Right, where you started was after tracking. When you track, you gotta pay attention to the peak meter, because you do NOT want to exceed 0dBFS at your converter. That will cause hard clipping, which sounds really bad. Hence, the -12dBFS peak advice.
Nobody wants to release a finished mix with peaks at -12dBFS though. So you need to bring that volume up. How? Well, let's say you have two vocal tracks, one is a singer, and the other one is a rapper who
spits hot fire 
Probably the fire-spitter had a more dynamic performance than the singer. If you were a good tracking engineer, you would have maybe had to use two different preamp gain settings to record the two, and both thus ended up peaking at -12dBFS.
So you singer is much louder than your rapper at the moment, because the singer's dynamic range, or what's called "crest factor" (difference between peak level and RMS level) is smaller than the rapper.
You can't just mix them equally, because that will leave the singer louder. If you just turn up the rapper so he is 0dBFS peak and leave the singer at -12dBFS peak, that might be a little better, but the rapper is still probably way too dynamic to sit well against the def ill beat the producah dropped.
What to do? Compress the rapper until he's at the same crest factor as the singer. Now you can adjust volume (after the compressor) to taste, still with plenty of headroom.
I would recommend starting out with a very easy compressor to use. One of the LA2A emulations would be ideal; they only have two knobs and a switch, and they are pretty well set for vocals. I imagine there are some freeware VSTs out there for the taking . . .