The most valuable form of organisational capability lives in specific pairs of people, is not transferable to their replacements, and can be destroyed by a reorganisation in which nobody leaves the enterprise.
The most expensive loss in a reorganisation is usually the one that appears nowhere in the post-implementation review, because nothing measurable went missing. Headcount held; every skill on the capability matrix is still in the building. Then, months later, a class of decision that used to be settled in an afternoon starts taking three weeks, and the explanation offered is that the teams need time to bed in.
They may never bed in, because what was lost was held by none of them individually. It sat between particular people: a scheduler and a coordinator who reroute a late order before the customer notices; a project director and a controller who discuss a slipping forecast a fortnight before it becomes a report. Each is one capability held jointly by two people, and neither owns it.
That is the counterintuitive part. If the capability sits in the pair, replacing either party does not degrade it — it ends it. The successor begins a new relationship at zero, on a build cycle measured in years, while the enterprise still prices the work as though the old speed were a property of the roles.
Almost no enterprise treats this as an asset class, for an instrumental reason: everything used to design an organisation operates on roles and reporting lines, and a pair is neither.
The Strategic Context
Consider a specialist manufacturer of precision bearings and seals for heavy rotating equipment — hypothetical in its particulars, ordinary in shape. Its most consequential technical judgement is written down nowhere: when a customer reports a surface finish problem, someone must decide fast whether the cause is material chemistry or the grinding cell. Getting it wrong costs weeks, more if it reaches a warranty argument.
In practice two people settle it in a short call. The metallurgist knows what the cell leader means by "it sounded different on Tuesday" because they have stood at the machine together forty times. The cell leader says "it might be us" early, because eleven years established that saying so will not be used against him. Neither condition is a qualification, and neither is in a procedure.
Now consolidate technical services into a central function two sites away. Nobody is made redundant; the metallurgist keeps the title and gains a broader remit, the cell leader keeps his job. The judgement is gone, because the conditions producing it — shared physical reference points and a safe channel for early bad news — were properties of proximity and history, not of either person.
The same structure appears in a teaching hospital's referral network, where the asset is the consultant who takes a call from one particular registrar at another site and acts on the description without repeating the workup, knowing how that registrar tends to overstate or understate a case. Replace the network with a single point of referral and the protocol survives while the pair does not. The protocol was never what made it fast.
Three organisational levels are involved. Projects form pairs, because consequential work under pressure forces two people to find out whether they can rely on each other, and dissolve them when the work ends. Programmes are where the same people persist long enough for the asset to accumulate. The portfolio sets how often team composition churns — the real control on how much of this capability the enterprise holds, which makes reorganisation a portfolio act whose consequences land at programme level and first appear as project-level slowness.
What Leaders Commonly Misread: Mistaking It for a Knowledge Problem
The reflex, once the loss is noticed, is capture. Document what the departing party knew, record the heuristics, build the decision tree, train the successor on it.
This fails for a precise reason: the asset is not knowledge, it is standing. What the pair held was the right to raise a difficult subject early and be believed without first assembling evidence, plus enough shared reference points that most explanation could be skipped. Neither transfers through a document. A successor who has read the whole file still has to earn the right to say "it might be us" on day two rather than day ten.
The medium through which a lesson travels from those who learned it to those who need it is the subject of Article 53. The point here is prior: what a severed pair held was not a lesson, so no medium would have preserved it.
The second misreading folds this into key-person risk. That analysis examines nodes — who is irreplaceable, who lacks cover, what happens if they leave. The asset here is an edge. An enterprise can hold full succession cover on both people, lose neither, and still write off the pair by placing them in different structures — which is why restructures with no attrition are so costly and so hard to argue about: every node-based instrument reports that nothing was lost.
Reframing the Issue
Treat the working pair as an asset with an unusual profile: high build cost, paid in time and invisible on any ledger; carrying cost near zero; replacement lead time in years. And it is destroyed by an administrative act that costs nothing, requires no capital approval, and is recorded nowhere as a disposal.
Valuable, unpriced, free to destroy and slow to rebuild: any organisation will over-consume such an asset — not through carelessness, but because that is what rational actors do with a resource nobody must account for.
Two forms behave differently. The first is the strong pair inside a boundary: accumulated trust, shared reference points, a channel through which bad news moves early. That is what makes a hard decision fast. The second is the weak tie across a boundary — a link to someone in another function, site or organisation. As Mark Granovetter established, such ties carry information the dense connections inside a group cannot, because everyone in a group already knows what the others know; Ronald Burt's account of structural holes shows that whoever bridges two unconnected clusters occupies a position no embedded member can.
Cross-boundary ties are the scarce and fragile form. They exist because two people had a structural reason to interact and kept talking after it expired. Remove the reason and nothing formal replaces it.
Why the Chart Cannot See the Asset
The design instruments have no representation for a relationship
Every instrument used to reorganise an enterprise operates on two objects: boxes, which are positions, and lines, which are authority. Neither is a relationship between named individuals. Such a design cannot express "these two people, specifically, produce a judgement neither produces alone", so the point is never entered, never assessed, never traded against the saving being pursued.
This is a property of the toolkit, not a lapse in any one enterprise. The material teaching restructuring works entirely on roles and reporting lines; the material teaching social capital establishes that the value sits in the relationship rather than in either party. Both are internally coherent, and almost never applied to the same decision.
