Governance as Decision Architecture
Data that informs is not the same as data that decides, and a governance culture that cannot tell the two apart will drown in the first while starving for the second.
Executive Summary
Somewhere in the past five years, the governance of large-scale change acquired an impressive apparatus and quietly lost its nerve. Steering committees multiplied, programme boards began meeting on a fixed monthly calendar, and the board pack swelled past forty pages. Yet across a striking number of transformation efforts, the machinery built to steer had stopped steering. It reported. It assured. It created a trail. On the questions that actually mattered, it rarely decided.
This essay is about that substitution — the slow displacement of judgement by reporting — and it makes a single argument: governance is not a reporting function that occasionally takes decisions; it is a decision-making function that occasionally needs reports. The distinction sounds pedantic until you watch a programme fail with a flawless audit trail. When we treat governance as the production of visibility, we optimise for the wrong thing, and the wrong thing is comfortable. What follows traces how the reporting reflex took hold, why the structural forces of this decade have made it almost irresistible, and what it would mean to rebuild governance as decision architecture: the deliberate design of who decides what, by when, on what basis, and how a decision — once made — is escalated, revisited, or unmade.
This is not a licence to abandon assurance. In the world the corporate scandals left us, and under the compliance regime that followed, that would be both naive and irresponsible. It is an argument that assurance is the by-product of governing well, not its purpose — and that a programme which reports beautifully while deciding nothing is not well governed but dangerously ungoverned.
The Green Report and the Stalled Programme
I have sat in the meeting I am about to describe more times than I can count, across more sectors than I care to name, and the details barely change.
A large transformation programme — a core-systems replacement, say, eighteen months in, with several tens of millions committed — holds its monthly board. The pack goes out on the Friday: an executive dashboard, a milestone chart, a risk register, a finance summary, and a set of workstream reports each ending in a red-amber-green status. The board convenes. Most of the ninety minutes is spent walking the dashboard. One workstream lead explains why integration has moved from amber to red. Another explains why testing has moved from red back to amber. The programme director narrates the milestone chart. Questions are asked and answered. The minutes will record that the board “reviewed progress and noted the key risks.”
And beneath all of it sits a single question that everyone in the room can feel and no one quite places on the table: the data migration will not be ready, and the go-live date four months out assumes that it will. That question was raised, softly, two months ago. It was raised again last month. Each time, the board agreed it “needed more information” and asked for deeper analysis next time. The analysis arrives, thicker each month. The decision does not.
A board that receives forty pages and resolves nothing has not been governed lightly; it has been governed theatrically. The pack is the performance of control, and the performance is convincing enough that the absence of an actual decision goes unremarked — until the date is missed.
When the date is finally missed — and it is missed, because unmade decisions do not wait patiently — the review will find that the migration risk was “known and reported throughout.” This is offered as exoneration. It is, in truth, the indictment. The risk was known. It was reported. It was reported for ninety days. And in those ninety days the one thing this body existed to do — choose between holding the date, moving the date, or de-scoping the first release so the date could stand — was never done. The decision latency was three months on a call that a single accountable owner, holding the very same facts, could have made in a week.
That gap — between the density of the reporting and the poverty of the deciding — is the subject of this essay.
How Reporting Came to Stand In for Governing
It would be too easy to blame weak individuals. The people in that room are usually able, senior and conscientious. The reporting reflex is not a personal failing; it is the rational response to a set of structural forces that this decade has assembled with unusual completeness.
The first force is the long shadow of the scandals that opened the decade. The collapses of the early 2000s taught boards, regulators and executives one lesson above all others: the mortal danger is not knowing. The remedy prescribed everywhere — in the revised corporate governance codes, in the internal-control guidance, and above all in the compliance apparatus that followed Sarbanes-Oxley — was visibility, documentation and traceability. Control came to mean the ability to demonstrate, after the fact, that you had seen what was coming and had kept a record. Against fraud and misstatement, this was a reasonable response. But its logic leaked into the governance of change, where the problem is a different animal. In a transformation programme the danger is rarely that no one saw the risk; it is that everyone saw it and no one resolved it. Against that danger, a model tuned for documentation is not merely unhelpful — it actively rewards the wrong behaviour, because it treats the recording of a risk as the discharging of it.
