A few months ago, I spoke with Eric Bieschke, Pandora’s’ Chief Scientist, for title="Pandora Radio’s Dominance Built On Big Data Edge">a story about the company’s approach to analyzing the music listening data it gleans from its 72 million active listeners. Pandora’s hallmark has always been its nearly flawless ability to choose songs you like, based on what you’ve previously listened to. And in this earlier piece I noted that, “with its recent hire of former advertising executive Brian McAndrews as their new CEO, Pandora seems intent on exploiting its cache of big data to expand its revenue opportunities.”
This weekend, the New York Times ran a piece about Pandora’s growing use of the analysis of its customers’ listening habits, to determine the kinds of ads they’re most likely to respond to. This approach is no great surprise, of course. While Pandora offers a premium ad-free subscription for $4 per month, almost 90% of the company’s $427.1 million in revenue is generated by advertising. By serving up ads that its users are more likely to click on, Pandora can potentially demand more lucrative ad rates. And as Bieschke told the New Times, “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.”
What makes Pandora such an interesting player in the customized ad game is the sheer volume of user data it possesses (eight years’ worth), and how granular Pandora’s analysis of that data has been. Beyond using demographic basics like age, gender and zip code, which customers supply at sign-up, Pandora chooses which song to play by analyzing variables like the time of day you’re listening and the streaming device you’re using. And of course, Pandora has a vast trove of user thumbs up/thumbs down and skip interactions to further fine-tune its playlist. Such a nuanced view of customer listening preferences has relevance well beyond musical tastes. As the New York Times story points out, “Certain product or cultural preferences can give glimpses into consumers’ political beliefs, religious faith, sexual orientation or other intimate issues.” No one has spun this thread into gold quite like Google has, of course, but they have the advantage of leveraging portions of users’ email and phone data, as well as social media activity. What Pandora is doing, based largely on musical preferences, is quite impressive from a data-mining perspective.
This ability to make reasonably accurate inferences about our worldview is what makes Pandora a very attractive advertising platform. And this speaks to the larger world of online services and apps. From politics to car preferences, our tastes in music, clothes, books, etc., can offer insights into our behavior on a level that few of us are likely to contemplate during the course of our online listening, shopping or reading sessions. And for Pandora, the ability to predict our responses to ads, not just songs, may turn out to be its most profitable trick.