Streaming music is gaining popularity fast, but lags behind the web in understanding its users.
Internet content sites can tag, track, and categorize people as they move about the web, creating deep and accurate user profiles that follow most of us as we move among digital properties. Combined with programmatic ad buying, it means that when a user visits a certain website for the first time, that site can display an ad that is more-or-less precisely relevant to that user. Advertisers pay more for effective targeting.
Dedicated music apps like Pandora or Spotify exist as islands, separated from broader engagement with the oceanic Internet. They are years behind the web in understanding their users. Two advances have sparked what promises to be an intensely developed intelligence layer that can understand, profile, and target the users of streaming music services.
First, Pandora announced audience segmentation based on user-registration information. (RAIN coverage here.) The first segments released by Pandora to advertisers are Hispanic users and Spanish-speaking users. More segments are doubtless forthcoming as Pandora develops and refines its testing and predicting methods.
Second, and more recent, The Echo Nest released its Music Audience Understanding platform, which leverages that company’s immense data intelligence about music choices to better profile online music users. (TargetSpot, a leading digital audio advertising network, was the launch partner and first client.) The premise of The Echo Nest’s development is that a proper analysis of music preferences can predict non-music attributes of a person and cohorts of people.
Does it work?
“The Echo Nest’s technology delivers predictions that are at least as accurate as registration data, or more so,” said Mitch Kline, CEO of TargetSpot. “Some advertisers think registration data is gospel, but we believe The Echo Nest’s technology is more accurate.”
It is certainly more ambitious than Pandora’s first venture into segmenting. The Echo Nest has released 20 audience segments that cover demographic categories (age, gender) and lifestyle inclinations (automotive, parenting, etc.). That starts to get very interesting to advertisers.
“It is increasingly important,” Kline affirmed. “Our clients are trying to get more and more targeted. Dumb inventory doesn’t do anything for them anymore. People are now looking for advanced targeting. Automobile manufacturers want to target auto intenders. Or, an insurance company wants to target people who are in the car market because they will need auto insurance. In digital audio, this is fairly new.”
Can music taste really predict whether someone likes cars, or is a parent? We put that question to Jim Lucchese, CEO of The Echo Nest, and he described elaborate methods of testing against the “ground truth” of music-service users, gleaned from multiple sources. Correlating music choices against populations that live in a certain “ground truth” segment (like auto enthusiasts or parents), results in a prediction reliability score.
“That’s part of what took us so long getting to this point. The segments that we’ve released are ones where we’re confident that the results are predictable enough, and reliable enough, to take to market. There were a number of things that weren’t successful. Music can’t predict everything about a person.”
Big Data and privacy
Though it might seem that the inference level is high when connecting music choices to lifestyle interests, that’s what “Moneyball” and Big Data are all about, and why there is such promise that analytics can illuminate previously unknown connections between all sorts of things.
Lucchese told us that the Music Audience Understanding platform was in pre-release development for two years, and that the launch timing, close on the heels of Pandora’s audience-targeting system, was coincidental, and probably favorable to both companies.
“[The timing] was awesome! The internal conversation we had was, ‘OK, cool, there’s going to be market discussion about this.’ Prior to that, you didn’t read much about it. [Our] timing was set well in advance — we needed to have an initial customer lined up; we needed to have a product ready. We saw the Pandora announcement as we were preparing our own announcement. I saw that as a huge plus, educating the market around music as a powerful predictor of people and applying a level of data vocabulary around streaming music.”
Interestingly, Lucchese also noted, “I don’t view Pandora as a competitor; I view it as a prospective customer. Pandora is one of the most forward-thinking companies in the space.” (Pandora does not use The Echo Nest’s music intelligence platform, which is utilized by over 400 other music services.)
All of this makes privacy advocates uncomfortable. Lucchese hit that issue head-on: “We really wanted to make sure anything we were doing was not only compliant from a privacy standpoint, but better the current state of the market. [Our system] is not only non-personally identifiable, we’re not even tracking users. Because we can look at anonymous clusters of listeners, and make predictions based on those clusters, we’re not tracking people around the Internet or dropping cookies.”
Mitch Kline expresses the natural enthusiasm of a launch partner, while looking beyond the first-mover advantage: “We believe in The Echo Nest and this technology. We’re the first ones to dive into this space, and we think others will dive into it also. Then we’ll have to think of other ways of targeting!”