The second force is the assurance economy that has grown up around the compliance burden. By the middle of the decade an entire vocabulary — governance, risk and compliance, spoken now almost as a single word — had fused three once-distinct disciplines into a common reporting stack. Whole functions exist to populate it. The unit of work in that world is the report, and the natural tendency of any apparatus is to produce more of whatever it is built to produce. A programme sitting inside such an organisation inherits its reflexes: it reports because everything around it reports.
The third force is quieter and more human. Visibility feels like control. A well-built dashboard, refreshed monthly, colours in, provides the genuine and pleasant sensation of being on top of things. But the sensation and the substance come apart precisely when they matter most — when the dashboard is amber everywhere and the honest reading is that nobody in the room is willing to own the hard call. The colours give comfort; the comfort suppresses the discomfort that would otherwise force a decision.
“Visibility is not control. It is the sensation of control — and the two diverge exactly when it matters most.”
The fourth force is the plain arithmetic of incentives. Consider the individual sitting on that board. If they force a decision and it proves wrong, the decision has their name on it. If they call instead for more analysis, they appear rigorous, they buy time, and if the programme later stumbles, the record shows they flagged the risk. Under those incentives, deferral is not weakness; it is the dominant strategy. The reporting reflex is what you get when a whole room, each member behaving rationally, converges on the option that protects everyone and commits no one.
There is a fifth force, and it is the one this library too often overlooks: the nature of the information itself. The modern board pack is thick with data — burn-down curves, defect counts, resource utilisation, a risk register scored on likelihood and impact. Almost none of it is shaped for a choice. It is data as exhaust — the by-product of running the programme, swept up and presented — rather than data assembled backward from a decision that has to be made. A migration risk scored “high / high” on a register tells the board to feel concerned. It does not tell them whether to hold or move the date, because it was never built to. Data that informs is not the same as data that decides, and a governance culture that cannot tell the two apart will drown in the first while starving for the second.
What Governance Is Actually For
Strip the apparatus away and ask what governance is for. In any complex endeavour with distributed authority and irreversible commitments, governance exists to answer four questions and only four: who is entitled to decide what; on what basis; by when; and how a decision, once taken, can be challenged, escalated or reversed. Everything else — the dashboards, the packs, the standing reports — is instrumental to those four questions or it is waste.
Read against that standard, most transformation governance is answering a different question entirely. It is answering what is the current status? — a question worth asking, but a subordinate one. Status is an input to a decision, not a substitute for it. A governance body that spends ninety per cent of its time establishing status and less than ten per cent making choices has inverted its own purpose.
To govern as decision architecture is to design the decision-making of a programme as deliberately as we design its technical architecture. It means treating decision rights, decision timing, decision information and decision reversibility as first-class objects to be specified — not as things that will sort themselves out once the reporting is in place. The contrast is stark enough to set out directly.
| The reporting reflex | Decision architecture |
|---|---|
| Organises the forum around status | Organises the forum around decisions required |
| Produces information as a by-product of running the work | Assembles information backward from the choice to be made |
| Treats a recorded risk as a discharged one | Treats a risk as open until a decision closes it |
| Meets on the calendar | Convenes when a decision is due |
| Gates confirm progress | Gates can stop or redirect the work |
| Success looks like a complete, current pack | Success looks like a short decision latency |
The right-hand column is not more bureaucracy. In almost every case it is less — fewer pages, shorter meetings, smaller forums — because once a body knows which decisions it owns, most of what currently fills the pack is revealed as material no one needs in order to choose.
The Strongest Case for the Reporting Reflex
An essay that only prosecuted the reflex would be a lighter thing than the problem deserves, because there is a serious case on the other side, and it is not the case of the timid. It runs like this.
We do not live in a world that rewards travelling light on documentation. The board of a listed company carries statutory duties of oversight; auditors and regulators expect a traceable basis for material commitments; the compliance regime of this decade exists precisely because organisations once acted on judgement without a record and destroyed themselves doing it. A transformation programme consuming tens of millions is exactly the kind of commitment that oversight is meant to catch. Against that backdrop, a dense reporting apparatus is not theatre — it is the discharge of a duty. And a “decision-first” governance that prizes speed and travels light on the paper trail is not brave; it is the very thing the scandals warned us against, dressed up as dynamism.
This is the strongest version of the objection, and it deserves a straight answer rather than a caricature.
The answer is that it rests on a false opposition. Decision architecture does not abandon the record; it produces a better one. A well-made decision generates a superior trail to a well-made report, because it captures what a status update never does: the options that were genuinely on the table, the basis on which one was chosen, the person accountable, and the date by which the choice will be revisited. That is precisely the material an auditor, a regulator or a successor most wants, and precisely what the forty-page pack, for all its bulk, does not contain. The pack records that a risk was seen; the decision log records what was done about it.
