The Information Asymmetry Problem in Complex Programmes — Why More Reporting Buys You Less Truth
A status is a conclusion with the reasoning removed.
The Meeting Where Everyone Already Knew
The programme board meets on the third Thursday of the month, and by half past ten the pack has done its work. The summary slide is green. The milestone chart reconciles to the plan. Earned value sits within tolerance, the risk log has been groomed to a respectable amber, and the finance line ties cleanly to the last approved forecast. The sponsor thanks the programme director, observes that things look “broadly on track,” and moves the meeting to the next item on a crowded agenda.
And yet around that table, and in the rows of chairs behind it, a dozen people know the release will not hold. The integration testing that was supposed to begin a fortnight ago has not begun. The vendor’s strongest engineers were quietly rotated onto another account in the spring, and the replacements are still learning the estate. The number on the slide is not false — every figure can be traced to a source that would survive an audit — but the picture those figures compose is not the picture the delivery team would draw if you sat them down in a different room with no pack in front of them.
This is not, in the main, a story about dishonesty. It is a story about information, and about who controls it. The gap between what the board saw that morning and what the programme actually knew is not a defect in the reporting that a better template will close. It is a structural property of the relationship between the people who commission complex work and the people who do it. And the orthodox response to that gap — demand more reporting, tighten the governance, add another assurance layer — reliably makes it wider, not narrower.
The Asymmetry Is Structural, Not Accidental
The economists gave this its proper name a generation ago. Where one party — the principal — engages another — the agent — to act on its behalf, and cannot directly observe what the agent does, the two do not hold the same information and do not share the same interests. Jensen and Meckling set out the shape of it in the 1970s; Eisenhardt sharpened it for management a decade or so back. The literature was written about shareholders and executives, but it describes a programme board and its delivery organisation with uncomfortable precision.
Consider what the board can actually see. It cannot watch the code being written or the data being migrated. It cannot sit in the vendor’s stand-up. It sees the programme only through an instrument — the monthly pack — and that instrument is built, populated and calibrated by the very people whose performance it is meant to measure. The programme reports on itself. The status is self-assessed. When a workstream lead colours a milestone amber rather than red, no independent party verifies the choice; the colour is an opinion dressed as a measurement, and the person forming the opinion has spent eighteen months being rewarded for green.
That last point is the one the governance frameworks are strangely quiet about. The asymmetry would be manageable if the agent were merely a neutral relay with imperfect visibility. It is not neutral. Every incentive that matters — the next phase of funding, the supplier’s milestone payment, the director’s standing, the team’s sense that it is winning — pulls toward optimism. Bad news is expensive to deliver and cheap to defer. A slip disclosed today becomes a problem you own; the same slip disclosed in three months may, with luck and a following wind, have solved itself. So the rational agent does not lie. The rational agent rounds down the doubt, rounds up the mitigations, and reports a defensible best case. Do that across forty people and a dozen workstreams, month after month, and the aggregate picture drifts steadily away from the ground truth while every individual figure remains honest.
Information asymmetry in a complex programme is not a reporting failure to be fixed. It is a permanent condition of the principal-agent relationship, and it must be governed as a condition rather than cured as a defect.
I have watched this drift produce its most striking effect on a core-systems replacement — a composite of several I could name but will not — with a budget a little north of forty million pounds. For eleven consecutive reporting cycles the programme was green. Not aggressively green; sensibly, plausibly green, with the usual honest ambers on the risk log. Then, in a single cycle, it went from green to red — not through the courteous amber that governance imagines as the warning it will receive, but in one step, because the testing that everyone had privately known was in trouble could no longer be described any other way. The board’s genuine question afterwards was not “why did it fail?” It was “why did we not see it coming?” The answer is that they had been looking at the one instrument structurally incapable of showing it: a report written by the people who most needed it to stay green.
What the Textbooks Prescribe — and Why It Misses
The programme management canon treats this as a plumbing problem. Managing Successful Programmes, PRINCE2, the gateway review, the earned value discipline — each assumes that if you specify the right reports, define the right roles, and hold the right boards, accurate information will flow to the people who need it. Get the pipes right and the water runs clean. It is a coherent worldview, and within its own terms it is not wrong. It is simply answering a different question from the one the asymmetry poses.
| The textbook assumes | The practitioner finds |
|---|---|
| Reporting is a flow of facts upward | Reporting is an account produced by an interested party |
| The right template surfaces the truth | The template shapes which truths are cheap to tell and which are expensive |
| More frequent, more granular data means more visibility | More data widens the gap the principal must interpret, and the agent still curates it |
| Governance boards interrogate the position | Boards ratify a position they lack the time and expertise to interrogate |
| Amber is a warning received in time | Amber is a negotiating position; the real warning arrives as red |
The deepest of these is the third. The instinctive fix, when a board feels blind, is to ask for more — a weekly pack instead of monthly, twenty metrics instead of eight, a dashboard rather than a summary. But more data does not reduce asymmetry; it relocates it. A principal who could not interpret eight numbers cannot interpret twenty, and the labour of deciding which of the twenty matter falls back onto — the agent. The programme office helpfully “brings out the key messages.” The curation that the extra data was meant to defeat simply moves up a layer and puts on a suit. Volume is not visibility. Past a point, volume is camouflage.
