Larry,
Thank you for bringing this up. It is a common misperception.
VDP is the point of accountability separation in these participatory models. The number of contacts per VDP is almost entirely a function of the dealer's inputs (photos, seller's notes, video). Yes, a dealer needs to understand how well their VDPs convert into phone, email, chat, and walk-in contacts. In order to determine the break-even point for cost per VDP, it is also important to know the close ratio from these leads. However, those things can be dealt with in aggregate.
The discussion here is about determining the value of one listings service compared to another or the value of a basic service versus a premium package. For this it is best to hold constant as many variables as possible that are attributable to dealership activity. Professional research attempts to isolate the measurement to that which impacts the decision. If your pricing, merchandising, and lead handling are the same for all services, there will still be variances from VDP to sale due to statistical variation and small sample sizes.
Temporarily, my analysis is somewhat unfair to Cars.com because they offer chat and AutoTrader.com does not. This is a downstream variance caused by differences between the services. However, AT will soon have chat and we will be back to cost per VDP as the best metric for deciding which service or level of service provides the best value. This is the number dealers should hold their services accountable to. If a vendor is coming in to raise rates 20%, they better have raised VDPs 20% since the last agreement, fair and simple. However, if the inflated cost per VDP is still cheaper than the other listing services, then the decision is whether to pay the increase or buy none of the services at all. That decision is where your analysis comes in.