SOMERVILLE, Mass. — The scrappy music/tech startup known as The Echo Nest occupies two floors of a brick building in Somerville’s Davis Square. It looks and sounds like the busy bed of contemporary geek-hacker culture that it is. Tunes waft out through speakers, filling the communal workspace. Young engineers huddle around computers, tapping away.
You may have never heard of The Echo Nest, but if you’ve ever listened to music streaming sites like Spotify, Rdio or iHeartRadio you have used the local company’s musical data. It creates the behind-the-scenes song recommendation platforms for those clients, along with a slew of others (Clear Channel, MOG, SiriusXM…), and it serves 35,000 app developers.
But things are changing at The Echo Nest. Sweden’s Spotify recently acquired the American venture for what some industry reporters have estimated to be $100 million.
On a tour of the office, The Echo Nest’s soft-spoken and decidedly low-key co-founder Brian Whitman takes me to a conference room and kindly interrupts a meeting. He jokingly asks his two employees, “You’re doing work, right? You’re having a business conversation?”
They laugh, nod and tell him they’re talking about “user listening.” Whitman replies, “Oh, that’s my favorite thing.”
“User listening” is one of the many somewhat-mysterious tech-speak phrases in the lexicon here. This once-tiny startup is said to be the industry’s leading source of music-related data.
Ajay Kalia, the guy Whitman affectionately grills, is a product manager for what they call “taste profiles.” He tells me those say something about a music listener’s personality or interests. Kalia also says he’s totally psyched about Spotify’s recent acquisition of The Echo Nest because — until now — his parents could not understand what he does for a living.
“I tried to describe my job to them before, and it involved a lot of terms like ‘API’ and ‘services’ and ‘machine learning,’ ” Kalia said. “Now I can pull out my phone and show them Spotify and pull up any song that they heard on ‘Dancing With the Stars’ last night and then show them here’s other artists that are very similar to that. ‘Here’s music you might have heard when you were growing up, and here’s what we can understand about you based on the past five songs I just saw you pick.’ ”
Teaching A Computer To Listen To Music
It can be hard to get your head around the technical lingo. One way to grasp what The Echo Nest does might be to imagine it as an engine operating under Spotify’s hood. The folks here gather a mind-blowing amount of data — about songs and about how people listen. Then they crunch it, with a lot of help from computers, to essentially design and customize the music users hear through streaming sites on their computers and mobile devices.
So when we like or dislike a song on Spotify, well, The Echo Nest’s complex algorithms do their best to predict what we want to listen to next.
But how do they do that? Co-founder Tristan Jehan explains his part this way: “What I do is teach the computer how to listen to music, that’s probably the easiest way to understand it.”
The scientist/engineer’s software extracts musical attributes from songs, such as tempo, key, time signature, harmony, color, danceability — the list goes on. All of this “musical intelligence” gets organized and indexed to feed the recommendation engines.
Jihan and Whitman concocted their concept as grad students at MIT’s Media Lab. They founded The Echo Nest in 2005 with seed money from a few angel investors and professors. Now they have 65 employees. Jihan says the team has analyzed about 200 million songs.
But The Echo Nest also analyzes us, to understand how we listen, when we listen, where we listen — like at a dinner party, the gym or in the car. Also how we talk about music globally. They comb (or “scrape”) through millions of Web pages and blog posts about music for descriptive terms and sounds that are emerging. That’s where Glenn McDonald comes in. His title at The Echo Nest? “Data alchemist.”
McDonald shows me his vast online genre map. It’s called “Every Noise At Once,” and it organizes genres visually on two axes — left to right and top to bottom. McDonald clicks through samples of music — more than 1,100 types — that are related in some way. It’s something like a color palette for song styles, and some of the genres are pretty obscure. We move from “hard core industrial” to “Belgian hip hop” to “dim core” to “orgcore.”
Then he gleefully hits… “Viking Metal! It speaks to us in our common heritage in the north,” McDonald says with a laugh.
Down at the bottom of the map you’ll find the most organic, acoustic, classical sounds.
“Here’s our beautiful harp music,” McDonald explains with a relieved sigh. “Ah, so free are we from the relentless drum machines and artificial synthesizers of the modern era.”
McDonald’s genre map is like a happy rabbit hole for the musically curious. It’s packed with so many things the international scene has to offer. But while it’s mind-opening, entertaining and undeniably cool, how does it help The Echo Nest fulfill its mission?
“It all sort of relates,” McDonald answers. “The map is a way of understanding the data. That data also goes into playlist generation and into recommendations and into personalizations.”
So if a Spotify user is interested in Belgian hip hop, The Echo Nest’s algorithm might deliver songs in that style from Brazil or Australia.
“Our goal is to basically connect more people to music,” explains CEO Jim Lucchese, adding that joining with Spotify will give both companies a data boost that should help them grow.
“And though there’s a lot of math involved, really all we’re doing is trying to understand the world of music in a really in-depth way, and then try to understand each fan in a really in-depth way,” Lucchese says. “Just like your friend who you rely on to connect you to music.”
Our Music Listening Future
“I think that’s kind of just corporate speak,” says music business reporter Ben Sisario, who’s been following The Echo Nest for The New York Times.
“Connecting people with music — I mean that sounds nice — but really what they’re trying to do is understand consumers at a very deep level when it comes to their music choices, and that’s to continue to sell you the product — which is the music,” Sisario says. “It’s also to let advertisers sell you their products.”
Like toothpaste. Or zit cream. At the same time Sisario says he’s a fan of music streaming sites — even if their computer algorithms are far from perfect. And he says online listening companies keep getting bigger and more valuable in the music industry landscape. He believes The Echo Nest is poised to have everything to do with our music listening future. And, Sisario adds, besides telling you want songs you might like, the rich data The Echo Nest miners gather from global chatter could very well predict the next Lady Gaga.
But even with all that potential, at this poignant moment in time, the team at The Echo Nest admits there’s a certain nostalgia that comes with being bought by a multinational company the size of Spotify.
“We’ve built a culture here and it was us — this was The Echo Nest,” Paul Lamere tells me. He’s worn many hats over the past five years here, including “developer evangelist.”
“It’s a little bit sad to see that phase of the company go away,” he says. “We’re no longer a scrappy startup in Somerville. We’re now part of a much larger company. And so that’s great — but that’s what happens when you grow up.”
Lamere has created some really creative music apps and “hacks,” as they’re called, including, “Boil The Frog,” “The Infinite Jukebox” and “6 Degrees of Black Sabbath.” And while he’s clearly a bit wistful, Lamere acknowledges that after the initial growing pains subside, the Spotify acquisition just means The Echo Nest’s scientists, engineers and self-proclaimed music freaks will have way more data to play with.