I first ensure my mix is what I want it to sound like, I use  a lot of parallel techniques on my subs 
I then add/enable on my Master bus the Lurssen Mastering Console (adjust as needed)
I also use Sonarwork's headphone reference plugin for when I'm using headphones to mix
Last I use IK Multimedia's metering plugin to look at levels and frequencies etc.....
 
Referencing bitflipper's post I find it efficient and easy to correct issues that the mastering process can expose. Although I know many like the separate mastering process I like hearing the end product without having to rerender my mix.


Good luck

I too have a Mastering Chain in my project which I always mix into.
 
As Dave (Bitflipper) says, it's much easier to revisit track related problems if everything is in the same project.
 
My chain is made up of several T-Racks modules and the excellent Metering plugin from Bluecat
 
When I want to export for album compilation, I export from my Pre-Master buss so I don't get the mastering Fx, but the chain is duplicated in the album template.
 
Any tweaks to individual modules are saved as a preset and loaded up as necessary.

I use Ozone 7 as well.  The first step I do is pick a preset that you think would closest fit the sound you're going for, and if you like it, start tweaking until it sounds like you want.   If the first preset doesn't do it for you, try another.
I don't mix down the song to a stereo mix, I always use Ozone on the project with all the tracks still intact because you might find you need to adjust levels on a specific track after Ozone is applied.  I know you should have the perfect mix BEFORE you start mastering, but I'm not a professional and sometimes I don't get it right the first time.

I have made a list of customized presets in Ozone, and usually start with them.  The Mastering purists think Ozone 7 is "cheating".  I don't care, I think it's the fastest and easiest way to master your songs (except that it is a bit of a memory hog, so I only apply it after all the tracking and mixing is done)

I work inside the project as well.
I often bounce 2-3 slightly different stereo mixes of the project, then mute all original tracks (archive the ones with heavy VSTs) and go on.
 
Say, there's a hi-hat or something that I'm not quite sure about volume-wise. I make two mixes, one with louder hi-hat, one with a quieter one. Then, when I've done my "mastering" of both mixes I compare how the versions sound in other reproduction systems.

I'm with Bit, I tend to master and finalize mix at the same time...depending on processing power...those linear phase EQs can be quite taxing :).

My mastering signal chain runs like this, first Waves NLS console emulation (sorry Cake, brought this plug a few weeks ago, and it really is the bomb), next compression, I tend to use for rock/pop and really full mixes Waves HComp, for less full mixes Waves VComp, and for acoustic sort of stuff Cakewalks C2, next I use Cakewalks linear phase EQ set to best quality, I use this EQ to really clean a final mix up, then I use Waves Master Tape to get the final levels right. Really what I'm doing is emulating a typical analogue mix master session, the master tape is then fed into Waves RS56 EQ for final tonal adjustment and finally I use Cakewalk's Adaptive Limiter, which is the bomb by the way. Of course of the mix needs less processes...so be it.

Ben