One of my DJ buddies asked me how I normalize my needledrops so I thought I would put my answer here as well since it seems relevant to the discussion.
(BTW if a pro mastering engineer type (does Bob Katz read changstar?) happens to pass through and has comments or better ideas I would love to hear them)
Anyhow, conventional wisdom would simply have you normalize to 0 (or -0.03) dbFS and be done with it however you will find that with needledrops you are invariably going to end up with some big ass transients few of which will be much larger then anything in the recording. If you are playing your drops back to back in a playlist (or set) vs modern masters they are often several dB quieter because of this. In extreme cases they can be as much as 6dB RMS lower or even more.
To offset this without smashing the dynamic range what I do is gain everything up just past the point where the biggest transients clip. The goal is to get maybe 100-200 or so overs in each channel (in iZotope RX4 use alt-D to bring up the loudness info/over calculations window) which usually ends up being somewhere between +0.75 and +3dBfs on the louder of the two channels. If you end up in the thousands or a single clipped interval ends up being >100 samples or so you need to dial it back. It will probably take you a few tries to dial in a number you are comfortable with.
Once you are happy and have the waveform just barely clipping you use the "Declip" function in RX4 set at -0.1dBFS threshold to re-interpolate the overs down to 0dBFS. Remember you are interpolating a few hundred samples which at 96K sample rate is a few milliseconds total over the entire song. Nobody is going to hear that and you have gained typically 1-3dB RMS loudness in exchange.
Good deal IMO.
