What the research shows
The research on organisational governance is older and less flattering than most leadership teams expect. It does not describe governance failure as an event. It describes it as a slow divergence between how an organisation believes decisions are made and how they are actually made, and it locates the cause in structure and information flow rather than in individual competence.
The most direct evidence sits in work on decision rights. Paul Rogers and Marcia Blenko, both partners at Bain & Company, drew on a survey of executives at more than 350 global companies for their 2006 Harvard Business Review article on decision roles, and reported that only about 15 per cent of those surveyed said they had an organisation that helped the business outperform its competitors. That figure is worth sitting with. It is not a measure of strategy quality or market position. It is a measure of whether the organisation itself — its reporting lines, its authority, its meetings, its escalation paths — was experienced by senior people as an asset rather than a drag. Roughly one in seven said yes.
Rogers and Blenko also described where decisions characteristically stall, and their description is useful precisely because it is descriptive rather than prescriptive. They observed decision-making getting stuck at recognisable boundaries: between global and local, between the corporate centre and the business units, between one function and another, and between the organisation and its outside partners. These are the seams of an organisation. What the observation suggests is that decision problems concentrate where authority has to cross a boundary, not where it sits within one. A team that runs its own work well can still be part of an organisation that decides badly.
The second body of evidence concerns why arrangements that once worked stop working. Larry Greiner’s account of organisational growth, published in Harvard Business Review in 1972 and republished as an HBR Classic in 1998 with his own later commentary, made an argument that has held up better than most: the practices that resolve one period of an organisation’s life become the problem in the next. Greiner wrote that managers experience “the irony of seeing a major solution in one period become a major problem in a later period.” Applied to governance, this is a strong claim. It says that a governance arrangement causing visible difficulty today may be the same arrangement that made the organisation successful five years ago, unchanged and still working exactly as designed. The arrangement did not break. The organisation moved.
A third strand concerns what leadership teams know. Amy Edmondson’s field study of 51 manufacturing work teams, published in Administrative Science Quarterly in 1999, found that team psychological safety predicted learning behaviour, while team efficacy did not once psychological safety was controlled for, and that learning behaviour in turn mediated team performance. The finding is usually read as a point about teams. Read as a point about governance, it says something sharper: the willingness of people to raise problems is a property of conditions rather than of character, which means an oversight arrangement can be formally sound and still receive incomplete information.
Michael Beer and Russell Eisenstat, then both at Harvard Business School, made a related argument in Sloan Management Review in 2000, describing a set of mutually reinforcing barriers to strategy implementation and organisational learning — among them unclear priorities and poor vertical communication — that operate largely out of sight of the senior team. Their central point is that these barriers reinforce one another, so examining any one of them in isolation understates the difficulty.
Finally, there is a strand of work on the sheer load that coordination places on an organisation. Jay Galbraith’s 1974 paper in Interfaces argued that task uncertainty is what drives organisational form: the more uncertain the work, the more information has to be processed during execution, and the organisation’s structure is in large part a response to that demand. Governance arrangements are among the mechanisms carrying that load. When the load rises faster than the arrangements, the symptom is rarely a formal failure. It is delay, duplication and quiet re-interpretation.
Governance rarely fails loudly. It stops fitting, and the organisation adapts around it without deciding to.
— Zoom Riddle
Where decisions actually get made
Ask a leadership team who decides a given class of question and you will usually get an answer. Ask three people separately and the answers diverge more often than the team expects. This is not evidence of dysfunction. It is evidence that authority in most organisations is a mixture of what is documented, what is remembered and what is inferred, and that the three drift apart at different rates.
The documented version is the org chart, the delegated authority schedule, the terms of reference. It changes rarely and by deliberate act. The remembered version is what people learned when they joined, or when they last watched a decision of that type actually get made. It changes when people change roles or leave. The inferred version is what someone concludes from the behaviour around them — who was consulted last time, whose objection stopped something, whose silence was treated as consent. It changes constantly, and it is the version most people actually act on.
