What problem does it solve? Account health reviews often rely on gut feel or a single blended score that hides why an account is at risk. This Skill turns raw product usage and support signals into two ranked watch lists where every row carries named, citable evidence, so owners can see exactly which accounts need attention and why. ## Core Features & Use Cases - Dual ranked watch lists: Builds separate usage-health and support-signal rankings over a stated window, then joins them to surface accounts showing both declining usage and rising support friction. - Evidence-backed ranking: Every ranked row includes a stable account id, a one-line reason, and 2-4 cited evidence items (metric, window, value or delta, source); rows that cannot meet that bar drop to an unranked watch-closely note. - Run-over-run diffing: Compares against the prior watch lists to report new entries, exits, and rank moves of three or more places with the signal that caused each move. - Use Case: A customer success lead preparing a weekly review asks which accounts are quietly declining; the Skill reads usage and support data for the trailing 14 days and returns two dated, ranked lists plus the single account that most needs a decision. ## Quick Start Ask the assistant to run an account health review over the last 14 days using the current account roster, usage metrics, and support ticket data, and produce the two ranked watch lists with evidence per row.
