We build a measuring instrument, not a scoreboard.

We report your position as a band, and we show the interval around every per-cell estimate rather than only its midpoint. When you have answered a great deal in a domain, the interval is narrow and you should act on it. When you have answered little, the interval is wide and the honest reading is that we do not know yet — which is itself the most useful thing the report can tell you, because it says exactly where the next hour should go.

The corollary is that a band can stay put through a good session. That is not the report being unresponsive; it is the report declining to move on evidence too thin to move it.

Alongside the overall band we report a Safety-True band, weighted toward items where a wrong answer would put a client at risk — prioritisation under deterioration, recognising a cue that requires escalation, an intervention that is contraindicated rather than merely suboptimal.

These two bands are allowed to disagree, and when they do, the disagreement is the finding. A candidate who is strong overall but weak on safety-critical judgement has a specific, fixable problem, and a single blended number would have averaged it out of existence.

Every miss is classified by the kind of mistake it was. A missed priority cue, a correct action taken in the wrong order, a genuine gap in knowledge, and an item read too quickly are four different problems. Re-reading the content only fixes one of them.

Pace is reported next to accuracy for the same reason: in a raw percentage, rushing and not knowing look identical, and they call for opposite responses.