For as long as the three stay roughly aligned, the difference is invisible. They come apart under specific conditions. A reorganisation moves a function but not the responsibilities that had accumulated to individuals inside it. A senior hire takes on a title without inheriting the informal authority the previous holder had built. An acquisition brings in a group operating on an entirely different inferred model of who decides. In each case the documented version is unchanged, which is precisely why nobody notices it has become inaccurate.
The consequence that leadership teams tend to notice first is not a wrong decision. It is a decision that will not stay made. A question is settled in one forum and reopened in another, not out of bad faith but because the second forum genuinely believes it holds the authority. Each round of reopening is individually defensible. The cumulative effect is that the organisation spends a growing share of its senior attention on questions it has already answered, and the people below become correspondingly reluctant to act on any answer until they see whether it survives.
A second consequence is escalation that does not resolve. When ownership is ambiguous, escalation looks like the safe move: pass the question up, and let clearer authority settle it. But if the ambiguity is horizontal — two functions with equal claim — passing it up delivers it to someone who must now reconstruct context they do not have, in order to arbitrate between two parties who both believe they were right to escalate. The question comes back down slower and no better resolved. Over time people learn that escalation is expensive and slow, and stop using it for the marginal cases, which are exactly the cases where early visibility would have mattered.
There is a third pattern worth naming because it is so often misread. Decisions can be made perfectly well and still not be governed. A capable manager who consistently makes good calls in an area nobody has formally assigned is producing good outcomes and, simultaneously, an unrecorded dependency. The quality of the decisions conceals the absence of the arrangement. Nothing appears wrong until that person changes role, at which point the organisation discovers that a body of judgement it relied on was never held anywhere but in one person’s head.
This is where reporting lines become misleading as a description of authority. A reporting line describes who manages whom. It does not describe who may commit the organisation, who must be consulted, whose objection is decisive, or which decisions belong to a role rather than to the person currently in it. In practice, organisations often find that these two things — the management line and the decision structure — were never the same, and that they had been treating one as a proxy for the other for years without difficulty, right up until the moment growth made the difference material.
Informal governance and the point at which it stops working
Informal governance is not a deficiency. It is an efficient arrangement that suits a particular set of conditions, and it works genuinely well while those conditions hold. In a small organisation, the people who need to know something are in the room. Context is shared without being transmitted. Authority is legible because everyone can observe how it is exercised. The cost of writing any of this down would exceed its value, and organisations that resist doing so early are usually right.
What makes this territory difficult is that the conditions do not fail visibly. They erode along several dimensions at once, and none of them announces itself.
The first is headcount, though less directly than people assume. What matters is not how many people there are but how many of them can observe decisions being made. An organisation of forty in one building may have better shared context than one of fifteen across three time zones. Distance, whether geographic, functional or temporal, removes the observation that informal authority depends on.
The second is tenure mix. Informal arrangements are transmitted by exposure. A team where most people have been present for years carries a great deal of unwritten understanding. The same team after a period of rapid hiring contains a majority who never observed the original arrangements and are inferring them from a much thinner sample. The arrangements have not changed. The proportion of people who accurately understand them has.
The third is the number of decisions the organisation must make per unit of time. This is the dimension Galbraith’s argument bears on most directly: as uncertainty and volume rise, more information must be processed during execution rather than settled in advance. Informal governance handles low volume well because each case can be reasoned through individually. It handles high volume poorly for exactly the same reason.
The fourth is discontinuity. Reorganisations, acquisitions and senior departures all break the transmission chain. Greiner’s observation is precise here: the arrangement was a solution, and it is now a problem, and it is the same arrangement. The organisation’s difficulty is not that it did something wrong. It is that it did something right and then changed the conditions under which that thing was right.
