How much of the mix's own measured shortfall the evidence licensed the fitter to attempt — not how well the cascade hit its target (it hits it within ~1 dB regardless). A ratio of the priced target to the raw deficit, summed over the band where there's a real deficit to speak of.
Σ priced target ÷ Σ max(−before_db, 0)Conditional precision of temporal peak–quiet contrast, weighted by requested boost. Quiet frames must represent additive noise; tracking must not be leakage. Neither assumption is proved by this score. Mastering causality remains unavailable. Legacy records retain their original score and are labelled in their notes.
weighted max(1 − z SE / contrast, 0)Share of the cascade's peak gain sitting below filter_floor_hz — a descriptive
level-invariance boundary. This legacy measure does not separate mastering from source
coloration; plateau-relative correction remains a shaping preference throughout.