Though we're often creatures of habit, we know human behavior is about a lot more than habit. In its essence, it's about change and finding new ways of expressing needs and intentions. It
should follow then that "targeting " behavior should be creative and responsive to change as well. Ironically, however, behavioral targeting has been surprisingly slow to move from a static
to more dynamic model, as Rodney Webster, senior product manager of Mediaplex, the technology division of ValueClick, explains below.
Behavioral Insider: ValueClick and
Mediaplex have been very active on the behavioral targeting front over the past year or more with new additions and enhancements. How does the integrated dynamic targeting suite you just introduced
take that evolution forward into 2009?
Rodney Webster: Behavioral targeting from the start has been about getting the right message to market at exactly the
right time. That's been the real goal. But for most marketers in actual practice so far it's really only meant a small fraction of that value proposition. It's meant basically getting a
message to the right person. The other parts of the equation, the right message and the right time have been far less refined.
advertisement
advertisement
What dynamic behavioral targeting brings to the mix is
real-time data, which can come from other non-behavioral data sets, demographic, geographic or contextual. So depending on the exact characteristics you wish to target, you can overlay all the other
data sets. Beyond that you can use all of that real-time data to optimize creative. Rather than have at best a few pre-set creatives to work with and be constrained by that, you can start with a
simple creative template and then customize it based on evolving information you have. The key then is not so much to just target certain predefined behaviors, something that's become child's
play, but targeting particular creatives based on certain behaviors, and doing it on the fly.
BI: How are these creative elements optimized, tested and scaled?
Webster: This can become quite powerful when you're also able to integrate a full gamut of testing methodologies, whether it's AB or multivariate testing, in real-time into a
behavioral campaign. You can simultaneously run a variety of creative ads, for instance,, against a particular behavioral segment and very quickly discover which creatives are optimal without needing
to waste hundreds of thousands of impressions finding out whether something works.
A big limitation, a practical economic limitation, of the way behavioral targeting has been provided, is
that it's become enormously resource-intensive, both in terms of time and money. By making testing and constantly changing campaign elements an automated process, those constraints are much more
easily overcome.
BI: You're alluding to dynamic targeting as a self-learning system. How does the system learn and how is that learning directed towards marketing goals?
Webster: At the end of the day, of course, conversions are the crux of the matter for direct response advertisers. But increasingly the lines between traditional
branding and direct response are blurring. Advertisers are also seeing that behavioral targeting can also add insight into behaviors that advertisers weren't specifically targeting and may never
have understood were even relevant before. They can, for instance, discover, that while they were targeting people who were browsing content X, in actuality clicks and conversions were coming from
places they hadn't predicted they would. In the past those behaviors would have been missed, or viewed as aberrations, because marketers were singlemindedly tracking just what happened regarding
Behavior A.
The key point is now that behavioral targeting no longer needs to mean just Behavior A means someone belongs to Behavioral segment B and therefore needs to be served promotional C.
Conventional behavioral targeting, ironically, has been based on the assumption, implicit or explicit, that behaviors are a fixed target. But the truth, everyone realizes, is that behavior is a
complex, ever-moving target. What we need to be doing from here going forward is shifting and adapting our model to account for that.