Founders occupy a particular position in this, and the usual account of it is unfair. The standard story is that founders struggle to let go. What the pattern more often looks like from inside is that the founder is the only remaining holder of a large amount of context that was never externalised, because externalising it was never necessary. Decisions genuinely are faster and often better when they route through that person, and will remain so until the context exists somewhere else. The reluctance is frequently an accurate reading of the current state rather than a failure of temperament. That does not make the dependency less real, but it does mean the difficulty is structural rather than personal, and treating it as personal tends to make it worse.
The point at which informal governance stops working is therefore not a threshold anyone crosses. It is a widening gap between the number of decisions requiring shared context and the number of people who hold it. Organisations typically detect the gap through symptoms that appear unrelated to governance: slower decisions, more escalation, more rework after decisions, and an increasing number of conversations that begin with someone explaining background that everyone used to have.
Oversight, bureaucracy and compliance
Three distinct things are routinely treated as one, and the conflation does real damage to the conversation.
Oversight is the capacity to see what is happening and to intervene where necessary. Bureaucracy is procedural weight — steps, approvals and documentation that consume effort. Compliance is conformity with external requirement. They overlap, which is why they get confused, but they are not versions of one another, and an organisation can have any one without the others.
The most consequential confusion is between oversight and bureaucracy, because it makes the trade-off appear simpler than it is. If oversight and bureaucracy are the same thing, then more of one necessarily means more of the other, and the question becomes a dial to be set. In practice, the relationship is much weaker. Some of the heaviest procedural apparatus produces very little visibility — approvals routed to people with no basis for judgement, reviews conducted on documents nobody reads, sign-offs that record that a form reached a desk. Conversely, some arrangements that produce genuine visibility cost almost nothing procedurally, because they consist of information already generated by the work being routed somewhere it can be seen.
What this suggests is that procedural weight is a poor proxy for oversight, in either direction. An organisation reducing bureaucracy may or may not be reducing oversight. An organisation adding oversight may or may not be adding weight. Leadership teams that treat the two as interchangeable end up making the wrong argument with each other: one side defends visibility by defending procedure, the other attacks procedure and is heard as attacking visibility, and the actual question — what does the organisation need to be able to see — never gets asked directly.
There is a genuine tension underneath, and it does not resolve cleanly. Any mechanism that produces visibility consumes attention, and attention is the scarcest resource in a growing organisation. The costs of oversight are immediate, visible and fall on identifiable people. Its benefits are diffuse, delayed and largely counterfactual. Nelson Repenning and John Sterman of MIT Sloan made this asymmetry explicit in their work on process improvement, published in California Management Review in 2001, describing how organisations under pressure default to working harder rather than working smarter, and how the people who prevent problems receive no visible credit for problems that never occurred. Oversight is squarely in that category. Its successes are absences.
Compliance sits differently again, and the difference matters more than it appears. Compliance answers a question posed from outside: does this organisation meet a defined external requirement? Governance answers a question posed from inside: does this organisation know what it is doing, and can it steer? An organisation can be fully compliant and poorly governed, because the external requirement covers only what an external party thought to require. It can also be well governed and technically non-compliant, having judged a requirement to be met in substance by different means.
The reason this matters practically is that compliance is measurable and governance largely is not. Compliance produces artefacts, dates and pass rates. Governance produces the absence of surprises, which no one can count. V. F. Ridgway’s 1956 paper in Administrative Science Quarterly on the dysfunctional consequences of performance measurement is the durable warning here: he examined single, multiple and composite measures and found all three capable of producing undesirable consequences for overall performance, concluding that the cure is sometimes worse than the disease. Where a compliance metric is the only available reading of governance health, the metric tends to become the object of attention, and effort migrates towards the measurable portion of a broader question.
In practice, organisations often find that their compliance record was strongest in the period immediately before a governance problem surfaced, because the two were never measuring the same thing.
What governance can and cannot see
Every governance arrangement is a viewing position. It sees some things clearly, some things late, and some things not at all, and the pattern of what it misses is a property of the arrangement rather than of the people operating it.
