• by Phil Leggiere , November 21, 2007

In these still early days of online video, targeting remains more trial and error than art or science. Two major schools, however, have already emerged, one concerned with contextually serving ads based on deep meta-tagging of editorial contents; another, on leveraging the demographic profile information available in social network registration data. In the conversation below, Keith Richman, CEO of video entertainment site Break.com, suggests that the next phase of video targeting will focus on the learning behaviors through which consumers use tools such as recommendation engines to match their personal interests and content choices over time.

Behavioral Insider: What has been Break.com's motivation for using behavioral data on the site?

Keith Richman: Our goal is to be the entertainment destination for guys in the 15 to 34 demographic.  We've built a platform that delivers video content that suits our audience's interests.  So for us, demographic targeting is a given. 

 From that starting point of knowing exactly who our audience is,  taking it to the next level entails learning ever newer and better ways of helping that audience find the specific content they're interested in.  We show 400 million clips a month to approximately 18 million users.  So the challenge is to learn about their interests in order to transform or personalize the demographic profiles we already have into far more granular psychographic segments.

BI: How is data compiled?

Richman: People who use our site often visit specific categories to view videos, say sports or horror films or what have you.  And, of course, whenever a video is uploaded it's given a description by the users.   But those ultimately are fairly skimpy bits of information in order to truly understand what that user is looking for. To get a deeper, more personalized interest profile, we need to go beyond those superficial descriptions of information and capture more dimensions.

One thing we do is look at the duration of time spent on particular categories of videos and look at viewer ratings.  A user may, for example, describe himself as liking to watch both pranks and martial arts movies. But if we find they're spending more time on martial arts and giving a thumbs up rating on particular types of films within that genre, it gives us a good basis on which to recommend films.  If someone says they like sports or car videos but ends up watching and rating martial arts, to use that example again, their behavior is actually a better key to personalizing our video recommendations than what the consumer himself says.

BI: How are recommendations derived and tracked?

Richman: The recommendation engine is based on users' stated interests as well as their demonstrated interests as shown in the behaviors described above. It also analyzes and predicts what videos a user will be most interested in based on what they've rated and what friends of theirs have rated.   As a result of being able to isolate relevant content, Break has extremely high engagement relative to other male video entertainment properties.

BI: How are you integrating ad serving?

Richman: Advertisers and advertising provide an additional layer of the site experience that can be closely tied into the content interests of their audience.  They're a side benefit that don't detract or distract from our main purpose.  We're not a social network so we don't know or want to know so much about whom your first girlfriend was or what your nickname is.  We leverage the data we know, and our recommendation engine in particular, to better serve advertiser interests. 

First, we enable advertisers to target audience based on specific categories of video.   Advertisers such as Electronic Arts, Eidos and Vivendi Games have taken advantage of this.   Even better, however, if someone never visits the video game category but continually watches videos about Halo 3 and Guitar Player, we can show them gaming videos whether they are in a video game category or not.   Second, we enable advertisers to target specific users directly through our recommendation engine by integrating their brand.   For example, in a recent campaign for Saw IV, users were shown 'personalized recommendations,' 'friend's recommendations' and 'Jigsaw's Recommendations.'  The campaign actually integrated elements of the brand itself.  Third, we allow advertisers to target featured videos by rating 'G,' 'PG' etc. That way there's total transparency for brands and a feeling of control over placement.  We can also target ads to users who tend to watch a certain rating of a video.    This safety level has enabled us to properly serve brands such as Burger King, Keystone Light and over 100 others.  The differences between just having a general sense of demographics and a deeper psychographic profile are enormous.

BI: Now that your behavioral platform is in place, what are your goals?

Richman: Looking forward into 2008, we're planning to more aggressively educate advertisers about how the behavioral layer added to existing demographic profiles unlocks entirely new possibilities to advertise by psychographics. Imagine that you have people who are big fans of a specific horror genre and director. Then you provide these people with personal recommendations by that submitter.  A further step toward customization of ad message is the introduction of an embeddable video player. This makes it possible for advertisers to create a look and feel for their video, including color and graphics and other aspects of ad formats. This opens up the potential for dynamically changing the viewer experience of ads, based on their interests, preferences and habits.