Why RAG status tells you a programme is failing too late

I have sat in a lot of steering meetings where a programme went from green to amber to red across three consecutive months, and everyone in the room was surprised each time. Nobody was lying. The reports were accurate. They were just describing the wrong thing.

RAG is a lagging indicator, and that is fine

A weekly RAG on budget, schedule, scope, and resources tells you what has already happened. Budget goes amber once the money is spent. Schedule goes red once the date is missed. Resources go amber once the people have left.

None of that is a criticism. Lagging indicators are how you confirm reality, and a programme without them is flying blind. The mistake is treating them as an early-warning system, because by construction they cannot be one. The signal arrives after the event it is signalling.

What was actually true, months earlier

Programmes rarely fail because a task slipped. In my experience they fail for a small number of structural reasons, and every one of them is observable long before it shows up in a RAG:

• The steering group has no real decision authority, so escalations sit for

weeks and the delivery team routes around them.

• The programme has a committee rather than an owner, so nothing that spans two

functions ever gets settled.

• The business case was written to secure funding and then filed away, so

nobody notices when the programme stops being able to deliver it.

• The change impact on the roles that have to absorb the new way of working was

never assessed, so adoption collapses two weeks after go-live.

• Architectural decisions have no written rationale, so each team change

re-litigates them.

Notice what these have in common. None of them is a number you can pull from a system. All of them are things the people on the programme already know. The information exists; there is just no instrument pointed at it.

Asking the question systematically

The instrument is not complicated. It is a structured question set, answered by a panel rather than by one author, scored the same way every month so the trend means something.

TransformRadar's Programme Health assessment covers six dimensions — governance, stakeholder engagement, scope and delivery, architecture and technology, change and adoption, and value realisation. Each is a handful of statements a respondent rates as yes, mostly, sometimes, or no. "The steering group has clear decision authority and uses it." "Benefits are tied to specific people who will be measured on them post-go-live." "The sponsor would still approve this programme today."

The scoring is deliberately dull: yes is 4, mostly 3, sometimes 2, no 1, averaged per dimension and rescaled to 0–100. Green at 75, amber at 50. There is no model in there doing anything clever, because the value is not in the arithmetic. It is in having asked at all, and in having asked the same way last month.

The disagreement is the finding

The part that surprises people is what a panel gives you that a single author cannot.

When the sponsor answers yes to "escalation paths are clear and decisions are made within days" and the delivery lead answers no, the average is a meaningless 2.5. The useful output is not the number — it is that two senior people have opposite views of how the programme is governed, and until this moment neither knew.

A dimension where the panel disagrees is worth more steering-meeting time than a dimension where it agrees and the score is low.

So the assessment flags wide spreads rather than averaging them away, and reports by role rather than by individual, with role breakdowns hidden until at least two people in that role have answered. People are candid when they cannot be individually identified, and the whole exercise depends on candour.

Reading it next to delivery

A strategic assessment on its own is a monthly opinion survey. It becomes a health score when you read it against what is actually being delivered and what is actually being realised — which is why the composite weights delivery highest at 45%, strategic at 35%, and benefits at 20%.

The pattern worth watching is not a low number. It is divergence: delivery still green while strategic health falls. That combination is what a programme looks like roughly a quarter before it goes red, and it is invisible if you only have one of the two.

What to do with a bad score

Nothing automatic, is the honest answer. The score is a conversation opener, not a verdict, and a governance tool that issues verdicts gets ignored by exactly the senior people it needs.

What it does is change what the steering meeting is about. Instead of forty minutes walking through a status deck everyone read on the way in, you spend the time on the two dimensions that moved and the one where the panel does not agree. That is a better meeting, and it is available from the first assessment.

More detail on the method is on the Programme Health page.

• Programme Health
• governance
• steering