The pairs that matter most are the ones nobody assigned
Assigned pairs — a manager and a report, two members of a mandated committee — are visible and protected by default, because the structure that created them usually survives. The load-bearing ties are typically unassigned: a supplier quality engineer and a customer's maintenance planner; a bid manager and a technical authority in another division; the applications engineer whose customer-site relationship surfaces problems months before they arrive as claims.
None has an owner, and nobody is paid to maintain one. Whether a reward architecture compensates anyone for sustaining a relationship they were never assigned — and what its mix of collective and individual reward says about what an enterprise believes a team to be — is the subject of [Related article: The Reward System Tells You What You Think a Team Is]. This article takes the reward system as it stands.
A further boundary sits upstream. Whether to formalise the developmental relationships from which many such pairs originate — trading wider access against a lower return per relationship — is the subject of [Related article: Formalise the Relationship and You Widen Access and Lower the Return]. That article governs who is granted a relationship; this one concerns what the enterprise holds once one exists, and what it destroys without recording.
Destruction is free, rebuilding is not
A restructure appraisal counts severance, recruitment, systems change and communications. It does not count the rebuild interval — the period during which a class of decision reverts to its slow form. That interval is real, often measured in years, and appears on no schedule.
This article does not convert it into money; the computable price of a day of delay for a revenue-generating asset, and how seldom it is computed, is the subject of [Related article: Every Day of Delay Has a Price]. The claim here is prior: the interval exists, nothing in the appraisal acknowledges it, and its length is a property of trust rather than process design.
Decision Framework: The Pair Inventory
The pair inventory is a register of the enterprise's load-bearing relationships, built annually and re-run before any structural change. It takes a fortnight and requires no system.
Collect edges, not nodes. Ask the twenty to thirty people who run consequential work two bounded questions. Name the three people outside your reporting line you would ring first if a decision had to be made today on incomplete information. Name the one person you would tell bad news to first. Open lists produce address books, not data.
Find the recurring pairs. Names recurring across respondents identify load-bearing individuals; pairs recurring across them identify the asset. Expect concentration: a few relationships carry a disproportionate share of the enterprise's speed.
Classify on two axes. Does the pair cross an organisational boundary, and how long did it take to form? For the second, ask each party how long they had worked together before they would have said the difficult thing — the honest rebuild estimate, usually longer than anyone senior expects.
Mark each pair against the proposed change. Intact: both parties retain a structural reason to interact. Stretched: one end moves, but a shared forum, joint review or co-signature keeps a reason alive. Severed: no remaining reason to speak.
Apply the rule. No structural change proceeds until every severed cross-boundary pair has either a named substitute route with an owner and a date, or a written acceptance by the executive making the change, recording the rebuild estimate. Severance without one of the two is not a saving; it is an unrecorded write-off.
One threshold is worth setting formally: where a single change severs several cross-boundary pairs in one function at once, stage it. Rebuild cycles run in parallel only while the counterparties still exist; sever the set together and nobody holds the old reference points to rebuild against.
The inventory has a second use. Preserving a tie needs only a structural pretext — a standing review, a joint sign-off, named representatives on a working group. Pretexts are cheap, and designing one in when somebody moves is the difference between stretching a relationship and ending it.
From Strategy to Execution
Immediate. Run the two questions against the next structural change already in design, before the consultation document issues, and give the severed list to the executive deciding. The aim is not to save every pair, but that the write-off is chosen rather than discovered.
Medium term. Add the pair inventory to the standard appraisal for any operating model change, alongside cost and headcount, and require rebuild time in the same document as the saving.
Long term. Treat team composition churn as a portfolio decision with a capability consequence. An enterprise that rotates people frequently buys breadth of exposure and pays in dyadic capability; one that leaves people together buys speed and pays in narrowness. Either is defensible; rotating for administrative convenience pays the cost and buys nothing.
Signals to Monitor
Watch for a class of decision whose handling time has changed while its complexity has not — the clearest evidence that a pair was severed upstream. Watch for problems arriving already escalated: when bad news enters through formal channels rather than a phone call, the early-warning ties have gone.
Watch for a rise in requests for written confirmation between functions that used to transact verbally — standing being rebuilt from zero. And watch for "they just need time to settle in" attached to the same team for more than two quarters, or for cross-boundary meetings that quietly lose their informal attendees.
Questions for the Leadership Team
- In our last reorganisation, which working pairs were severed, and has anyone since asked either party what it cost them?
- If our thirty most consequential operators named who they would ring first on incomplete information, how many names would sit outside their own reporting line?
- Which decisions here are fast because of a relationship rather than a process, and what happens to each if one of the two moves?
- How long did the pairs we rely on take to form, according to the people in them rather than according to us?
- What has changed in the handling time of our routine cross-functional decisions over two years, and what did we attribute it to?
- When we last consolidated a function, what did the appraisal say about rebuild time, and if it said nothing, who decided that was acceptable?
Closing Perspective
An enterprise cannot preserve every relationship and should not try. Some pairs ought to be broken, because closeness also produces shared blind spots and mutual protection. The failure is not that restructures sever relationships; it is that they sever them invisibly, and the enterprise then attributes the slowness to the people who inherited the structure.
What remains inside an enterprise after a capability investment is made and the budget closed belongs to Article 17 in this series. The asset described here was never bought, valued or entered anywhere — which is why it is spent so freely.
The responsibility becomes unavoidable the moment the inventory exists. A leader who has seen the severed list and proceeds has made a defensible decision. One who reorganises without ever building the list has decided nothing; they have spent an asset they declined to look at.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.