Three states carry a judgement well and have nowhere to put an absence of one.
Why it matters when the plan changes
RAG works because it is fast to produce, fast to read and comparable across unrelated workstreams. Those properties are why it has survived every attempt to replace it with something richer. Most criticism of RAG is really criticism of how it is governed, not of the scale itself. The strain shows most where reporting crosses ownership: Sull, Homkes and Sull found in Harvard Business Review, 2015, that only 9% of managers say they can rely on colleagues in other functions and units all the time, and a single colour often stands in for exactly that kind of partial view.
The tension is what happens to unknowns. With three states, a workstream nobody has evidence for is reported as a colour, and green is the path of least resistance. A reader cannot tell a considered green from an absent one, so the greens that were earned are worth less than they should be.
In practice
A monthly report shows fourteen workstreams, eleven green. Two of the eleven have had no substantive update in six weeks, because the person who produced them changed role. They are green because nothing has been reported against them, and they read identically to the nine that were genuinely checked.
Evidence
Cross-unit commitments are the least reliable part of execution, and they are also the parts least likely to have a single reporting owner.
Donald Sull, Rebecca Homkes and Charles Sull, Why Strategy Execution Unravels and What to Do About It, Harvard Business Review (2015)
What it cannot tell you
RAG status compresses many possible conditions into three states, all of which encode a judgement. It has no state for missing evidence, so an unassessed workstream and a genuinely healthy one can produce the same colour. The convention tells a reader what someone concluded; it cannot tell them whether anyone was in a position to conclude anything at all.
Questions
Conventionally: green means on track, amber means at risk or needing attention, and red means off track or requiring escalation. The exact thresholds are set locally, which is one reason the same colour can mean different things in two parts of one organisation.
No. It is fast to produce and read across unrelated workstreams, which is why it has outlasted richer alternatives. Cross-unit reporting is where it strains most: Harvard Business Review found in 2015 that only 9% of managers rely on colleagues in other functions all the time, and RAG rarely shows that gap.
It gives an unknown somewhere to go. Without it, a workstream with no evidence is reported as a colour, usually green, and reads identically to one that was genuinely assessed. With it, the reader can tell which greens were earned.
It adds one question: is there evidence for this, or not. That question takes seconds to answer and is the one most reports never ask. The cost is small and it falls on the person reporting rather than on the person reading.
Every colour should carry the evidence behind it, a named owner and a date for revisiting. Governance matters most in cross-unit reporting, the area Sull, Homkes and Sull (Harvard Business Review, 2015) found most fragile, since so few managers say they can rely on other units consistently. Without those attached, a colour is a mood, not a status.