The Strongest Case for the Orthodoxy
It would be too easy to leave the defence there, so let me put the orthodox case at its strongest, because it is not a foolish one.
The reply runs like this. Yes, the agent reports on itself — which is exactly why mature governance does not rely on self-report alone. That is what independent assurance is for: the gateway review conducted by people with no stake in the answer, the internal audit, the external technical review, the peer challenge. And yes, a colour is an opinion — which is why earned value exists, to replace opinion with arithmetic that is far harder to fudge, because it measures delivered value against a baseline rather than asking anyone how they feel. Layer independent assurance over disciplined measurement, the argument concludes, and you break the asymmetry by refusing to take the agent’s word for anything that matters.
This is the best objection, and it deserves a real answer rather than a dismissal. Here is where it breaks.
Independent assurance is periodic, and the programme knows the dates. A review that arrives at a gateway is a review the programme has weeks to prepare for; the estate is tidied, the narrative is aligned, the awkward workstream is given a fortnight’s air cover. Assurance samples a prepared position at a known moment. It does not live inside the programme between the moments, which is precisely where the drift accumulates. And earned value, for all its arithmetic rigour, measures progress against a baseline that the agent set and re-baselines against a scope the agent defines. If a task is booked as ninety per cent complete, earned value faithfully computes the consequences of ninety per cent — a figure that is itself, once again, a self-assessment. The discipline hardens the maths around a soft input. Rigour downstream of an interested estimate is rigour serving the estimate.
“Assurance samples a prepared position at a known moment. The drift accumulates in the moments in between.”
None of which is an argument against assurance or against earned value. Both are worth having. It is an argument against believing they dissolve the asymmetry. They raise its cost at the margin. They do not change who holds the information or which way the incentives point.
What Actually Changes the Picture
If the problem is structural and incentive-driven, then the effective responses are the ones that work on structure and incentives — not on the volume or frequency of the reporting. In my experience the boards that see furthest are not the ones with the richest dashboards. They are the ones that have changed what they ask for and what they reward.
- Ask for the decision, not the colour. A status is a conclusion with the reasoning removed. Replace “what colour is this?” with “what decision does this status imply, and what would have to be true for it to be wrong?” A colour can be defended indefinitely; a decision and its falsifying conditions cannot. The question forces the reasoning back into the room.
- Reward the disclosure, not the absence of bad news. As long as the first person to say “red” is treated as the person who caused red, every rational actor will wait for someone else to say it first. The board that visibly thanks the workstream lead who raised the alarm early — and never punishes them for it — is buying itself months of warning. This is the single cheapest intervention available and the one most consistently declined.
- Separate the reporter from the rewarded. Where the person who assesses status is the person whose bonus, standing or next contract depends on that status being green, the assessment is compromised by design. It need not be. The assurance function, the delivery confidence view, the person who colours the milestone — pull these apart from the people carrying the delivery incentive, and the self-report stops being a self-report.
- Buy unmediated contact. The most valuable thing a sponsor can do is spend an hour, without the programme director present, with the people actually doing the work — the tester, the migration lead, the analyst. Not to bypass the director, but to hold a second channel that the pack cannot curate. Ground truth is available; it simply does not travel through the reporting line, because the reporting line is the thing being examined.
Notice what these have in common. Not one of them asks for more information. Every one changes the conditions under which the information the programme already holds becomes cheap to tell. The asymmetry is never abolished — the agent will always know more than the principal, and should, or you have hired the wrong agent. But the terms on which that private knowledge surfaces are not fixed by nature. They are set by what the board rewards, and the board can change them any month it chooses.
Governing a Condition You Cannot Cure
The uncomfortable conclusion, for anyone raised on the frameworks, is that you cannot audit your way out of an incentive problem. The temptation in a tightening climate — and the business cases crossing programme boards this year are being interrogated more sceptically than they were eighteen months ago — is to answer a felt loss of control with more control: more reporting, more assurance, more governance. It is the natural reflex, and it treats the symptom while feeding the disease, because every additional layer of reporting is another surface the agent curates and another demand on the principal’s scarce attention.
The programmes that stay honest are not the most heavily governed. They are the ones whose leaders have understood that information asymmetry is a permanent feature of delegating complex work — not a bug to be patched, but a condition to be managed for as long as the programme runs. You govern it by working on the incentives that shape what people tell you and by holding more than one channel to the truth, so that the day the single instrument finally turns red is not the day you first suspected it might. The board that internalises this stops asking its reporting to do something reporting cannot do, and starts asking its people the only question that ever really mattered: not “what colour is it?” but “what do you know that this pack is not built to tell me?”