Brian Fisher shares some approaches for when a vocal is too dynamic to sit right, even after running it through two or more compressors in series with moderate gain reduction on each.
One place to start is clip gain. Cutting the vocal into regions and adjusting their individual gain levels can even out the signal before it hits the compressor, giving it a more consistent input to work with rather than constantly chasing extremes.
If heavy gain reduction is still needed - 10-15 dB or more - that level of compression can start to sound aggressive and upfront in ways that don't always fit the song. Running that compressor in parallel is one way to manage this: blending the heavily compressed signal back with a mix knob at around 50-60% can preserve much of the dynamic control while taking the edge off the character. Fisher demonstrates this using the UA 1176's built-in mix knob.
Automation after compression is another layer worth considering. It can help restore natural phrasing, bring down quiet passages that still feel uneven, and tidy up word endings in ways compression alone tends to miss.