Masking is decided where speech lives, roughly 200 Hz–8 kHz and weighted toward the 1–4 kHz consonants. Every bed is measured two ways the moment it lands: spectrum (an Even masker fills that band; a Scooped bed has a mid hole a voice can surface through) and steadiness (a bed that plays with quiet gaps lets a line peek through its pauses, so it grades one step worse).
The level suggestion already accounts for the grade, and the live coverage scope in the mixer gives you a live check while the program plays.