And there is a sharper point. The control failure the scandals warned against was never, at root, a failure to see. In most of those collapses the warning signs were visible and, in some cases, documented. The failure was a failure to act on what was seen — to escalate it, to halt the thing, to make the unwelcome decision. A governance model that faithfully reports a fatal risk for ninety days and never resolves it has not avoided that failure mode. It is that failure mode, in a more respectable suit. Assurance that assures you of a programme quietly heading for the rocks is not assurance at all.
Designing the Decision Architecture
Diagnosis is the easy half. The harder and more useful question is what a governance body actually does differently on the Monday morning after it decides to stop reporting and start deciding. The following are not a maturity model or a framework to be certified against; they are design principles, drawn from watching the reflex give way in the places where it has given way at all.
- Name the decision, not just the status. Every governance forum should carry a standing item that is a list of decisions required — each with the choice framed, an accountable owner, and a drop-dead date beyond which deferral is itself a decision, and a worse one. If a forum has no decisions before it, it need not meet.
- Assign decision rights explicitly. Be precise about who decides, who must be consulted, and who is merely informed — and about which decisions belong to the board, which to the programme director, and which should never have climbed so high in the first place. Ambiguity about decision rights is the soil in which deferral grows: when no one is quite sure they own a call, no one makes it.
- Design information backward from the choice. A paper exists to enable a specific decision. Begin with the decision, ask what a reasonable person would need in order to make it well, and provide that — and only that. If a chart changes no one’s action, it is not analysis; it is decoration, and it should be cut.
- Make gates genuine forks. A stage gate that has never once stopped or redirected a programme is not a gate; it is a toll booth with a ceremony attached. A real gate has a live possibility of no — and the organisational nerve to say it. Design the gate around the two or three questions whose answers could genuinely halt the work, and be willing to act on them.
- Set cadence by decision need, not the calendar. Some decisions cannot wait a month; some forums need not meet at all this month. The monthly rhythm is an administrative convenience mistaken for a governance principle. Convene when a decision is due, at the tempo the decision requires, and let the reporting settle into whatever quieter rhythm is genuinely useful.
- Separate the reversible from the irreversible. Not all decisions carry the same weight. A choice that can be walked back cheaply should be delegated low and made fast; the cost of being wrong is small and the cost of delay is real. A choice that cannot be undone — a data cut-over, a decommissioning, an irrevocable contractual commitment — deserves the board’s fullest and slowest attention. Governing both at the same ponderous cadence is how organisations agonise over the trivial and rush the irreversible.
- Measure decision latency. Track the elapsed time from the moment a decision becomes necessary to the moment it is made. It is the one vital sign the red-amber-green report cannot show, and it is the truest measure of whether a governance body is governing. A programme whose risks are all amber but whose decisions routinely take a quarter to surface is not stable. It is stalled, and reporting its stability.
None of this dispenses with reporting. It puts reporting back in its place — as the servant of decisions rather than their substitute. A short, honest status update still matters; it is simply no longer allowed to be the meeting.
The Longer View
Step back far enough and the reporting reflex tells us something uncomfortable about the state of our craft. We have become extraordinarily fluent in the machinery of change — the methods, the frameworks, the stage gates, the registers, the packs. We are far less fluent in the thing all that machinery exists to serve, which is judgement: the willingness of an accountable person to look at an imperfect picture and choose. The apparatus has outrun its purpose. We have built ever more elaborate instruments for observing a programme and, in the building, quietly forgotten that observation was only ever meant to serve steering.
This is, at bottom, the same gap that opens between every transformation’s intent and its reality. The intent is stated in the language of decisive change — new capabilities, retired legacy, a different way of working. The reality is too often a beautifully documented drift, in which the programme is watched with great care as it fails to arrive. The distance between the two is not usually a distance of effort or of information. It is a distance of decisions not made, or made too late, by bodies that mistook the reporting of a problem for the resolving of it.
The remedy is not more governance. It is governance that remembers what it is for. If a board can look at its own last six months and count the decisions it actually made — not the reports it received, not the risks it noted, but the forks at which it chose one path over another and owned the choice — it will know very quickly whether it has been governing or merely watching. Most boards, asked that question honestly, are chastened by the answer. That chastening is the beginning of the repair, because the first decision any governance body can make is the decision to start deciding.