Generally, you want a flat level across the frequencies on a spectra.
Everything peaking at some point at around the same level, with lots of dynamic movement across the spectra.
I use them to pick out any bias my ears may have missed, to check for naughty spikes, and good frequency distribution.
Your ears obvisouly come first, but they can be fooled.
With dance music you`ll tend to see a bit of a bulge at the low end probably around 80hz, as dance music is a little bottom heavy asa rule.