Five questions ask what the market is. The sixth asks what you actually did.

The shape of the set

Five reads and one record, and the record is what turns reads into evidence.

Read in order, the first five questions narrow from the general to the immediate: what asset, how volatile it is structurally, what coefficient the lookup assigns, what the live volatility evidence is saying now, and whether the intelligence read should differ from the static one. Each is answered by a computation the operator consults. The sixth question breaks the pattern. It asks what was actually used, and the panel cannot answer it — the operator has to write it down. That asymmetry is easy to read as an administrative afterthought and it is the opposite: without the sixth answer, the first five are a service that is never graded.

Why the log is the load-bearing one

A recommendation nobody records is a recommendation nobody can grade.

Suppose the panel suggests a wider coefficient on a given pair and the operator takes something narrower, for reasons that were sound at the time. Six weeks later the pair's stop-out rate looks poor. Without the log, the review has two candidate explanations and no way to separate them — the panel's guidance may have been wrong, or it may have been right and ignored. With the log, that is a lookup. The audit column is what converts the panel from a source of opinion into a component with a measurable hit rate, and a component with a measurable hit rate can be improved, retired, or trusted more. One without cannot.

The freshness stamp

Each of the five reads carries the window it was computed over, and staleness is a real failure.

Every answer the panel serves is stamped with its data window, which sounds like housekeeping and is a defence against a specific error. A volatility classification computed against a baseline that has not been refreshed describes a market state that may have ended, and it will present with exactly the same confidence as a current one. The stamp lets a consumer judge staleness at a glance and lets the weekly review verify that no downstream decision quoted an answer older than doctrine permits. A volatility answer without a timestamp is not data; it is something someone remembers.

• Structural bucket and live read refresh on different cadences — check both stamps.
• A stale read is not a wrong read, which is what makes it dangerous.
• The review's job is to confirm no decision quoted an expired answer, not to re-derive it.

The order is not arbitrary

Structural before live, because live evidence has no meaning without a reference.

The sequence matters because each question conditions the next. The structural classification establishes what normal looks like for this instrument, and a live reading only becomes informative once that reference exists — an ATR value that is extreme for one pair is unremarkable for another, and no amount of staring at the live number reveals which case you are in. Asking the live question first produces a figure the operator then rationalises against an intuition of the instrument's usual behaviour, which is the eyeballing the panel exists to replace, arriving one step later and looking more official.

When the fifth answer is yes

Disagreement between the static assumption and the live read is the panel's most useful output.

The fifth question asks whether the intelligence coefficient should differ from the static one, and a yes is worth more than either number alone. It says the instrument is currently behaving unlike itself — the structural classification, which is built from quarters of history, and the live evidence, which is built from now, do not agree. That is a specific, actionable state: it argues for wider management or for standing aside, it is temporary by nature, and it should be expected to resolve. Recording it in the log with the coefficient chosen is what later allows the review to ask whether acting on the disagreement was worth doing.

The key idea

A panel that is only ever consulted becomes an opinion. A panel that is logged becomes a component.

The difference between those two states is one column in a journal. Consulted-only, the panel accumulates a reputation based on the readings the operator happens to remember, which will be the dramatic ones. Logged, it accumulates a record, and the record can say that its guidance beat the operator's deviations on this pair and lost on that one. That is the only route by which a volatility tool gets better rather than merely older, and it runs entirely through the question the tool cannot answer itself.