Drawing on its rich data set and leveraging best-in-class intelligent automation, the Swedish company is testing a new, hyper-personalized feed on its app home screen. The feed is meant to introduce users to new music for them to enjoy. Every day, the feed recommends fifteen songs accompanied by “canvas” loops – the dynamic Spotify GIFs that appear in the background of songs, and which are presented in a TikTok-inspired vertical video feed. Spotify listeners are more likely to continue listening to songs that are accompanied by canvas videos, as well as more likely to share the track, either via text or on other social platforms like Instagram where the canvas loops also appear.
The daily song recommendations are able to be “refreshed” with Spotify’s Enhance feature, which gives the app’s powerful algorithm the go ahead to automatically add in songs it thinks will fit into the “vibe” of the playlist being curated. What’s unique about the new feed is how Spotify is sourcing the songs it recommends to listeners: by mood and emotion.
There are more than four billion user-created playlists on the Spotify platform, with more being generated every day, and the composition and naming patterns of these playlists give Spotify’s data scientists important clues into the emotional content and context of the songs they contain.
For example, if a song is in, say, twenty thousand playlists with the word “Happy” in their title, then it's pretty likely that the song is upbeat and happy. Whereas if another song is in a “Rainy Bus Ride” playlist, it is more likely to be reflective and introspective. By triaging these insights and combining them with user-specific behavioral data, Spotify can provide appropriately tailored playlists and track recommendations that offer just the right fit for the experience a particular user is hoping for when they head to the platform.
This article originally appeared in the PSFK iQ report, Optimizing The Personalization Of The Customer Journey.