Distance is the first constraint. Information degrades as it travels, and it degrades asymmetrically. Quantitative information travels well — a number arrives at the top looking much as it did at the bottom. Contextual information travels badly, because it is the part that has to be summarised. Ambiguous information travels worst of all, because summarising it requires someone to resolve the ambiguity, and whoever does so is making a judgement that then becomes invisible. By the time an item reaches a board, it has typically been resolved several times by people who did not think of themselves as deciding anything.
This is where Edmondson’s finding becomes structurally, rather than culturally, interesting. If the willingness to raise a problem depends on conditions rather than on individual courage, then the same organisation can have entirely different reporting behaviour in two adjacent units, and the difference will not show up in any formal description of its governance. The arrangement is identical. What arrives through it is not. Beer and Eisenstat’s argument runs parallel: the barriers to information reaching the senior team are mutually reinforcing, so the senior team’s view of its own visibility is itself formed through the compromised channel.
The second constraint is that governance sees events far better than it sees drift. Discrete occurrences — an incident, a breach, a missed deadline — have edges, dates and owners. Drift has none of these. A responsibility that migrates gradually across three reorganisations never generates an event. A standard that is relaxed by small increments never crosses a line. A control that is worked around occasionally, then routinely, then permanently, produces no moment at which anything changed. What Erik Hollnagel, Robert Wears and Jeffrey Braithwaite described in their 2015 white paper on safety, published jointly by the University of Southern Denmark, the University of Florida and Macquarie University, applies directly: everyday work is routinely adapted from how it was formally imagined, and things go well largely because people make sensible adjustments to the demands of the situation rather than because they follow procedure exactly. The adaptations are usually improvements. They are also, from a governance point of view, invisible, which means the organisation’s documented understanding of its own operation becomes progressively less accurate in ways that nobody has done anything wrong to cause.
The third constraint is that the reward structure surrounding governance frequently points away from disclosure. Steven Kerr’s 1975 paper in the Academy of Management Journal on the folly of rewarding one thing while hoping for another remains the clearest statement of the general problem: people determine what is actually rewarded and do that, sometimes to the exclusion of what is stated. Applied to governance, the question is what an organisation actually rewards when someone raises an unwelcome and uncertain signal early. If early raising is met with additional scrutiny of the raiser, additional work, and no recognition when the concern turns out to be well founded, then the stated preference for early visibility is not the operative one, whatever the policy says.
Expansion sharpens all three constraints at once, which is why governance questions tend to multiply after a period of growth rather than during it. New geographies introduce genuine differences in local requirement and practice, and the centre cannot always distinguish a necessary local variation from an unnecessary one. Acquisitions import an entire second set of inferred arrangements, which coexist with the first for far longer than anyone plans. The volume of decisions rises before the arrangements for making them do. The board’s own position shifts in this period too: the questions it can usefully ask about a business its members can still hold in their heads are not the questions available to it once the business exceeds that, and the shift is gradual enough to go unremarked for some time.
Why this matters
The costs in this territory are unusually easy to underestimate, for two reasons.
The first is that they are almost entirely opportunity costs, and opportunity costs do not appear anywhere. Senior attention spent re-settling a question that was already answered is attention not spent on something else, but no ledger records it. A decision that took six weeks rather than two produced an outcome, and the outcome looks the same as one that took two weeks. A capable person who left because they could not get decisions made is recorded as a departure, not as a governance cost. None of this shows up as a line item, which means the organisation’s own information systems are structurally incapable of surfacing the price it is paying.
The second reason is compounding. Ambiguity in decision rights generates behaviour that increases ambiguity. When people cannot be certain a decision will stand, they seek broader consultation before acting, which draws more parties into the decision and makes future ownership less clear. When escalation is slow, marginal issues stop being escalated, which reduces the information reaching the people who would have needed it. When oversight arrangements produce weight without visibility, they attract justified criticism, which makes the next attempt at visibility harder to argue for. Each step is a reasonable response to the previous state.
