Playlist recommendations key to streaming service success
Spotify developer tells Web Summit its Discover Weekly feature generated more than one billion streams in 10 weeks
Matthew Ogle, senior product developer at Spotify: streaming firm’s Discover Weekly feature, which provides a personalised “mix tape” in playlist form for each of the company’s 75 million users, generated more than one billion streams in 10 weeks
Matthew Ogle, a senior product developer at Spotify, took to the Music Summit stage to dispel some myths about music discovery, including the notion that it should be left up to users to seek it out.
He drew a comparison between Spotify’s Discover Weekly feature, which provides a personalised “mix tape” in playlist form for each of the company’s 75 million users, and the regular “browse” screen on the music service. The former generated more than one billion streams in 10 weeks; the latter does okay, he said, but most users never visit the screen. “It does ok, but it’s not setting the world on fire,” he said. “When you’re implementing a discovery feature, what you call it matters.”
It’s not something just for “music nerds” either, Ogle said. New users on Spotify that would be considered mainstream spend almost 80 per cent as much time as the power users of Spotify on the discovery feature.
Discover Weekly creates individual playlists for users based on their preferences, using an algorithm that builds playlists like a human”. That was achieved by feeding the algorithm the billions of playlists created by users on the Spotify platform since it was set up, giving it the chance to learn.
It also dispels a second myth - that you either need more data to solve the music discovery issue, or that only the human touch will do. The computer versus human battle is a “false fight” he said.
“It doesn’t even exist. The reason it’s so pervasive today is that it’s really seductive to talk about humans versus robots, but it doesn’t really describe how we work in Spotify or how any company doing tech on any scale works,” he said.
“Algorithms are people too. We didn’t have a computer watch High Fidelity 10 times and make mix tapes.”
Mr Ogle has the experience to back up his claims. After spending six years at Last.fm, which was one of the first products to bring in collaborative filtering for music, he moved to Echonest, which was involved in examining music data on the web, and helped establish This is my Jam, before moving to Spotify.
“Discovery is not a problem to be solved,” he said. “It’s a huge opportunity.”