Building An Architecture For Cooperation

by Phil Leggiere, May 2, 2007, 12:30 PM

• Comment
• Recommend (1)
• Tweet
• 

Leverage in behavioral targeting, or in any form of targeting, is widely assumed to be best pursued on the basis of closely guarded proprietary user data. Sometimes, though, as Scott Case, lead product manager of Atlas Publisher, explains below (and as the experience of the direct mail catalog industry demonstrates), it pays for Macy's to tell Gimbel's. A key challenge of the next phase of targeting, he says, is to create the right architecture for data-sharing and cooperation.

Behavioral Insider: As an ad-server working with publishers, what are the biggest gaps you see in their current thinking about behavioral targeting and its role?

Scott Case: I think on the publisher side there are still misconceptions and points of confusion by many as to what behavioral targeting works for and what it does. You encounter unrealistic performance expectations frequently where a publisher is expecting a 1000% lift based on the notion that using behavioral data gives them a magic bullet.

Essentially the fact is that when the data you have available is limited to just anonymous publisher-side data about content categories, that's not likely to get you any big jumps, not by itself. It can be a way of extending premium inventory. If you're a Yahoo, to take an example, and you have a strong core of financial users, you can identify them and 'peel them off' when they go to e-mail, which is generally perceived as less valuable inventory. So that's a plus, but it's limited.

BI: What are some of the things you use to transcend those limits?

Case: Where you really begin to leverage behavioral data in a serious way is if you can relate sell-side and buy-side data. That's something we are offering on an opt-in basis. The first big category of this kind of data is cookie data where a consumer has shown interest in a product but has not purchased. You can use offers in those cases to push them over the edge. Another area is recent search history as it relates to relevancy of interests.

In addition, there's buy-side behavior which can be gathered offline, and then finally you can develop pooled buy-side data from more than one advertiser. This is something that's been effectively practiced for a long time in the catalog world. One rule of catalog marketing is, if you've bought from a catalog you're likely to buy something else from that same vendor. But it's also true that you're likely to buy from another catalog as well.

BI: What kind of research have you done on what works?

Case: To get significant lift from leveraging behavior means combining buy and sell side. But one of our core ideas is that behavioral targeting in itself is a silly idea. Our sister company, the ad network DrivePM, has done a lot of research, and what they've found is that you get a small lift from just using publisher surfing data, a bigger lift from combining that with buy-side cookies, but far and away the biggest lift from integrating these two with other kinds of data -- demographic, geographic and what we call technographic data, the technological systems and devices used and so forth. So there's an evolutionary scale, as it were, that's based on learning how to combine these disparate pieces.

BI: I know Atlas has been working on extending-ad serving capabilities beyond online into things like VOD. How does that relate to what you're talking about?

Case: What we're most focused on doing as a next step is using behavioral data across channels, specifically to relate online behavior to other digital channels such as on VOD cable. We have projects going on right now with Comcast and Charter which we expect to scale significantly in the next 6-12 months. The challenge there is architectural. How do we do cross-channel targeting to discreetly identify consumers across three digital screens including the Web, cable VOD and mobile. The next horizon is to achieve that kind of seamlessness, and to do that means again to create higher and higher levels of data sharing between buy- and sell-side and across channels.

BI: What are the major challenges you see looking ahead to that kind of data-sharing synergy?

Case: Actually the more we develop technologically, the more relevant the experience of good old-fashioned catalog marketing becomes. Because in a funny way, as advanced and different as online targeting seems, it's actually retracing the steps catalog marketers took 30 or 40 years ago. The question is how you create an architecture for cooperation and opt-in dating sharing. There's a reluctance at first to share any information that might be seen as proprietary -- but the fact is, much publisher data is relatively useless in a standalone context. And the same actually goes for advertiser data. The more widely data can be shared to deepen segmentation profiles, the bigger a lift behavioral advertisers will see across the board.

• Comment
• Recommend (1)
• Tweet
• 

Be the first to comment on "Building An Architecture For Cooperation"

Leave a Comment