It is also worth being direct about what the evidence does not support. Claims that a specific percentage of organisations fail at governance, or that governance failure costs a specific proportion of revenue, circulate widely and are almost never traceable to a stated methodology. Mark Hughes of the University of Brighton examined five separate published instances of the frequently repeated claim that 70 per cent of change initiatives fail, and found no valid or reliable empirical evidence supporting any of them. His subject was change rather than governance, but the lesson transfers: a figure that everyone repeats and nobody sourced is not evidence, and building an argument on it weakens the argument.
What can be said with more confidence is narrower and more useful. Cross-boundary work is where the difficulty concentrates. Donald Sull of MIT Sloan, with Rebecca Homkes and Charles Sull, drew on a multi-year study of more than 250 companies and several thousand managers, reported in Harvard Business Review in 2015, and found that only 9 per cent of managers said they could rely on colleagues in other functions and units all of the time, with around half saying they could rely on them most of the time. Their broader conclusion was that the central execution problem is cross-unit coordination rather than vertical alignment, and that alignment processes in most firms are sound. Governance is largely a set of arrangements for coordinating across boundaries. If that is where reliability is weakest, it is a reasonable place to expect the cost to sit.
How this presents in practice
Consider an organisation where a compliance requirement is missed. The investigation that follows is short, because the facts are not in dispute. Three people each believed one of the others was responsible for it.
The obvious reading is individual carelessness, and in some organisations that reading would be correct. Here it is complicated by what the investigation also finds: all three had reasonable grounds for their belief. The responsibility had migrated informally across two reorganisations. On each occasion the work moved with a team, or stayed behind when a team moved, in ways that made sense at the time. No document was updated, because no single change was large enough to obviously require it. Each of the three inherited an assumption from a predecessor, and each had good reason to trust it.
What is notable is that nobody involved made a mistake in the ordinary sense. The requirement had an owner at every point in its history. What it did not have, at any point after the first reorganisation, was an owner who knew they were the owner. Ownership was never explicitly transferred. It was inferred, repeatedly, by people acting in good faith on incomplete information, and each inference was individually reasonable given what that person could see.
Organisations of this size often find that the same conditions apply to a good deal else. If a responsibility can migrate this way once, the arrangement that allowed it is not specific to that responsibility. The question a leadership team is left with after an episode like this is less about the missed requirement than about the population it belongs to: how does an organisation know which of its responsibilities have owners and which merely have assumptions?
That question is genuinely hard to answer from inside, because the two look identical until tested. An assumed responsibility and an owned one produce the same appearance of coverage, right up to the moment coverage is required.
This scenario is an invented composite constructed to illustrate a pattern. It does not describe any actual organisation.
Continue reading
More on governance
- Decision Rights: The Written Rule and the Rule in Force
- Why Governance Questions Multiply After Expansion
- Why Informal Governance Works Until It Doesn’t
- The Board’s Role as an Organisation Scales
- Why Decision Rights Become Blurred in Growing Companies
Related areas
Looking for your own sector? The Other Sectors page points to the closest match.
Questions worth considering
These are open questions rather than a checklist. They are the questions this territory raises, not a procedure for working through it, and there is no order implied.
- Where in the organisation would a disagreement about who decides be most likely to surface, and what makes that the place?
- Which decisions currently work well because of a particular individual rather than because of an arrangement, and how would anyone find out?
- What was the last decision that had to be made more than once, and what made the second forum believe it held the authority?
- Which parts of the organisation can observe how decisions are actually made, and which are inferring it?
- What does the organisation currently reward when someone raises an uncertain concern early and turns out to be wrong?
- Which of the existing oversight arrangements produce visibility, and which produce mainly weight?
- What would distinguish them from the outside?
- What has changed since the current arrangements were set, and which of those changes altered the conditions the arrangements assumed?
- If a responsibility migrated quietly over the past two years, what would have made that visible, and would anything have?
Related perspectives: Strategy Execution and Leadership.
Your team is facing similar challenges?
Subscribe to The Riddle Ledger
Perspectives are sent as they are published, with no promotional email, and you can unsubscribe at any time.