You can think of Echo Nest as an algorithmic think tank which crunches the petabytes (you'll be hearing a lot of that word in the future) of data that its clients, including Spotify's tens of millions of users, generate every time they pause, play or playlist one of the service's 30-million-plus songs.
The center of Lamere's talk was on the four well-known segments of the music listening audience: savants (most engaged), enthusiasts, casuals and indifferents (least engaged). He broke down their constituent percentages as well, with savants composing 10% of the audience, engaged at 20%, casuals at 30% and indifferents at 40%.
Lamere said that, for the most part, Spotify and its competitors had already conquered the savants and, to grow to the scale needed or desired by these companies, they would have to target casual -- or 'lean back' -- listeners, those for whom turning on the radio in the car is the most engagement they have with music. "If they really want to grow, there's 70% of the market that thinks $10 a month is too much," Lamere said.
So, he said, "you have to find a way to engage the people that are going to be intimated by a search box [sitting] in front of 30 million songs."
The way to do that, Lamere contends, is to make it easy for casual listeners to discover music on their own terms. He cited some programs he'd written either for the Echo Nest or one of the many Music Hack Days he's taken part in. One is "Music Popcorn," an application which presents a cloud of genres for the listener to choose from, after which it plays examples from that genre. Another is the Infinite Jukebox, which makes unending versions of users' favorite songs. The end result is the user seeing an easy-to-understand interface which they can engage with (and be monitored by) and, the hope is, hear new music they would feel they'd played a part in finding. (Also worth a peek are Lamere's other applications: "Girl Talk in a Box," "Boil the Frog," and "Six Degrees of Black Sabbath," with a special mention to "The Bonhamizer."
While getting these listeners more engaged with their music is ideal, so is giving them exactly what they want, when they want it -- a problematic solution for math equations to square. "There's a problem here: You can generalize a room, but when you try to define a single individual, you run into trouble." Meaning these algorithms will need to collect contextual data on listeners in order to be able to predict wide-ranging tastes on an individual level. Contextual data is what you listen to while at the gym, or on Monday morning before work, or when you and your significant other are alone. The way this data is collected is, well... creepy. And a common refrain.
My Day: Jim Lucchese, CEO of the Echo Nest
"Phones know where we are, what we're listening to while we're there... this is a really important thing. Headphones will know when you're nodding your head to the beat -- our devices will know whether we're excited, stressed, and [they] will play us the right song based on that."
The talk served as a restrospective preview of what Spotify may look like in the near future, though Lamere declined to answer questions about his company's existing contracts with Rdio and other music services. "I won't be answering any questions like that." He'd begun the talk with a similar warning: "I prepared this talk long before I was an employee at Spotify... so no secret plans that Spotify might be having, or secret data that only Spotify would know about."
It's an alarming future for those averse to being watched -- but for the casual (in all senses) among us, it might be the beginning of a new love for a lot of new music.