Behavioral Targeting's Predictive Future
by Kaila Colbin , Tuesday, July 29, 2008  THE TOUR DE FRANCE ENDED this past weekend, and if you watched any of it on TV, you were overwhelmed with ads for high-end bicycle brands like Colnago and Pinarello. It makes a certain amount of sense that these companies would advertise in conjunction with the largest and most-watched cycling event in the world. Of the millions who follow the race, though, how many are actually in the market for a $10,000 bike? That disparity between the Tour ads and the Tour audience is representative of the issues with contextual targeting -- and it explains why the next frontier in personalization lies in behavioral. Contextual targeting puts things together that go well together. Behavioral targeting aims to put things together with the people who are likely to want them. And, yes, this is Search Insider, and, yes, behavioral targeting has a reach that goes beyond search, but here is what I want you to keep in the backs of your minds: the search industry rests on delivering relevant results and relevant ads. That's where semantic search is going. That's where personalized search is going. That's why behavioral is important in this context. And that's why the big news at OMMA Behavioral last week was ValueClick Media's launch of Precision Profile, which, according to the company, "uses a proprietary predictive algorithm to identify a marketer's best prospects in hundreds of consumer interest segments." Following their keynote, I had a chat with Matthew Boyd, ValueClick Media's senior vice president, and Joshua Koran, ValueClick's vice president of targeting and optimization. Koran contrasted his company's new product with the "traditional" behavioral targeting approaches of clusters and business rules. For the record, clusters group people together based on existing data ("I've seen you looking at auto ads, so you're an auto guy"), and business rules rely on if-then relationships ("If you've done three auto page views within five days, you're an auto guy"). Clusters and business rules rely on a certain amount of behavior in a segment before they become useful -- e.g., they don't recognize your interest in cars until you've already been shopping for a few days. That lag also prohibits those technologies from recognizing when a segment no longer applies, such as after you've bought your car. And, as Koran points out, coming up with business rules is difficult. You have to know the right frequency and the right recency, and the answer changes for each client: what's right for BMW isn't necessarily right for Jaguar. ValueClick Media's new system, on the other hand, uses a predictive algorithm to detach past behavior from future likelihood to click or convert. This forward-looking approach is self-learning and adapts on the fly to data from more than 13,000 publishers and thousands of advertisers. (For full disclosure, our company, VortexDNA, takes a conceptually similar -- but technically different -- approach, so obviously I'm inclined to support it!) BT faces some significant challenges, and privacy sits at the top of the list. ValueClick's platform doesn't contain any personally identifying information at all, which sounds great. Unfortunately, as the recent Senate hearings demonstrate, there has yet to be any consensus amongst citizens, politicians, or businesses about what is and isn't acceptable from a privacy perspective. The other big BT challenge -- and one that's potentially more significant from a commercialization perspective -- is avoiding the creepy factor. Everybody wants relevance, but nobody wants to feel like they're being spied on. ValueClick Media, like every other behavioral targeting company, will have to find the optimal balance between "personal enough to be useful" and "so personal it freaks me out." These concerns aside, the beta period for Precision Profiles generated promising results. Although Boyd declined to speculate on what kind of increase clients could expect, the press release contains some impressive anecdotes: "One online games client achieved a 298 percent lift in conversion rates over an optimized control group... [and] conversions for one mobile campaign outperformed other behavioral targeting vendors 11 to one." With numbers like that, it seems predicting the future is the future of behavioral targeting. | Kaila Colbin blogs for VortexDNA, whose technology can improve relevance for search engines, ecommerce sites, or any other recommendation service. |
|
No comments:
Post a Comment