A New Breed of Behavioral Targeting

• Tweet
• Buffer

Two products related to behavioral targeting recently appeared on my radar. One just launched this month; the other has been around since 2004 in stealth mode. It's now being formally introduced to the market.

U.S.-based online marketing company ValueClick Media will launch a platform this summer offering advertisers behavioral targeting based on predictive analytics.

My direct-response antennae were raised by this news, as I spent years working on predictive modeling strategies for circulation plans with traditional catalog clients. I wondered if that concept had morphed into the online world and, if so, how.

I called Tony Winders, ValueClick Media's VP of marketing, to get a better understanding of the new offer. Though the product hasn't officially launched yet, Winders was able to give me a bit more detail about the beta. Our conversation also touched on privacy issues, the consumer's lack of understanding of cookies, the need for guidelines, and so on.

Until now, behavioral targeting methodologies have fallen into one of three camps: retargeting, clusters, or custom business rules. Retargeting is the most effective form of behavioral targeting because it targets individuals who have already interacted with a brand. Clusters assign each visitor to only one group, while custom business rules provide marketers with a high degree of control. Both, however, rely on humans to interpret data to determine how to classify a visitor. ValueClick Media hopes to popularize a new, predictive approach to behavioral targeting by allowing its technology to automatically determine what category each visitor belongs to.

ValueClick Media's algorithm takes into consideration visitors' observed behavior and creates predictive models for future behavior, all based on attributes provided by anonymous cookie data, not personally identifiable information. The resulting profile data define each visitor as belonging to one or more categories (mobile, finance, retail/shopper, travel/air, etc.). ValueClick can then leverage its extensive inventory (over 130 million unique visitors per month) to give advertisers greater potential per category than a smaller ad network could supply.

Another product, dubbed aCerno, has been in stealth mode for nearly four years. It collects anonymous information from an association of over 375 major multichannel retailers' Web sites (that aren't identified to one another), representing 140 million shoppers. The information is completely private and tagged only with an ID, with no cross-reference to personal information. ACerno compiles this anonymous data using cookies. The concept borrows heavily from the blind cooperative databases catalogers have been using for over a decade. In fact, aCerno is a wholly owned subsidiary of I-Behavior, one of the major co-ops.

ACerno's analytics provide two exclusive tracks of predictive information:

• Who customers are. Knowing what people shop for correlates strongly to who they are. For example, someone who buys a dress and a crib is almost certainly a woman with a baby.
  
  
• What customers will buy next. Large populations with similar purchase behavior patterns can be discovered and sold.

ACerno clients' best prospects are identified with modeling and profiling techniques, finding users who look most similar to their best customers. Once these high-value prospects are recognized, aCerno uses its massive advertising network to deliver targeted advertising messages directly to them, creating brand consideration or incremental transactions, which can be purchased on a CPA (define) basis

The company's extensive network of high-quality Web sites, publishers, and portals is targeted exclusively at the cookie level with banner ads and rich media to achieve maximum reach within the target audience. The network reaches over 95 percent of the Internet population with more than 80 percent of the impressions served into sites on comScore's Top 500.

This is predictive analysis; scoring million of cookies against hundreds of variables to create models. The concept was so successful to catalog mailers that it accounted for the vast majority of their prospecting efforts and budgets, ultimately replacing the use of individually rented mailing lists. Going forward, the same may hold true for retailers that can use custom models of their customers. These can change per promo or over time, and by using the profiles as selection criteria for branding efforts or to drive traffic to a store.

Both companies are extremely sensitive to the privacy concerns that may surface regarding their products. They've gone to great lengths to issue very thorough privacy statements and to educate both the consumer about cookies and the publishers/site owners about modeling with cookies. Ultimately, the success of either product will be a combination of results and the willingness of consumers to allow the use of cookies for better-targeted offers.

Both the New York and Connecticut state assemblies have been wrestling with bills that define online privacy and what information can be gathered about an individual. AOL, meanwhile, has come out with a campaign featuring a penguin to educate its visitors about cookies