All 10 rubric factors, org-wide average, worst-first within each group — the enablement priority list.

Session type × the 3 weakest factors org-wide — reveals whether a gap is org-wide or type-specific.

What Claude's per-session recommendation most often points at, aggregated org-wide. #1 = next training topic.

Session volume and avg score per team, last 30 days. ("By Surface" was dropped — every session today is the same surface, so it carried no signal.)

Users repeatedly (≥3 of their last 10 sessions) triggering the same recommendation category — all-time, a real outreach list.

Auto-generated from threshold rules (avg below 0.5, n≥10) — no model call, fully explainable.