The Knowledge Drain Nobody Measured
It was weighing saving against nothing, and nothing always loses.
Executive Summary
Every outsourcing and offshoring business case written this decade rests on two numbers. The first is easy to defend: the saving — fewer people on the payroll, a lower day rate, a transition cost that pays back inside a couple of years, a figure a board can hold against last year’s cost base. The second number is never taken at all. It is the value of the institutional knowledge that leaves the building when the people do — the tacit, undocumented, load-bearing understanding of how the work actually gets done, which no transition plan fully captures and no ledger records as it quietly depreciates.
This essay is about that missing number. It argues that the great transfer of work now under way — to shared service centres, to nearshore and offshore providers, to the retained-organisation model that has become the default shape of the modern back office — is running on an accounting that can see cost but cannot see capability. We have become fluent in costing the people we release and almost illiterate in valuing what they carried. The consequence is not a single dramatic failure but a slow one: a steady erosion of the organisation’s ability to explain itself to itself, visible eighteen months after the celebration, long after the sponsor who signed the deal has moved on. The deeper reflection is uncomfortable. The knowledge drain is not really a story about outsourcing at all. It is a story about what organisations can and cannot see — and about how carelessly we change the things that never made it onto the balance sheet.
The Morning After Go-Live
There is a particular quiet that settles over a function about six months after a transition has gone live. The programme is closed. The steering committee has stood down. The service-level agreement is signed and, on paper, being met — response times inside the threshold, tickets closed within target, the monthly report a reassuring wall of green. The transition team has been thanked and redeployed. The people who used to do the work have taken their redundancy or their retained-organisation reassignment, and most of them have gone.
And then something breaks that is not in any runbook.
A reconciliation that has always balanced stops balancing. A batch that has always completed by six in the morning starts finishing at nine, and no one can say why. A month-end that used to take five days takes eight. The new team, diligent and capable and working precisely to the documented process, does everything the documentation says and cannot make the problem go away — because the answer was never in the documentation. It was in the head of a woman who had run that close for eleven years and knew, without knowing she knew it, that the figure from one feeder system arrived a day late every quarter and had to be accrued by hand. She was not a senior person. She did not appear in any knowledge-transfer plan as a risk. She has been gone for four months, and no one thought to ask her the question, because no one knew the question existed.
What leaves in an outsourcing transition is rarely the knowledge anyone decided to lose. It is the knowledge no one knew they had — and therefore no one thought to keep.
This is the shape of the thing. Not a catastrophe, which would at least be legible and attributable, but a diffuse and deniable decline. Each individual instance is small enough to be explained away as teething trouble, or the new provider bedding in, or an unlucky quarter. Only in aggregate, and only to someone willing to look, does the pattern resolve into what it actually is: the organisation has lost the ability to do something it used to be able to do, and has no idea how much of that ability it lost, because it never measured the thing in the first place.
The Asymmetry of What We Count
To understand why this keeps happening — and it keeps happening, across sectors and across the full range of providers, with a regularity that ought to trouble us more than it does — we have to look at the asymmetry buried in the business case.
Cost is legible. A full-time equivalent has a fully-loaded rate. A contractor has a day rate. A vendor has a price, itemised and benchmarked and negotiated. These numbers are precise, comparable, and auditable; they sit in systems designed to track them; they roll up cleanly into the figure the sponsor is accountable for. When forty people leave and a contract of a certain value replaces them, the arithmetic of saving is trivial and satisfying.
Knowledge is illegible. It has no rate card. It does not appear as an asset on any statement, because accounting convention treats the people who hold it as an operating expense, not as capital — which means that when they depart, the books record the removal of a cost and nothing else. There is no line for the write-down of an intangible, no depreciation schedule for institutional memory, no impairment charge when a decade of accumulated judgement walks out through reception carrying a cardboard box. The saving is booked in full and at once; the loss is booked nowhere and revealed slowly.
“We measure the people we release to the nearest pound, and the capability we release not at all.”
The old management maxim — that what gets measured gets managed — is usually offered as encouragement. Here it functions as an indictment. Knowledge did not get measured, and so it did not get managed, and so it was disposed of with a carelessness we would never permit ourselves with a physical asset of comparable value. No competent organisation scraps a working plant because the maintenance line looks expensive this quarter, without first understanding what the plant produces. Yet we do the equivalent with human capital routinely, because the plant is on the fixed-asset register and the capability is not. The instrument shapes the decision. We manage what our instruments can see, and our instruments were built to see cost.
What Actually Walks Out the Door
It is worth being precise about what is being lost, because a great deal of energy in transition planning is spent on the part that matters least.
Knowledge-transfer methodologies — and every serious provider now has one — are overwhelmingly concerned with the capture and handover of explicit knowledge: process maps, runbooks, standard operating procedures, system documentation, the shadowing period during which the incoming team sits alongside the outgoing one. This is necessary work and it is done, on the whole, competently. The trouble is that explicit knowledge was never the part at risk. It is, almost by definition, the part that can be written down, and much of it already has been.
