Targeting Taste: The Emerging Science Of Preference Profiling

From the time I started covering behavioral targeting back in 2001, much of the field was shaped by the purchase funnel. Using content consumption, click histories, etc. to locate consumers who are “in-market” for specific purchases was the understandable low-hanging fruit for a nascent industry that had enough of a challenge just finding these most glaring purchase intent signals at scale. And to this day, for all the supposed sophistication of the ad tech economy to profile consumers and find segments everywhere and anywhere online, retargeting recent visitors to retail Web sites remains the bread and butter of behavioral targeting.

But in recent years we have seen a richer approach to user profiling, based less on where someone is in a specific purchase process for a particular item than on who they are and their receptivity to certain messages. One glaring success story in this model was the Obama 2012 presidential campaign, which used TV viewing tastes to super-segment audiences for highly targeted political messaging. For those who missed the incredible walkthrough of that process at OMMA DDM last July, check out the video of the keynote by Larry Grisolano, CEO, AKPD Message and Media, and former head of paid media buying for Obama for America. It was a textbook case of how viewing tastes were rendered as voting preferences.

I was reminded of this by a New York Times news story on music service Pandora and its experiments in ad targeting based on musical preferences. Pandora has been targeting taste for over a decade now. Its core technology focuses on a musical “genome” tracking common elements that run deeper and subtler than just genres. It can assemble personalized channels of music that had as much to do with beat, tempo, instrumentation, etc.

Like Amazon, Pandora’s is a remarkable story of how algorithms and automated recommendation engines can over time build relationships with users. For instance, Pandora recently experienced the biggest challenge to its ubiquitous Internet radio platform from Apple, which launched a rival service that is baked into the iOS 7 operating system. Despite Apple’s home field advantage, Pandora does not appear to have lost a step, in large part I believe because longtime users are hesitant to leave a system where their profiles have been built for years. My wife, who is otherwise tech-averse, regards her Grateful Dead station on Pandora with personal affection.

The same technology that has recorded and targeted her taste is now being applied to ad targeting, Pandora tells the Times. Eric Bieschke, chief scientist at the company, says, ““It’s becoming quite apparent to us that the world of playing the perfect music to people and the world of playing perfect advertising to them are strikingly similar."

It's not just about targeting a user’s musical tastes, but also detecting shifting moods reflected by the tracks a listener is choosing or liking during different times and even in different places. Pandora describes scenarios where listeners may be more receptive to promotions for adventure travel, for instance, when they are listening to certain types of music on the weekend rather than they might be when listening in the office on a weekday. By correlating tastes, times, places and click histories, the company is looking to find affinities most of us never considered before.

Correlations have been made in the past for many different behaviors. Musical taste and politics, for instance, is a no-brainer. The article quotes a computer scientist who suggests that attitudes toward gun control or the environment map well against certain musical tastes. Actually, Pandora is already doing this -- targeting ads for specific parties and candidates according to musical tastes.

Sure, that's a no-brainer. But the real potential here is in branding. With so much attention in behavioral targeting traditionally focused on intent, recency, retargeting, much of the energy goes to direct marketing. For much of display advertising, and despite a decade of promises to do otherwise, the technological focus has been on direct marketing that can be tied to ROI. By emphasizing broader profiles that leverage vaguer qualities like sensibilities, taste, moods, however, we may see a more natural path to target the bigger brand stories to specific consumers.

Recommend (2)
1 comment about "Targeting Taste: The Emerging Science Of Preference Profiling".
  1. Gray Hammond from Quire , January 6, 2014 at 4:34 p.m.
    We have long known that reading (magazines, books) and radio/music are strong psychographic indicators. Too bad most "old media" planners never really leveraged that - maybe big digital data players (following Amazon's lead) will be smarter?