The Most Dangerous Three Colours
The board does not see a status. It sees the accumulated optimism of everyone beneath it.
{ “Title”: “The Most Dangerous Three Colours: Why RAG Status Reassures When It Should Warn”, “Subtitle”: “How a reporting convention built to surface trouble became the mechanism that buries it”, “Series”: “Programme”, “Mode”: “Standard”, “Type”: “Perspective”, “Beyond Hook”: “Beyond the Traffic Light”, “Quote”: “The board does not see a status. It sees the accumulated optimism of everyone beneath it.”, “Year”: “2007”, “pretend_date”: “2007-08-25”, “real_date”: “2026-07-26”, “Article Version”: 1, “Template Version”: “v1.0”, “content”: “<verbatim paper content stored in the fenced code block below>” }
The pack was green
I have sat in the room where a steering committee received a programme report that its chairman summarised, with evident relief, as “broadly green” — and watched the meeting break up early because there was, it seemed, nothing left to discuss. Eleven of the twelve workstreams carried a green light. The twelfth was amber, and had been amber for four consecutive reporting cycles, but amber was “in hand.” Some months later the same programme was the subject of an emergency review: its go-live pushed out by two quarters, its sponsor asking, with real bewilderment, how a green programme could have run aground without anyone raising a hand.
It had not run aground without warning. The warning was on every page of every pack. It was simply printed in a colour no one had been trained to fear.
The Red-Amber-Green convention — the roadside traffic light transplanted into the board report — is the most trusted and least examined instrument we have. We reach for it because it crushes a mass of detail into a single glance, and a busy board is grateful for the glance. But the compression that makes RAG convenient is exactly what makes it dangerous, and the danger is not a matter of one organisation using it sloppily. It is built into the three colours themselves. RAG does not fail because people misuse it. It fails, reliably, when they use it exactly as intended.
A colour anyone can agree to
A RAG status has the appearance of a reading taken from an instrument. It is nothing of the sort. It is a judgement, formed by a person, recorded in a document that travels upward through a hierarchy in which that person’s own standing rises and falls with the colour they choose. That single fact bends everything that follows.
Green is the resting state of the system. To report green is to close the item: no explanation required, no action logged, no return visit scheduled. To report amber is to volunteer for work — it obliges you to account for the slippage, produce a recovery plan, and submit to the question, next period, of whether the recovery held. Red is heavier again: red brings the sponsor into the room, and with the sponsor the unspoken suggestion that you are no longer in command of your own patch.
| Colour | What it invites | What it costs the person reporting it |
|---|---|---|
| Green | Nothing; the item closes quietly | Nothing today — the bill arrives later |
| Amber | Questions, a recovery plan, a follow-up next period | Visibility, scrutiny, the appearance of not coping |
| Red | The sponsor, an intervention, an audit of your judgement | Standing, autonomy, the benefit of the doubt |
Read that table as the person filling in the report, at nine o’clock the night before the pack is due, and the drift toward green is not dishonesty. It is ordinary human sense. The colour that costs least tonight is green, and the cost of green falls in a later period — on a later version of you, or on someone else entirely.
The same rounding then happens at every joint of the hierarchy. A team lead softens a hard amber to “amber, trending green” because the programme manager dislikes surprises. The programme manager, assembling twelve such judgements, reports the portfolio as green because eleven are green and the twelfth is “being managed.” By the time the colour reaches the board it has passed through three or four hands, each with a small, reasonable, local incentive to warm it by half a shade. The board does not see a status. It sees the accumulated optimism of everyone beneath it.
A RAG status is not a measurement of the programme. It is a measurement of how safe the people running the programme feel about telling the truth this month.
The information a colour cannot hold
Suppose we solve the honesty problem entirely — every status set by a saint, every criterion applied without fear. RAG would still mislead, because three colours cannot carry the information a decision actually needs.
A colour is a snapshot with no direction. It tells you where the programme stands and nothing about which way it is moving, or how fast. Two workstreams both report amber. One was red last month and is clawing its way back; its amber is the best news in the pack. The other has been amber for a quarter and is quietly sliding toward red; its amber is a slow-motion emergency. The report gives them the identical colour, and in doing so discards the one thing the reader most needs — the trajectory.
