-15 dBFS around 500 Hz or wherever the belly of the vocal is. Then EQ and compress away from that. What I do is use a brickwall limiter at +5 in front of the analyzer to simulate the mastered level at all the important points like kick -10, snare -10, -15 dBFS for instruments -20 or less, vocal -15 dBFS, etc. Then I turn off the peak limiter and work with this initial industry standard balance.