the kind of normalizing that you want goes like this...
the program looks at the WHOLE (I do this before tracking) file and determines the loudest point, let's say -8dB.
then the program raises the WHOLE file by 8dB so that the loudest point of the file is at 0dB, or whatever level that you specify, i use -.5dB. What this is specifically called and where to find it in SF, i do not know.
there are other algorithms that look at perceived loudness and other nonsense, ignore those.