the specified work and receives the unspecified free, while the supplier is paid for the specified work and would fund the unspecified out of margin.
The Saving Is Measured; the Loss Is Not
The saving appears as a number in a budget line, with an owner, a date and a variance report. The loss appears as nothing for a period, then as a rise in unrelated incidents, each with its own apparent cause: a failure caught too late, a complaint that escalated further than it need have, a decision made without context somebody used to supply. No control system attributes them to the boundary change, because none was watching the thing that stopped.
A related divergence sits outside this article. The control system a board is shown is denominated in accrual while the constraint that ends an enterprise is denominated in cash, and no instrument in the delivery discipline joins them; that is the subject of [Related article: Accrual Governs the Report; Cash Governs the Outcome]. The divergence here is harsher: the saving is recorded in a ledger and the loss nowhere, on any basis of accounting.
The post-implementation review compounds it. It asks whether service levels were met. They were: response times inside tolerance, volumes delivered, audits passed. The function is performing exactly to specification, which is the problem stated in the language least likely to reveal it.
Reframing the Issue
The useful question is not what the function costs, but what it does that nobody has written down.
A specification is not a description of a function. It describes the part somebody could describe, at one moment, for one purpose — usually budgeting, assurance or contracting. A function's value is its specified output plus a discretionary layer, and the transaction prices the first while valuing the second at zero.
The obvious response — specify more — fails structurally. Most of the discretionary layer's value lies in the judgement of when it applies, and judgement is what specification cannot carry. An enterprise can specify an inspection frequency; it cannot specify noticing. It can require a monthly report; it cannot require the observation that something looked wrong on a Tuesday. Each attempt also converts a free act into a priced one, raising the price and lowering the frequency.
What the Margin Actually Stops
Work that is free inside and priced across a boundary
Consider a hypothetical municipal parks and open space function. The crew mows to a schedule. It also notices a cracked limb over a playground; mentions a drainage failure before it becomes a washout; tells the events team a lawn will not recover before a festival; knows which irrigation zone has failed for six years and why; and settles a resident's complaint on the spot.
Contracted out, the arrangement specifies mowing frequency, sward height, response times and rectification periods, all of which the contractor meets. The cracked limb now sits in a separate arboriculture scope, or in nobody's. A resident reports the drainage failure three weeks later. The festival lawn fails publicly. Each event is small, avoidable, attributable to something else — and none is a breach.
The economics are not about goodwill. Reporting the limb costs an employed crew nothing. It costs a contractor a conversation, a form, a record that it saw a defect, and a real chance of being asked to make good at its own expense. A rational contractor reports what the contract requires. That is not the contract failing; it is the contract working.
The loss of the reason behind the work
The second casualty is the knowledge of why: why that irrigation zone fails, which slope slumps after heavy rain, which fittings fail early, which decision three years ago explains an arrangement that now looks illogical. None of it is recorded, none of it compensated, and none of it transfers with the work.
Its loss compounds: it degrades the enterprise's ability to buy the function in future. Specifications get written by people who no longer know what they are for, evaluation is done by people who cannot tell a good technical answer from a plausible one, and every later contract carries a price for risk the buyer can no longer describe. Managing what has been bought also consumes senior attention on a scale no business case contained, which is the subject of [Related article: From Execution Attention to Governance Attention]; this article is about the work that ceased when the margin went in, not what governing the arrangement costs.
Advice sought only while it is free
A hypothetical specialist legal practice shows the same mechanism professionally. In-house counsel sits in commercial meetings and says small things: that a sentence should not go in an email, that the way a service has been described changes the indemnity, that a clause means something else in the jurisdiction where the work will happen. None would have been commissioned, because nobody knew they were needed until they happened.
Move the work to an external panel on hourly rates and each intervention requires somebody to judge it worth a call, open a matter and accept a bill. Low-threshold interventions stop first, and those were the ones preventing most, because they came before anyone had recognised a problem. The enterprise now buys advice only once someone has identified the risk — the moment advice is most expensive and least valuable. A priced service is consumed less than the same service free, whatever it was worth.
Decision Framework
The unspecified-work inventory is run before a function crosses the boundary and again a year later. It takes about four weeks of part-time effort and produces a number where there is none.
