For the sake of illustration, say the path of the zipper crosses the chest and abdomen parts of the dress. You'll need [n] alternate geometries containing just the abdomen group, and another [n] containing just the chest group, where n is the number of unzipping stages: 2 in your case.

So the total geometry list would be something like this (names made up by me, of course):

Dress.obj (grouped in the normal way for conforming clothing)

abdomen_unzip_1.obj (contains only abdomen group)

abdomen_unzip_2.obj (ditto)

chest_unzip_1.obj (contains only chest group)

chest_unzip_2.obj (ditto)

Because you're going to slave the abdomen and chest geometry dials together, the only group boundaries you need to be particularly careful about are abdomen-hip, chest-rCollar and chest-lCollar, assuming the usual DAZ grouping. These boundaries need to be identical to the base mesh on all alternate geometries.

Before you go much further, I'd advise you make up additional versions of the dress using the geometries from the two stages of unzipping. These won't be part of the final product, as such, but it will enable you to see how well they conform. You see, the thing is that your alternate geometries have to use the same rigging as the base model (as far as I know), and this might limit how well the unzipped versions are going to conform. I.e. you may not be able to pose the figure as freely in an unzipped dress as you would in the zipped-up one. It's best to be aware of the limitations before you put in too much more work, I think.

I hope that all made some sort of sense.