But beyond just having the damage formula, we also need a way to get some useful data out of it.

The way I have been going about it is thus:

In Excel, using the given model and a given set of stats, I calculate a base potency.
Then I add 5 new rows of stats, where in each set 1 parameter is increased by 1. So WD+1, Int+1, Det+1, Crit+1, and Spd+1.
I calculate the potency for each new row using the same formula, then subtract the result of each new row from the original unaltered stats. This gives me the potency gain for increasing a given parameter by 1.
Lastly, I divide each of the results from the above procedure by the increase from Int, putting everything onto an Int scale.

Once everything is done, for BLM, I typically get something that looks like this:
WD = ~6
Int = 1
Det = ~0.27
Crit = ~0.19
Spd = ~0.20

These are the resultant stat weights, roughly, that I get from analyzing small controlled variations in the damage model.