“Which means,” Conrad said, “we can innovate in the user experience, we can add new features, and experiment with new advertising formats in ways we never could before.”
Currently, Pandora just does audio ads, said Ian Geller, VP of business development. “Not to say we’ll go in this direction but when we have control of the browser, you can imagine us experimenting with more visually contextually appropriate advertising on a TV,” Geller explained. “Why should a TV just play audio ads? It’s a TV.”
This updated delivery method also gives Pandora the potential to build better targeting and personalization across devices. “We have control over the application,” Geller said. “We know what people are doing in that particular time and there’s no disparity between what the client knows and what the server knows.” Context is especially important for sending appropriate ads given a listener’s circumstances. For instance, Pandora doesn’t want to send an ad with a call-to-action to someone driving a vehicle in rush hour traffic.
Beyond basic appropriateness, Pandora also hopes to enable a seamless cross-device experience. Conrad posited a scenario in which someone listening to Pandora in his car arrives home, the station synchronizes on his phone as he enters his house, then onto the television where his wife votes thumbs down on the Justin Bieber song he was listening to.
“Pandora understands at that moment that this is a personalization context of you and your spouse, not just you,” said Conrad.
Pandora has steadily been rolling out ad targeting products over the last year, many of which are powered around Pandora’s ability to segment its 250 million-plus pool of registered listeners based on traits like ethnicity, income and political leaning. This involves cross-referencing third-party data like census information or Gallup polls (which break down political affiliation by zip code, and which Pandora uses as a key to probabilistically determine whether an individual is Republican or Democrat) with Pandora’s information about users’ listening habits.
For instance, Conrad explained, Pandora might be able to dive into a pool of 80 million-plus people who listened to Pandora last month and extract concrete data of the 2 million users the radio service knows are Hispanic. “Using data science, we can take that pool of two million and project it intelligently across our user base by looking at what radio stations they listen to and with unbelievable accuracy, predict that the remaining 78 million are also Hispanic,” Conrad said.
Listening habits naturally play a large role in determining listener traits. It’s not a surprise, for instance, that Republicans tend to listen to country, whereas Democrats tend to listen to jazz, reggae and new age.
Jack Krawczyk, who leads Pandora’s product team for ads, pointed out that while this may be implicit, a listener’s musical diversity and preferences for specific genres within that diversity tend to provide clearer indicators of traits like political leanings.
“We can figure out via a probability model, it’s not just what genres you like but how much you prefer them within your listening spectrum,” Krawczyk explained. “A preference for jazz is only a strong indicator if it takes up around 35% of what you listen to each month. If you have a jazz station but only listen to it 5% of the time, it isn’t a good indicator.”
Still, these new capabilities and ad products exist in territory into which some marketers are stepping cautiously. Conrad acknowledged that many partners still want third-party validation that Pandora’s segments are accurate, and that the company is in the process of getting that validation to ease those concerns.
Greg Polsca, manager of digital marketing at dietary supplement provider Pharmavite, who attended the Pandora tour, indicated that right now his company’s presence was mostly exploratory. Pharmavite currently has a relationship with Mindshare, but isn’t a Pandora advertiser, and Polsca said his interest in learning about Pandora’s offerings had much to do with connecting with consumers and learning about the “demographic and psychographic factors” that affected the way they interacted with the world.
And for Conrad’s part, there’s some reluctance in overemphasizing Pandora products for advertisers instead of products for consumers. When asked about leveraging real-time location to push ads, Conrad said that while he has no trepidations about providing that feature, he feels “a certain obligation to deliver consumer value beyond just better advertising using location, before I ask for permission to track for location. I’d like to do some non-advertising-related consumer benefits as it relates to real-time location first, or maybe in conjunction with turning it on for advertising.”