A system that can say it cannot see something is the only kind that can honestly say green.
Why it matters when the plan changes
Three-state reporting has no way to distinguish a considered green from an absence of information, so both arrive as green. The reader cannot tell which they are holding, and the one that is actually a gap in evidence is the one most likely to be acted on. A fourth state removes the ambiguity by making thin evidence reportable, which is the same principle behind scoring forecasts that carry an explicit statement of uncertainty rather than treating silence on confidence as confidence itself.
The tension is that blind is unpopular. It reads as a failure to deliver an answer, and there is always pressure to round it up to yellow. Holding the line costs credibility in the room and buys it everywhere else, because a rating scale that never admits ignorance is not being read as a measurement; it is being read as reassurance, and reassurance is not what a leadership team is asking for when it opens a readiness review.
In practice
A readiness review covers eight pillars. Six carry real evidence; two rest on a single document from before the reorganisation. Under three states those two would show green. Marked blind, they become the first two questions the leadership team asks, which is where they belonged.
Evidence
Forecasts that carry an explicit statement of uncertainty and are later scored improve, which requires a state for insufficient signal.
Superforecasting and the Good Judgment Project (2026)
What it cannot tell you
The four-state rating tells you how confident a judgement is, not why the arrangement is under strain or what should change first. Blind flags an absence of evidence but does not diagnose its cause, and the scale is silent on severity within a colour: two reds can carry very different urgency.
How Atlas reads it
The colours describe the arrangement around a person, never the person. Every rating carries a confidence, an owner and a review date, which makes it a forecast with a deadline rather than a status. Blind is the scientific state and it is preserved deliberately: an unknown is marked blind and never rounded to green.
Questions
A traffic light has three states and no way to report missing evidence, so an unknown is reported as a colour. The fourth state separates a judgement from an absence of one, similar to how the Good Judgment Project's 2026 forecasting work scores predictions only when uncertainty is stated explicitly.
That the arrangement needs adjustment to carry the demand, not that it has failed. Yellow is a weak signal with a proportionate response attached, and it is the state where the cheapest moves on the intervention ladder are still available and still likely to work.
No. The colours describe the work and the arrangement around a person: the role, the mandate, the support, the sequence, the interface. Nobody is ranked out, and the recommended move is almost always a change to that arrangement rather than to the person in it.
That the evidence exists, has been read against the current demand, and supports the judgement within the stated horizon. Where any of those is missing the rating is blind. The Good Judgment Project's research on forecasting, published in 2026, shows judgements scored well only when uncertainty is stated rather than guessed.
Yes. Every rating carries a review date, because it is a time-bounded judgement about a specific plan and arrangement. Any of the six triggers firing before that date brings the review forward, since the thing being rated has changed underneath it.