What resists capture is tacit knowledge — the kind that Michael Polanyi had in mind, four decades ago, when he observed that we can know more than we can tell. It is the accumulated, mostly unconscious understanding that lets an experienced practitioner recognise a situation as familiar and act correctly without deliberating: that this reconciliation looks wrong in a way that matters and that one looks wrong in a way that does not; that this business unit’s numbers are always late and always fine, while that one’s being late is the first sign of something serious; that the workaround introduced three years ago for a system that has since been replaced is now doing nothing but should not be removed without checking, because two downstream processes quietly depend on its side effects.
- It is judgement — the ability to tell the significant exception from the routine one, which is precisely the part of the work that cannot be proceduralised, because if it could be, it would already be in the procedure.
- It is context — the memory of why things are the way they are, which turns an arbitrary-looking rule into a sensible precaution and lets someone know which rules can be safely broken and which cannot.
- It is relationships — the knowledge of whom to call when the documented escalation path fails, which is most of the times it matters, and the reservoir of goodwill that gets a favour done at quarter-end.
- It is improvisation — the tested repertoire of what to do when the process breaks, assembled over years of the process breaking, and never written down because the breaking was never supposed to happen.
The scholarship on this is not new; the knowledge-management movement of the last ten years was built on exactly this distinction between the tacit and the explicit, and produced libraries of writing on how tacit knowledge is socially held and slowly transferred. Yet the transition plans that movement should have informed still, overwhelmingly, count success in documents delivered. We know, in the literature, that the valuable knowledge is the hard-to-move kind. We plan, in practice, as though the valuable knowledge were the easy-to-move kind. The gap between those two sentences is where the drain occurs.
The Forces That Kept the Drain Invisible
If the loss were merely an oversight, it would have been corrected by now; oversights that cost this much tend to get noticed. That it persists, structurally and across the industry, tells us it is held in place by forces stronger than inattention. Four seem to me to do most of the work.
- The incentive horizon is shorter than the loss. The executive who sponsors a transition is measured on the transition: on delivering it to time and budget, on booking the saving, on standing the programme down cleanly. That accountability window closes at go-live, or shortly after. The knowledge drain, by contrast, manifests a year to two years later, on someone else’s watch, in a different reporting line, as a general malaise rather than a specific fault. No one is ever standing where the cause and the effect are both visible at once. The person who could have prevented it is rewarded and gone before it appears.
- The loss is unattributable, and so it is deniable. When month-end slips, there are a dozen candidate explanations and no way to prove which is right. You cannot run the counterfactual; you cannot demonstrate that the outage would not have happened had you retained the woman who understood the feeder system, because she is gone and the experiment cannot be re-run. An invisible cause with no proof of authorship is, organisationally, no cause at all. It becomes background noise, “the way things are since we moved to the new model,” absorbed into the baseline and forgotten.
- Accounting sees the release as pure gain. Because the departing capability was never on the books as an asset, its loss cannot appear as a cost. The business case is therefore structurally incapable of representing the trade it is actually making. It shows a saving with no offsetting charge, which is not a description of reality but an artefact of the chart of accounts. A decision framework that can only see one side of a two-sided transaction will make that transaction every time.
- The provider has no reason to reveal how thin the transfer was. A vendor is paid to take on the service, not to tell the client how much irreplaceable understanding failed to make the crossing. Where the transferred knowledge is shallow, the rational response — for a while — is to absorb the failures quietly, rebuild what can be rebuilt, and let the client attribute the rough patch to transition friction. The one party best placed to measure the drain is the party with the least interest in naming it.
Notice that none of these forces requires anyone to behave badly. Each actor is responding sensibly to the incentives and instruments in front of them. That is what makes the pattern so durable: it is not a failure of character but a failure of visibility, and visibility is a property of systems, not of individuals.
A Composite Reckoning
Let me put a shape and some figures to this, drawn not from any one organisation but composited from the recognisable arc of many.
A finance-and-administration function of around forty people is assessed as a candidate for a shared-service and offshore model. The business case is clean. A retained organisation of six is kept onshore for oversight and vendor management; the transactional work moves to an offshore team of fifty-five at a materially lower blended rate. Even after transition and governance costs, the projected saving lands somewhere around thirty per cent of the run cost — a number worth having, and defended, correctly, as such.
The transition is run properly by the standards of the day: a documented knowledge-transfer plan, a twelve-week shadowing period, process maps signed off, a hypercare window after go-live. On the dashboard, it succeeds. The service-level metrics are green from month two.
Then the second-order costs begin, none of them on the business case:
| Measure | Before | Eighteen months after |
|---|---|---|
| Month-end close | 5 working days | 9 working days |
| Open items at close | ~40 | ~180 |
| Retained-team time spent re-explaining context | negligible | roughly half of it |
| Queries escalated to the retained six | a handful a week | dozens a day |
The saving on the run rate is real and still there. But it is now partly financed by the retained organisation, which was sized for oversight and is instead spending half its week reconstructing context that used to live in the heads of people no longer employed. Month-end has slipped by four days, which cascades into late management reporting, which erodes exactly the decision-making tempo the transformation was meant to sharpen. Nobody put a figure on any of this, because there was no figure to put it against; the business case had no line called “capability retained,” so there was nothing to show it draining away. The thirty per cent is still booked. The tax on it is paid in a currency the ledger does not record.
