I personally never use a multi-band compressor on my master buss, because if I'm personally mixing the track, if there are places where I could see using a multi-band compressor, I'll use it on the track that needs it, not the master buss.
When mixing, I rarely use one, but it usually ends up being on things like acoustic guitar and vocals most often, then maybe sometimes on something like overheads and maybe electric guitars. I don't tend to use it on much else, now that I think about it.
If you're wondering when to use it, you have to realize what the multi-band compressor is doing. An easy way to get an idea of this is to manually create a multi-band compressor in a throwaway session.
Basically what you can do is make a new session with just a vocal or acoustic guitar or something. Then, duplicate the track three times (for this example, it's going to be simulating a three-band multi-band compressor).
On one track, setup an EQ with a low pass filter set to say, 200Hz with a steep slope. When I did this in my class a few years ago, we just used the Digi III compressor that comes stock with Pro Tools, which maxes out at 24 dB/oct, but I think steeper slopes for this are supposed to be better to keep phase issues down (someone else correct me if I'm wrong!).
Set a second EQ on one of the other tracks to have a high-pass set to that 200Hz range, and then a low pass set to around 5kHz.
Lastly, set an EQ on the last track with a high-pass set to 5kHz.
You know have your guitar or vocal split into three bands. Now add a compressor to each track, with the threshold all the way up for now so they aren't compressing anything.
Now solo each individual section. You have a low, mid, and high band. If you're using an acoustic guitar, more often than not, my biggest issues have been with poorly recorded acoustics having issues with being boomy in the low end when playing lower strings, but then not being boomy when playing higher. If it's a vocal, the most common way you'll notice a multi-band being used is a de-esser, so you'll probably notice some dynamic issues in the top frequencies above 5kHz!
So now that you've noted these issues, you can apply a compressor to just the low frequencies to even out the issues with the low end on the acoustic guitar. This makes it so that you don't have essentially zero low end when low strings aren't played, but it doesn't get out of hand when the lower ones are played. Conversely, if this is applied to a vocal with harsh esses, I'm assuming you understand how a de-esser works, and this will level out the high frequencies so that those frequencies aren't an issue. If you were to apply static EQ to this vocal, the vocal might sound dull by pulling out too much top end to soften the esses.
This is of course just an exercise to help you more clearly understand how a multi-band works. The frequency ranges were essentially arbitrary in this example. They can be moved around as needed (for example, say it's just the sub frequencies you're having issues with, then you'd move the low pass on the low band down to maybe 80Hz, and then consequently move the high pass on the mid band down to 80Hz as well).
So now that that big bunch of garbage is out of the way, assuming you understand what the multi-band is doing, you're basically just needing to ask yourself if any track/mix has dynamic issues only in a specific frequency band. If so, utilize a multi-band compressor to take care of the issue. If not, then you don't need one.
TL;DR: multi-band compressors are used to fix dynamic issues that are only occurring in a specific frequency range. If you aren't having these issues, then don't use a multi-band compressor.
