Definition
The rank agreement between the confidence grade assigned to a claim and the realized fitness of that claim in the uses it fed, trended over time.
The Measurement Gap
There is no metric that tells you whether the grade distribution still corresponds to realized fitness for use in the consumers it feeds, which means the program optimizes toward a proxy after the correspondence has decayed.
Formula
GFC(t) = τ( g_c , φ_c ) over claims c with an observed fitness outcome in period t, where τ is Kendall rank correlation
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Whether the grade population can still be read as a statement about trustworthiness, or has become a statement about the grading configuration.