Great to hear you've got a successful body sim!
There's no hard and fast rules with this stuff.
You could try a two pass process. Here's a rough outline to get you started.
First pass:
Facial blendshapes are a ziva bone.
You need a mesh with depth to make into a tissue. This would be the equivalent of the fat on the body.
You can project the skin down onto the skull and some deep facial muscles (temporalis, masseter etc), and merge that with the outer layer to get your 'facial fat'.
Now you can attach the facial fat to the ziva bone (animated blendshape target performance that you imported earlier)
Duplicate the projected mesh you made earlier, and wrap it to the tissue.
This inner layer is the layer you export for the second pass.
Second Pass.
Import your baked inner layer from the previous pass, convert it to a ziva bone.
Import your facial fat mesh, convert it to a tissue.
Attach your facial tissue to the bone.
This should give you a better distribution and perhaps some secondary effects, plus some nice volume preservation.
Other notes.
Make your initial poses with the eylids and mouth a bit open. This is to minimize the amount of tet-welding between the upper and lower eyelid/lips respectively.
You'll need to animate the mandible as well to get the effect of having the inside of the facial tissue to collide with it correctly.
I'd just encourage you to experiment and find out what works best for you, as simulation for faces is pretty cutting edge 🙂