• by Steve Smith , Staff Writer @popeyesm, March 16, 2007

In the car branding world, it is all about getting into the consideration set, that handful of models a car shopper actively compares before purchasing. Edmunds.com got famous years ago by posting model invoice pricing online, and so it had a lucrative place among dealerships at the bottom of the decision funnel. In recent years, the site piled on more content and comparison tools, aiming higher in the funnel toward more branding dollars. At the same time, Edmunds has buckets of behavioral data: how we shop for cars, what we look for on which visit to the site, and the model pages we visit, and which ones we stop visiting. President Jeremy Anwyl watches 12 million unique users pass through his 13-year-old auto information hub. He tells us how he wants to use behavioral data in different ways to shape campaigns, customize messaging and even show precisely how and where brand advertising impacts behavior.

Behavioral Insider: What technology do you use for BT?

Jeremy Anwyl: When you talk about BT in the car business it's along the lines of somebody tracking consumers as they go from site to site or express interest in a particular vehicle. But we're building a different view of BT starting from a different place. By its nature, Edmunds is behaviorally targeted.... It is behavioral marketing on steroids, but not in the sense you may think of it.

BI: If you aren't using a format BT ad serving system, then how is behavioral tracking a part of your mix?

Anwyl: A lot of consumers are using the Internet to decide what car to buy, not just what to pay for it. And that sets up a whole other marketing opportunity, which is to work with clients to help them get on that list of vehicles that people are considering. And we've created a whole metrics system of base-lining consideration and analyzing consumer behaviors as they go through the site, [which shows] how  targeted advertising can increase consideration. It is different, because in this context click-throughs don't mean much. People are performing a task on the site. Just by watching their behavior, you can see how advertising influences the task they are performing. But we see on Edmunds that the people most likely to click on an ad are those with an existing interest, so you may be optimizing toward existing interest instead of optimizing toward expanding interest.

BI: So how do you leverage that tracking data into new programs?

Anwyl: This year we are turning it completely around by saying you have to put effort in expanding interest first. That is a very different sort of message. Picture a map of the site, and we ran an ad on the home page. Before the ad ran, 1% of the people on the home page considered your vehicle. We run an ad for a week and we get it up to 1.5 %. This is a key point.

You are doing two things. One is, you are able to show an incremental behavioral change. Not too many kinds of advertising can demonstrate baseline and then incremental activity as a result of the ad. And then you can say, based on that data there were an additional 50,000 people that considered your vehicle that wouldn't have otherwise, and it cost X per person. You can do a fundamental ROI not based on costs per click, but on costs per incremental considerer. We have pages that have vehicle information pages, so if you have someone looking at pricing or colors or options, that is what we label as consideration.

BI: And you are tracking also across sessions?

Anwyl: Yes. Here is an opportunity we haven't used. People come to the site, say, four times when buying. The first time, they're building this consideration set; the second time, they are focused on pricing. Car companies probably would love the opportunity to send different messages to those consumers depending on the different visits. That second time they are closer to a decision, and then you need a call to action. We haven't gone up to that level of sophistication.  

BI: Are advertisers themselves to the point where they will customize for creative to match these discrete stages and visits?

Anwyl: It makes sense for everyone to do it, because the relative change in performance is significant  -- 2X is not unheard of. There is a great science to be built around this stuff, but most clients are not structured [to do it] in this environment of cost-cutting. It is mind boggling to me that people will pay $1 million on a campaign and [then] they don't want to spend $20,000 optimizing it. Why not keep optimizing where you get to the point where it starts to plateau?

BI: What are the key data points you collect now on users?

Anwyl: Let's go back to the behavioral model. Someone is coming to the site, and the first thing we look at is total traffic, then we look at what numbers of people consider individual models. Then we look at accepter and rejecter data because sometimes consideration doesn't create sales. Let's say a Toyota vehicle is not competitive. Inducing more people to consider that vehicle is not going to do much. That is an important message for a marketer to get. All you can really do in the short term is come up with a pricing adjustment to re-balance the equations. Then we measure configurations, which we find have a very strong correlation with actual sales. Our score is about .85. The relationship between increasing the number of people who configure a model and sales is very solid. The rest of the industry is obsessed with clicks, but they missed the point that there is a whole behavioral model here, and you can build a marketing system around it.