I think you also have make sure it's apples to apples in what you are calling cost. Since many have bundles with other products that don't generate VDPs.
Part of my cost per VDP theory was that I would be able to add on/upgrade packages without seeing an increase in cost per VDP. If they wanted me to add cost, my return should be an increase in VDPs, therefore keeping the cost per VDP at the same level or lower. This played out for me in a couple of instances where increasing packages actually lowered my cost per VDP BUT increased VDPs per unit without a measurable increase in Leads. I don't know of many products options today that wouldn't be tied to a VDP or CTA that would stem from a VDP.
For classified value I totally see your point here @Dan Sayer how an overall VDPs increase that's unequally distributed to likely 20% of the most appealing VINs doesn't provide the same value as if they were spread evenly or even better yet to cars with less traffic. I really think this is a good way to assess value and not go past this point. This analysis is actually still pretty hard. I am going to think some more how I would set this up