Buying a capability does not release management attention; it converts execution attention into governance attention at an exchange rate no business case contains, and the conversion is frequently unfavourable.
The paper in front of the board runs to a page and a half. A function performed in-house will move to a specialist provider. Internal cost is stated, external price is stated, the difference is favourable, and the closing line notes that released capacity will go to work closer to the enterprise's strategic priorities. Nobody dissents: the arithmetic holds, the provider is competent, and the function has never been claimed as a source of advantage.
The decision looks simple because it has been framed as a comparison with one variable. What has been approved is a change in the kind of work the enterprise does, not a reduction in how much of it there is. Someone must still decide what good looks like, express it in language a third party can be bound to, judge what arrived against it, and settle the argument when that judgement is contested. Those acts were embedded inside doing the work. Now they are the work.
That is the conversion. Execution attention — the attention absorbed by performing a function — becomes governance attention: the attention absorbed by specifying, commissioning, assuring and adjudicating it. It happens at a rate, that rate is not one for one, and in ERANORTH's experience it appears in no business case.
It belongs at enterprise level because the two come from different pools. Execution attention is broadly distributed and largely delegable; more of it can be bought. Governance attention sits with a narrow band who can specify a requirement with authority, judge performance against it, and settle a dispute without escalating. That band also makes the portfolio's choices, and a sourcing decision rarely expands it. Every arrangement drawing on it draws on the capacity to decide what the enterprise does next.
The Strategic Context
The teaching literature of procurement is unusually clear that the price of a bought service understates its cost. Transaction cost economics, associated with Oliver Williamson, holds that the choice should turn on total cost — search, negotiation, monitoring, enforcement, termination — and that where uncertainty is high and the asset is specific to the relationship, market governance becomes expensive enough that making wins.
Yet the same material, having established that those costs exist, turns them into line items denominated in currency. Once expressed in dollars they join every other dollar in the comparison, which resolves in favour of the smaller column. The move looks rigorous and conceals what matters: money is fungible and senior attention is not. An enterprise can fund an extra hundred thousand dollars of contract administration from a dozen places. It cannot fund four hundred hours a year of its chief engineer's judgement from anywhere.
The problem compounds as work rises through the hierarchy. A contracted service inside a project consumes attention for that project's duration and stops. A programme delivering a capability change carries several such arrangements whose governance load lands in the same months. A portfolio accumulates them permanently: each is defensible alone, and the aggregate is a standing claim on the enterprise's scarcest resource that no approval ever authorised.
What the Business Case Cannot See
The material's own structure gives the mechanism away. It treats risk transfer as a benefit of buying, and separately establishes — in different chapters, without joining them — that commissioned work requires specification, evaluation, contract administration, performance monitoring, dispute handling and eventual exit. Each is a claim on senior attention the internal option did not make. The benefit is counted; the cost of collecting it is not.
More revealing is the process prescribed for reaching such a decision: strategic objectives, model design, validation, requirement definition, vendor selection. It is a sound sequence, and it has no branch along its length that returns a decision to retain. Once an enterprise enters it, every terminal state is a supplier, and validation tests whether the chosen model is well formed rather than whether buying was right.
A well-governed procurement is therefore a process that cannot conclude "we should keep doing this ourselves". Its rigour sits downstream of the only question that could have changed the outcome, and the input that would have forced it — the constrained band's time, for as long as the arrangement lasts — is never asked for. Governance attention is also uneven across a contract's life, clustering at the start and again in the tail; business cases are built on the quiet middle.
Reframing the Issue
The useful question is not whether to make or buy. It is what the enterprise is exchanging, and at what rate.
Framed that way, judgement calls become measurable. Whether a function is core is a question about competitive advantage; whether contracting it out releases capacity is a question about who does what, and has an arithmetic answer. Two arrangements at identical prices can have opposite effects on the enterprise's ability to decide anything else, and no price comparison distinguishes them.
It also relocates the decision. Sourcing authority is delegated on value thresholds: below a number a manager decides, above it a committee does. But the resource committed is not proportional to contract value. A modest contract for a technically ambiguous service, where every exception needs expert judgement, can consume more of the constrained band than a large one for a defined commodity. Delegation set by dollars approves the arrangements most expensive in attention at the lowest level of scrutiny.
Where the Conversion Turns Unfavourable
The specification interface
The rate depends first on how much of a function's value lives in judgements that cannot be written down in advance. Where a service is fully specifiable — defined output, unambiguous test, stable requirement — conversion is cheap, because the specification governs. Where the requirement moves, or the right answer depends on conditions found during the work, a person must.
Consider a hypothetical operator developing a geothermal resource. It buys drilling rather than building a drilling capability, since drilling is a mature market while its own advantage lies in resource assessment. Execution attention transfers cleanly. What does not transfer is the run of decisions taken while the hole is open: whether an unexpected formation warrants a change in the mud programme, whether to case earlier than planned, whether a day of non-productive time was caused by the operator's data or the contractor's equipment. Each was a technical decision taken by people who lived with the consequence. Each is now also a commercial event needing someone senior enough to commit the enterprise.
The adjudication tail
Every commercial event is a candidate dispute, and disputes draw from the very top of the band. This is the part of the conversion that estimates omit most reliably, because at the time of the estimate it has not happened.
It carries a consequence this article does not pursue. When an arrangement causes loss, recovery has its own structure: the party that caused the damage is often not one the enterprise can sue, and recovery through a chain of suppliers compounds against you at every link. That is the subject of [Related article: The Party You Cannot Sue]; the point here is earlier — the possibility of a claim alone lifts routine decisions to a level where they can be defended.
