When I think about personalised content, I think of Netflix and Spotify providing recommendations during the onboarding process based on a user’s tastes and preferences. These recommendations will then be refined and augmented as a user consumes more content. This level of personalisation is fully algorithm based, which makes it highly scalable but also prone to ‘inaccuracies’ (or simply not doing much for the discovery of new content).
Spotify has adjusted the ‘algorithm only’ personalisation model by introducing ‘algotorial playlists’. As the name suggests, these playlists are a combination of editorial and algorithmic content. This level of personalisation is based on Spotify’s algorithms with a personalised touch from Spotify’s curators. Oskar Stål, VP of Personalisation at Spotify, explains that algotorial playlists are “sets of songs our editors put together to evoke a certain mood or moment that are also tailored to the individual user.”
In a hugely helpful Medium post, Oskar Eichler — CEO at Songstats — breaks down how Spotify’s algotorial playlists are being generated:
Pinned positions for logged in users
Spotify’s editorial team curate a set number of tracks as part of a personalised playlist. If you access this playlist without being logged in, you’ll see the same list of tracks in the same order each time you access the list.
Once a user logs into Spotify, the Spotify algorithm starts shuffling tracks around and the track ordering changes as a result. However, logged in users will now see ‘Pinned Positions’: tracks that are always on the same spot in the playlist irrespective of which user views the playlist. These pinned positions are chosen by Spotify’s editorial team, in line with a certain mood that the playlist is trying to evoke (think of Spotify’s “Songs to Sing in the Car” or “Cozy Blend” for example). Each pinned position has a timestamp that determines a range of dates during which a specific track is pinned to a set position in the playlist. Based on the analysis that Songstat did, 10–30 tracks are being pinned to their position within a personalised playlist for about a week.
Priority Tracks
Spotify also selects ‘Priority Tracks’ that, when shuffled, will always average out to be positioned at the top of the playlist. These 10 or so tracks have been curated, and the subsequent tracks in the playlist have been personalised based on a users’ listening behaviour. Spotify’s editors can thus pin a certain to a top spot in a personalised playlist.
Main learning point: I’m curious about if the combination of machine and humanly generated content in Spotify’s algotorial playlists adds to the serendipitous effect of music discovery.
In a hugely helpful Medium post, Oskar Eichler — CEO at Songstats — breaks down how Spotify’s algotorial playlists are being generated:
Pinned positions for logged in users
Spotify’s editorial team curate a set number of tracks as part of a personalised playlist. If you access this playlist without being logged in, you’ll see the same list of tracks in the same order each time you access the list.
Once a user logs into Spotify, the Spotify algorithm starts shuffling tracks around and the track ordering changes as a result. However, logged in users will now see ‘Pinned Positions’: tracks that are always on the same spot in the playlist irrespective of which user views the playlist. These pinned positions are chosen by Spotify’s editorial team, in line with a certain mood that the playlist is trying to evoke (think of Spotify’s “Songs to Sing in the Car” or “Cozy Blend” for example). Each pinned position has a timestamp that determines a range of dates during which a specific track is pinned to a set position in the playlist. Based on the analysis that Songstat did, 10–30 tracks are being pinned to their position within a personalised playlist for about a week.
Priority Tracks
Spotify also selects ‘Priority Tracks’ that, when shuffled, will always average out to be positioned at the top of the playlist. These 10 or so tracks have been curated, and the subsequent tracks in the playlist have been personalised based on a users’ listening behaviour. Spotify’s editors can thus pin a certain to a top spot in a personalised playlist.
Main learning point: I’m curious about if the combination of machine and humanly generated content in Spotify’s algotorial playlists adds to the serendipitous effect of music discovery.