Step one — observe, do not survey. For four weeks, sit with the people who do the work and log every act no document requires. Record the trigger, who initiated it, who received it, and the counterfactual: what would have happened otherwise, and when. Do this with the people themselves; a manager's list will mostly contain work already specified.
Step two — classify. Sort the log into six types, which behave differently once a boundary exists:
| Type of unspecified work | What replaces it |
|---|---|
| Early warning | A later, larger incident with another apparent cause |
| Cross-boundary problem solving | An escalation, or a problem left unowned |
| Informal advice | A commissioned engagement, or no advice |
| Institutional memory | A rediscovery cost, paid repeatedly |
| Discretionary quality | Compliance at the contractual minimum |
| Relationship maintenance | Formal complaints and their handling cost |
Step three — price by avoided consequence, not by effort. For each item, estimate what the consequence would have cost had the act not occurred, and how often it occurs in a year. Multiply, then sum by category. The arithmetic is crude and the estimates contestable; both are acceptable, because the number's purpose is to exist.
Step four — decide each category explicitly. Three decisions are permitted, each with a named owner. Specify it, accepting the price and the loss of judgement. Retain it, funding the residual capability that performs it and saying so in the business case. Or accept the loss, recording the consequence the enterprise will absorb, with a review date.
Step five — apply the threshold. If the discretionary layer's estimated value exceeds the projected saving, the case fails on its own terms and should be reworked. If nobody can describe that layer at all, the function is not ready to be contracted out; that is a finding, not a delay.
Step six — re-run at twelve months. Against the same categories, ask which items still happen, who performs them now, and what the accepted losses cost. It is the only test that shows whether the enterprise bought what it thought.
From Strategy to Execution
Immediately. Run the observation window on the next function scheduled to move, and retrospectively on one outsourced two or three years ago, by asking those who remain what stopped. The retrospective persuades a board faster, because its findings are already true.
Over the next two to three quarters. Make the inventory a mandatory attachment to any business case moving a function across the boundary, and fund the residual capability it identifies inside the case rather than as contingency. Where the discretionary layer matters, design the contract to make judgement free at the point of use: a prepaid advisory pool, a fixed retainer, a no-fault obligation to report what is observed. The barrier is rarely the money; it is the transaction.
Over the longer term. Treat the enterprise's boundary as a portfolio position reviewed on a cycle, not a series of one-way decisions taken by whoever held the budget. That review is also where an enterprise re-enters the market, and pre-contract instruments used there are not free: a non-binding approach is an option the market eventually prices into every bid, which is the subject of [Related article: The Free Option You Give Away Before Anyone Bids].
Signals to Monitor
Incidents in the transferred domain that nobody saw coming, where somebody once would have. Problems first reported by customers, residents or regulators rather than by the people doing the work. Contract compliance excellent while satisfaction falls. A drop in requests for advice after a service moves to hourly billing, which will be reported as an efficiency. Specifications growing at each renewal, as the enterprise buys back what it used to receive free. And plainest of all: nobody left who can explain why something is as it is.
Questions for the Leadership Team
- For a function we outsourced years ago, what work stopped that no contract mentioned — and who did we ask?
- What is the annual value of the discretionary layer in the function we are about to move, and how was it estimated?
- Which incidents last year were first reported by an outside party, and would they once have been caught internally?
- Where a service moved to hourly billing, how has request volume changed, and what size of problem now justifies a call?
- How much of the specification we are drafting exists only because we lost knowledge we once held?
- For each function outside our boundary, who inside can still explain why it is configured as it is?
Closing Perspective
The margin is not the villain. A supplier that declines to fund unpriced work is behaving as the arrangement asks, and an enterprise expecting otherwise has misunderstood what it bought. The failure is upstream, in a comparison that counted every dollar of cost and none of the work that made the function valuable.
What a leadership team decides here is not a sourcing question but one of composition: which capabilities the enterprise holds, and what it will still notice about its own operations in five years. An enterprise can contract out a great deal and remain formidable. What it cannot do is contract out the parts of itself nobody wrote down, and then be surprised they are gone.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.