Great topic. I've frustrated myself to death over attribution modeling in automotive for years. There are too many online & offline unknown variables, and factors with car shoppers that effect a purchase.
You can't build a model which will know how to assign a value to a advertising source that makes absolute sense all the time. Although you can build a model in Google analytics that paints a picture of a path to purchase, or assigns shopping value, and use linear attribution model to understand your exposure strengths with online (& offline) advertising, but it's just a another guide, another tool.
You really don't have to get that deep with it though, there are probably less than 10 new & used online automotive sites (national & local) that are worth paying for, and most dealers can afford to be on all of those sites anyway. If I could prove what source brought them in to visit the dealership, then I would assign the highest value to that source, regardless if it's first or last. But did they contact me first, or just walk-in the door? See, too many unknown variables.
In traditional eCommerce online shopping websites, I can show a real conversion (checkout) that has a real $ value, I can see virtually everything about the customer once they enter the site and login, their entire path, and I know with very high probability how I enticed them to visit the website. It's a very simple system compared to automotive, and real estate, both are just different shopping journeys. Would a website that requires a login for shoppers work in automotive?
I think if dealers can focus on making their websites easy to search, view photos of the vehicles (great photos), list all options clearly, provide detailed pricing, and provide several ways to communicate for the shopper, then handle every customer interaction with professionalism... they are doing the majority of what's needed to lead their local markets. Once that's solid, now we can choose where to expose your inventory and dealership online.
