I'm using the master buss (since I'm in reaper) mostly for glue and tone shaping. Like Niki, exactly what goes on is genre (and song) dependent, though it'll always look similar.
The latest thing I've done* has:
- Waves NLS Buss (Nevo, med drive)
- Waves API 2500 (2:1, 10ms attack, .05s release, med knee, med thrust, feed back, 0db threshhold, HP filter in, ~2dB reduction)
- SPL Passeq (1.3K, 13K boosts)
- bx_2098 (3.25K shelf)
- Slate Virtual Tape Machine (1/2" FG456, med drive)
- Abbey Road Vinyl (saturation only, no click, wow, flutter etc)
- Ozone (stereo width, safety limiter)
Everything there except the EQs and the safety limiter (just there to avoid hitting 0dBFS) is to add at least one of sheen (subtle EQ changes + saturation) or glue. Each acts on a different part of the sound, which is why they are stacked. The idea is that it's stuff that I like the sound of, and is heavy on the hardware emulation.
Once I have a balance mix (with no effects) I'll set up my master (with everything except EQs and stero width) and move things closer to the sound in my head. Then I'll do the same to the tracks with something like Slate Virtual Tape and/or Slate Virtual Channel/NLS Channel/bx_console to get a base-line (in this case I used Virtual Tape -> Virtual Channel). Next step is to go bottom up with most important (or if there's something I need to bed in, most irritating!) track first.
Why do I set it up first? It's mostly stuff that is hard to do by effecting individual tracks and I want to mix into that so I'm getting straight to the sound I want. I don't put EQs on until late because I want to do that work on my tracks and busses first. In this case I found the whole mix a little dull but balanced well enough. The SPL got me close and I like the tone, but I couldn't push it enough without it sucking, so stacked the bx_2098 on it (and I preferred the sound of each doing a bit to getting the bx_2098 to do all the work). It's all on all the time so I hear what I'm printing and print what I'm hearing.
Sometimes I'll use the EQs to find a problem, but then fix it on the specific track/buss. Something like, 'there's a bit of honk here. Hmm, cutting 700 clears it but kills the piano, time to track down the culprit so I don't lose that piano.'
Why so much stuff? If I get it right this doesn't leave much for the mastering engineer to do, but it does mean when I'm done the band is hearing pretty much what they are going to hear once it is mastered. In the track linked I'm hit -14LUFS only touching the limiter at -0.8 couple of times _one on channel_, but I'm not loosing a lot of peak from my master chain so it's coming in pretty hot from the mix itself. If the band has budget, I'd still want an ME to go over it. I could squeeze out the last 5dB for an album version, but they could do that better; for a full length they'll spot any subtle shifts across the tracks and sit make everything sit together; I may be tonally together, but need some tweaks to fit into commercial radio or fix some problems.
Another thing worth noting is I've started mixing hot. Originally this was so I had plenty of level into my monitors, but now I've got it so that I can print a track I'm happy with without having to normalise it. I run Waves VU at -15dB=0VU and bx_meter at Katz-14 on my monitor buss (a Reaper thing - plugins after the Master buss, where you can put monitoring and room correction pluigns) and shoot for 0VU/0K. I'm doing this from the first balance mix, right through. I was doing -18dB=0VU but then I'd have all this headroom I didn't use. After watching the Richard Furch tutorial where he mixes into -15dB, I kinda went ¯\_(ツ)_/¯, I guess I can just drive it hard.