The band the attention comes from
A hypothetical subscription meal-kit business shows the same mechanism at low unit value. It contracts out pick, pack and dispatch: repetitive activity, competent market, strong case. But it changes its recipe range weekly, works with short shelf lives, and relies on substitution decisions trading a customer's experience against a supplier's cost. Inside, a shift supervisor made those calls in seconds. Outside, each is a variation, a service-credit argument or an unattributed complaint. The decisions are small; their frequency makes them a standing claim on someone senior.
Two adjacent boundaries sit elsewhere in this collection. How deeply an enterprise sees into a supplier's operation is a design decision with its own costs and its own article; the question here is not how much you can see but how much scarce attention seeing it consumes. And the counterparty may change while the contract does not: a contract binds a legal person, not the capability that person holds, so governance load can rise the day a supplier's owner changes. That is the subject of [Related article: You Contracted With a Company, Not a Capability]; this article assumes a stable counterparty and asks only what governing it costs.
Decision Framework
The attention conversion rate is the ratio of constrained-band hours consumed by governing a function after transfer to those consumed by performing it before. It is computed in four steps and is deliberately crude, because the answer is rarely close to the threshold.
Step one — define the constrained band. Name individuals, not roles: those whose time limits what else the enterprise can do, because they can specify a requirement with authority, judge whether it was met, and settle a disagreement without escalation. If the list runs past a dozen names it has been drawn too generously and will flatter the decision.
Step two — measure the "before" honestly. Count hours per month those named people spend on the function in any capacity, including interruptions, informal advice and problems brought to them because they are the ones who know. Interruptions are most often omitted and most likely to survive the transfer.
Step three — estimate the "after" by category. Estimate constrained-band hours per month at steady state for requirement definition; market engagement; contract administration including variations and claims; performance assurance; escalation; and exit readiness. Estimate the first year separately, since it will be heavier, and add a dispute loading: the share of months carrying an escalation, times the hours one consumes.
Step four — compute and act. Below 0.6, the transfer genuinely releases scarce capacity. Between 0.6 and 1.0 the release is modest, and the estimate should be re-tested by someone who has run a comparable arrangement. Above 1.0 it consumes more constrained attention than it frees, and needs another justification — access to capability, capital avoidance, surge coverage — written down. Above 1.5, redesign it or retain the function.
| Where the rate is set | Cheap conversion | Expensive conversion |
|---|---|---|
| Specifiability of the output | Objective, testable, stable | Judgement-dependent, moves with conditions |
| Frequency of exceptions | Rare and batchable | Continuous and individually consequential |
| Who decides an exception | Contract administrator | Named senior technical or commercial authority |
| Consequence of a wrong exception | Absorbed operationally | Safety, regulatory, reputational or claimable |
Two rules make this governing rather than analytical. The expected rate goes into the business case with a named owner, and the realised rate is measured at twelve months by someone else and reported to the body that approved the decision.
From Strategy to Execution
Immediately. Apply the rate retrospectively to the three arrangements consuming the most senior time, not the three largest by contract value; they are unlikely to be the same three. Publish the results without blame, so the measure is established before it is contested.
Over the next two to four quarters. Make the expected conversion rate a mandatory field in sourcing approvals above a defined threshold, and change the delegation basis so a high expected rate escalates the decision regardless of contract value. Build the "before" measurement into standard analysis so it is not reconstructed from memory afterwards.
Over the longer term. Hold a portfolio-level view of committed governance attention across every arrangement, as a share of the constrained band's available time. Treat it as a constraint on new commitments in the way a balance sheet constrains capital. An enterprise that cannot state this number is setting its strategic bandwidth by accumulation rather than choice.
Signals to Monitor
The earliest signal is calendar drift: senior diaries filling with supplier meetings nobody planned. The second is the informal escalation, where problems reach an executive by relationship rather than through the contract's mechanism, which shows it is not resolving them. The third is variation frequency outrunning variation value: many small changes signal a specification that cannot govern, and each costs roughly the attention of a large one. The fourth is the quiet return of shadow capability, where the enterprise rebuilds expertise to supervise work it already pays for. The fifth is silence on exit.
Questions for the Leadership Team
- For our three largest outsourced arrangements, how many hours of named senior time did each consume last quarter, and how does that compare with the internal function it replaced?
- Which of those arrangements would have failed an attention conversion rate test at approval, and would we have decided differently?
- What share of our constrained band's time is now committed to supervising work done by others, and who authorised that total?
- In the last two years, how many sourcing decisions were approved under value-based delegation that would have escalated on an attention basis?
- When a specification fails to resolve an exception, who actually decides, and is that person on the list we would have named as the constraint?
- For each major arrangement, what would exit cost in senior time, and when did anyone last test that estimate?
Closing Perspective
There is a second cost this instrument does not capture and should not be asked to. When a margin is placed between an enterprise and one of its functions, the work nobody specified stops — the early warning, the passing advice, the problem solved on the way to something else — and that unspecified work is frequently the reason the function was worth having. That is the subject of [Related article: When You Put a Margin Between the Enterprise and a Function]; this article is about what an arrangement consumes, not what it quietly ceases to produce.
Together they describe a responsibility that sits with the board and cannot be delegated to procurement. Every decision to buy rather than make commits the enterprise's scarcest and least expandable resource, for as long as the arrangement lasts, without naming it. Many such commitments can be carried before anything visibly breaks. What breaks first is not delivery. It is the capacity to consider anything not already on the agenda.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.