A colour also carries no confidence. A green declared because the manager has tested every dependency and found them sound is a wholly different object from a green declared because no one has looked. On the page they are indistinguishable. The most dangerous status in any pack is the confident green with nothing behind it, precisely because it is the one no one thinks to question.
And a colour hides the shape of a programme. The item on the critical path and the item with weeks of float both occupy one row and one light. I have watched a programme report seventy-eight per cent of its milestones “on track” and green in the very month its delivery date moved by five months — because the milestones being hit were the small, local, low-risk ones, while the single integration dependency that governed the end date slid, amber and unescalated, in a row that looked no different from the rest. The average of many local truths was a global lie, and the colour was the instrument that did the averaging.
“The tool is fine; the discipline is poor”
The seasoned objection to all of this deserves to be stated at full strength, because it is not foolish. It runs: none of these are faults of RAG. They are faults of governance. A well-run scheme defines objective, measured criteria for each colour, holds reporters to them, and rewards the honest amber instead of punishing it. Fix the discipline and the tool works. Abandoning RAG because it can be gamed is like abandoning the accounts because someone once cooked the books.
There is real force here, and better criteria genuinely help. But the objection concedes more than it intends. If a green is only trustworthy when it sits on top of rigorous, evidenced, independently held criteria, then it is not the colour doing the work — it is the evidence beneath it. The colour is merely a lossy summary of an analysis that had to exist anyway, and the reader who acts on the colour rather than the analysis is trusting the summary exactly where it is thinnest. Worse, the discipline the objection demands runs against the grain of the situation permanently. The reporting act is social, performed upward, under asymmetric consequences, in every organisation, in every period. You can suppress the drift toward green with enough governance effort, but the moment attention lapses it returns, because the incentive never left. A tool that is safe only under constant vigilance is not a safe tool. It is a hazard with a good safety record.
Demoting the colour
The remedy is not to abolish RAG. The colour is a serviceable headline, and boards will always want a glance. The remedy is to demote it — to insist that the colour is the headline of a story and never the story itself, and to rebuild the report around the three things the colour cannot hold.
- Report direction before level. Every status should state whether it is better or worse than last period, and why. “Amber, and worse than last month for the third month running” is a sentence that cannot hide; “amber” on its own hides easily. The trajectory is the information; the colour is only where it starts.
- Make green expensive and amber cheap. Invert the tax. Green should be the colour that must be earned — accompanied by the evidence and the tested dependencies that justify it — while amber should be the cheap, unremarkable, no-blame default for anything not yet proven. When green is the effortful colour and amber the ordinary one, the drift reverses.
- Attach the threat to every status. No status is complete without naming the single thing most likely to change it, and what would have to be true for it to turn red. A green with no stated threat is not a green; it is an admission that no one has yet thought about what could go wrong.
None of this requires new tooling or a heavier process. It requires that the person reading the pack refuse to accept a bare colour, and ask, every time: which way, how sure, and what would break it? A programme office cannot manufacture that discipline on its own. It has to be wanted from the top of the table.
The real subject of a status report
Beneath the argument about colours is an argument about organisations. A reporting convention cannot be more honest than the culture it reports into. If the person who declares red is thanked, red will appear early, while it can still be acted on. If the person who declares red is remembered as the one whose workstream went wrong, red will appear late — in the emergency review, when the colour has finally become undeniable and it is far too late to matter.
This is why the three colours are dangerous in a way no sharper criterion can cure. They present a social artefact — the collective nerve of the people filling them in — with the calm face of an instrument reading. We trust the face and forget the artefact. The colour on the page is real; what it measures is not the health of the programme but the organisation’s appetite for bad news.
“Green is not the safe colour. It is the colour we have agreed not to look behind.”
So keep the traffic light, by all means. But treat it the way an experienced hand treats any single number offered up as reassurance: as the beginning of a question, never the end of one. The most dangerous three colours in programme management are dangerous precisely because they are so easy to accept — and the discipline that saves programmes is the refusal, month after month, to accept them.