• by John Nardone , Op-Ed Contributor, April 15, 2008

If you are like many other marketers and media planners, you're probably struggling to make behavioral targeting work. Behavioral targeting confuses people. The term encompasses a wide range of tactics and methodologies and is further obfuscated by new media networks trying to differentiate themselves by claiming the latest and greatest approach.

As a result, clients and planners are advancing carefully, doing trial buys with many networks in order to learn what will work, what will not, and why.

To me, behavioral targeting refers to any targetable audience that is defined by a common Web behavior, and comes most often in one of three flavors.

First is remarketing, which refers to the targeting of cookies who have visited the client's own Web site. Remarketing can involve very sophisticated matching of offer and creative against specific on-site behaviors. For example, a consumer who abandons a shopping cart or a sign-up page is targeted with online ads containing sweetened offers over the next several weeks. In the best case, the offers reflect knowledge gleaned from their entire interaction on the Web site. Obviously, remarketing can be very effective, since it targets consumers who have demonstrated interest in the client's product. However, remarketing efforts often suffer from lack of scale because the volume of visitors to client sites tends to be very small.

Next, behavioral segmentation refers to a media seller's packaging of groups of cookies who have visited sites (other than the client's own) that suggest a need or receptivity for the client's category of product. For example, all cookies that have visited automotive-related sites over the last 60 days may be packaged as a segment of people likely to be in the market for a new car. Media sellers invest in building this package for one reason: to sell more of their less desirable inventory at higher CRMs. But currently, there are no standards to define behavioral segments, so every network has its own.

Lack of standards is not necessarily a bad thing insofar as it leaves media companies free to innovate.
However, this does put a burden on the media planner. To be successful, the planner either has to invest a lot of time to investigate and keep current on each network's specific methodology or (as is usually the case) rely on trial and error to find what works for each client. But thoughtful media pros are rarely satisfied with simply knowing that a tactic worked or did not. They want to know why. And here is where the behavioral networks often fall short: They provide little analysis (or even reporting) of the composition of delivered audience. Why? Because sellers put their analytic efforts into defining and packaging behavior segments to be sold; their systems are not set up to analyze a specific client's results. From the advertiser's point of view, they are a black box.

Last is predictive targeting, which is often lumped together with BT, but is actually quite different.

Predictive targeting is client-aligned, rather than publisher-aligned or seller-aligned. It is used by the advertiser for the advertiser to achieve their goals at the lowest CPM possible, not to support higher sell-side CPMs.

Predictive targeting uses more than behavior data and is transparent to the client. It includes objectively defined user data which, when combined with behavior, provides many more targeting dimensions for the advertiser to take advantage of. Importantly, clients have complete visibility to all the analysis and targeting rules, as well as audience composition related to impressions, clicks and response. There is no black box.

Predictive targeting is forward looking. It uses mathematical algorithms to predict what a user will most likely interact with, and then learns and is adjusted based on the actual results. By contrast, behavioral targeting is backward looking. It is based on cataloging what a user has done in the past and leaves the media planner to evaluate whether or not past behavior will be a good indicator of future performance.

Experienced marketers will incorporate all three approaches in their plans. They will use remarketing to take advantage of consumers who have "raised their hand" at the client's site. They will use predictive targeting to understand who those consumers are, and then max out the amount of that audience and response the predictive vendor can cost-effectively deliver. Finally, they will fill out their plans with behavioral networks who prove themselves by cost-effectively delivering client goals.

John Nardone is CEO of [X+1]. (jnardone@xplusone.com)