The point of the composite is not that the deal was wrong — often enough the economics genuinely favour the move — but that the organisation made a thirty-per-cent decision while blind to a cost it had no mechanism to see. It was not weighing capability against saving. It was weighing saving against nothing, and nothing always loses.
The Case for Letting It Drain
Here I have to take the strongest version of the opposing argument seriously, because there is one, and it is not a straw man.
Much of what we sentimentally call institutional knowledge, the argument runs, is nothing of the sort. It is accumulated dysfunction dressed up as expertise: the undocumented dependency that should never have been allowed to form, the heroic individual whose indispensability is itself a governance failure, the tribal workaround that persists only because no one has been forced to fix the underlying process. On this view the knowledge drain is not a loss but a purge — a healthy, if painful, forcing function. When the person who “just knew” is gone, the organisation finally has to write the thing down, standardise it, make it robust and transferable and no longer hostage to one memory. Outsourcing, precisely because it is unsentimental, does what years of internal good intentions never managed: it converts fragile tacit knowledge into durable explicit process. Losing the knowledge is how you industrialise it.
This is a serious argument and it is partly right. Key-person dependency is a risk, not an asset. A great deal of what long-tenured staff carry around is sludge — habits that made sense against a system retired years ago, precautions no one can justify, complexity that exists only because no one ever had a reason strong enough to simplify it. A disciplined new operator will, and should, strip much of that out, and the organisation will be better for it.
But the argument proves less than it claims, and it fails at a single decisive point: it assumes we can tell the difference — and the whole problem is that we cannot, because we never measured. The purge is indiscriminate. It does not lift out the dysfunction and leave the judgement intact; it removes the person, and the person was carrying both, fused together, indistinguishable from the outside. The forcing function forces documentation of the knowledge you already knew you had — the explicit part, the part that was never really at risk. It does nothing to surface the tacit judgement that no one knew to write down, because the entire definition of that knowledge is that its holder cannot fully articulate it and the organisation cannot fully see it. So the drain does not selectively remove the cruft and industrialise the value. It removes both, indiscriminately, and calls the result standardisation because the part that survived is, by construction, the part that was already written down. You cannot industrialise what you were never able to name. The steel-man is right that some of what leaves should leave. It is wrong to assume the leaving is aimed.
What the Drain Reveals
Step back far enough and the knowledge drain stops being a problem about outsourcing and becomes a mirror held up to the organisation — one that shows, with unusual clarity, the shape of what it can and cannot perceive.
An organisation changes what it can see. Its instruments are overwhelmingly financial, and financial instruments were built to track things that can be counted in the period they occur. Capability — accumulated, distributed, tacit, slow to build and slow to reveal its absence — is exactly the kind of thing those instruments were not designed to hold. So when we reorganise, outsource, delayer, or consolidate, we move the visible things with great care and the invisible things with none, not out of recklessness but because the invisible things do not appear on the instrument panel we are flying by. The knowledge drain is what it looks like when an organisation confidently changes something it cannot measure, and mistakes the silence of its instruments for the absence of a cost.
The gap between transformation intent and transformation reality is very often just this: we intended to change the organisation, and instead we changed only the part of it our instruments could see.
This suggests something more demanding than the usual remedy. The usual remedy is better knowledge capture — more documentation, richer runbooks, longer shadowing. That helps at the margin, but it misunderstands the problem, because it is still trying to convert the tacit into the explicit and the whole difficulty is that most of it will not convert. The more honest response is not a better capture technique but a change in what we hold ourselves accountable for seeing. Before we release a capability, we might at least ask the questions our instruments do not: Who here holds a “why” that exists nowhere else? What breaks that is not in the runbook, and who is the only person who knows how to fix it? Where is our judgement concentrated, and what is our plan for the day it walks out of the door? These questions do not yield a number, and we should be suspicious of any method that pretends they do. Their value is not precision but attention. They make visible, for a moment, the asset we are about to dispose of, so that at least we dispose of it knowingly.
That is a modest ambition, and deliberately so. I am not proposing that we can put institutional knowledge on the balance sheet, or price tacit judgement to two decimal places; false precision about the immeasurable is its own failure, and arguably a worse one than honest neglect. What I am proposing is that the first discipline is to stop pretending the number is zero simply because it is hard to name. A cost you cannot measure is still a cost. An asset you cannot price is still an asset. The organisations that will navigate this decade’s great transfer of work with the fewest self-inflicted wounds will not be the ones with the most elaborate knowledge-management systems. They will be the ones that learned to be uneasy about the confidence of their own instruments — that treated the smooth green dashboard not as proof that nothing was lost, but as a reminder that the dashboard was only ever built to show them cost.
The people carrying the box out through reception are, on the register, a saving realised. Whether they are also a capability lost is a question the register cannot answer. The least we can do is ask it before they reach the door.