Getting Targeting To Scale

Advertisers, to use the late philosopher Isaiah Berlin’s terminology, have often been forced to divide their time and efforts between being “hedgehogs” who focus on one big thing,  mass scale and reach, and “foxes,” who focus on many little micro-segments. But, as Eric Eller, senior director of product marketing at, explains below, one of the biggest emerging challenges of online advertising today is how to go beyond that dichotomy: how, in other words, to do precision targeting, especially behaviorally based targeting, that can seriously scale.

Behavioral Insider: Have the primary challenges to making targeting on the Internet live up to its potential been technological, or do they stem from other sources?

Eric Eller: The technology for doing more precise kinds of behavioral analysis and targeting has been around for longer than advertisers and publishers have really been ready to fully leverage it. Two issues have held it up: how to decide and prioritize what kind of behavioral targeting you want to do; and when and how to do it at a high enough level of focused scale to insure real impact.



BI: So it sounds like one of the greatest edges an ad network could provide is to provide that “focused scale.” Right?

Eller: The challenge is always scale, or more precisely, how to encompass as wide a scope as possible without stretching your advertising, your brand and your customer knowledge too thin. For instance I remember seven or eight years ago when the first wave of ad networks was getting under way. There was incredible width and breadth now available when you had 20 thousand, 30 thousand, 40 thousand or more sites to theoretically choose from. The problem, of course, was that the overhead of managing relationships with all those sites was not equal to the results you’d obtain. Smart networks gradually learned that the thing they needed to do was to selectively focus and reduce the number of sites, but at the same time to expand scale. Over the past year or so, something similar has started to happen with behavioral targeting. There are now so many individual sites that have deployed a variety of different approaches and methods, and many are yielding results on a small scale. The problem has been a lack of any seamless way of deploying targeting that can scale to ten million or twenty million or more impressions.

BI: There are, as you well know, a wide-ranging, confusing panoply of behavioral targeting technologies, methodologies and platforms. What has found to be the most practical ways to deploy for scale on your network?

 Eller: We’ve found in our practice there are three modes of behavioral targeting that lend themselves to scale. Retargeting, which we’ve found really beneficial especially to direct response, is one. [Then there’s] post-keyword search, where keyword based query patterns are used to target more appropriate follow-up display ads. And, most recently, audience based targeting, in which we cluster types of consumers together by identifying similarities in online search, web surfing and buying patterns.

BI: How does retargeting work?

Eller: Retargeting is based on the understanding that users who visit an advertiser’s Web site, but don’t stay long enough to fill out a registration form or make a purchase, are, or should be, great future prospects. Identifying these users for retargeting and then subsequently serving ads to them across other sites has a demonstrable effect on bringing them back. One test we did on retargeted ads versus non-targeted ads showed a nearly 200 percent  increase in click-through and 167 percent increase in conversion.

BI: What types of campaigns lend themselves to particular approaches?

Eller: We generally look to both retargeting and post-keyword search for more direct response campaigns. For more branding-oriented campaigns, [there’s] customer-based targeting, which allows advertisers to target by online behavioral patterns, grouped together by similarities to other groups of online visitors. For instance, we can identify people who’ve been intensely researching cars online over a certain time period. One test study we did involved a campaign for Volkswagen of America, which was launching a Jetta A4 model. We compared results obtained doing traditional demographic targeting of likely Volkswagen customers versus targeting of car buyers by previous online shopping behavior. The behaviorally targeted campaign scored 27 percent higher in information requests and a third higher in conversion rates.

BI: Looking ahead, you recently announced an all-video network. What will it entail, and how do you see targeting evolving on it?

 Eller: For our video network we’re aggregating pre-roll film from an initial group of 30 to 35 publishers which we’ll expand. The short-term goal is to begin tracking and targeting viewers by types of video content viewed, as well as deploying the other types of targeting we’ve found are scaleable on a network level. The thing we need to keep reminding ourselves is how early on in the process or learning curve behavioral is. The deeper the amount of data points and behavioral history we can aggregate and organize, the richer and more relevant the targeting.

Next story loading loading..