It's that easy?
And is that the more definitive way or is listening the more definitive?
No, the definitive way is to measure equivalent input noise of every piece of gear at the intended gain setting, and measure headroom (defined as point below which 1% THD is experienced, although less distortion is always advisable), and operate within those boundaries. For example, if you find that your mic preamp has EIN of -125dBV at 40dB gain, but -110dBV at 10dB gain (contrary to popular belief, preamps typically have less EIN at higher gains), then it behooves you to use enough gain to overcome that. That is, of course, if your signal needs that much gain in the first place. If you signal/mic combination is such that the noise floor going into the preamp is say -100dBV, then it doesn't matter at all. And of course if the signal is so loud that your preamp is clipping even at 10dB gain, well then, turn it down or use a pad.
So you can go to those lengths if you want, but if you have reasonable quality gear designed by non-idiots, they will have done the heavy lifting for you already. Given that recording raw tracks generally results in a peak-to-RMS difference of say 20dB (could be more for drums), your gear should have about that much headroom above nominal operating level. If it doesn't, turn it down a bit.
Now, converters. Converters can be kinda guilty of only offering maybe 15dB of headroom above their nominal operating level. Not all, but some. So don't clip them. This might require operating at a lower than nominal level. Where? If we're talking 6dB, probably doesn't matter, but in theory you'd reduce right before the converter to minimize noise. Unless the rest of your gear isn't really up to nominal level either.
Once you are in the box, generally IF you have a 32-bit floating point system that doesn't truncate anywhere before the master output, you can blithely mix at whatever level you like, so long as you reduce (or increase) the master fader such that the output doesn't clip. That said, can you guarantee that? All major DAWs these days work properly, but plugs are still the Wild West, especially some of the obscure freebies. Some pass float data and don't truncate, some might truncate internally but pass data that looks like float, some might be intentionally designed to distort/saturate/whatever above 0dBFS. So it probably won't kill you to keep all points inside your box-chain below 0dBFS.
Then we could have an exciting thread about intersample overs and what illegal output does to certain DACs, but honestly I don't find the topic that interesting anymore . . .
