How Spotify selects songs to recommend to us

Spotify offers an impressive track recommendation system (overall audio content) that is based on machine learning and sophisticated algorithms. However, for some users, the repetitive recommendations of songs they have already heard may not be very pleasant listening. The company recently revealed some aspects of its recommendation algorithm, which may help users understand why certain songs keep popping up. He brought the news XDAdevelopers.

Spotify’s algorithm relies on three main pillars of recommendations

1. Collective filtration: This method compares similarities between users based on their musical tastes. If two users tend to listen to similar songs, the app can suggest songs that both of them like. This system is also very popular among other platforms, including Netflix.
2. Content-based filtering: In this method, Spotify recommends music based on the characteristics of the songs themselves, such as genre, tempo or mood. For example, if someone often listens to upbeat pop songs, the algorithm can recommend other songs with similar elements. Sometimes even me after songs that have a different tempo or direction.
3. Deep learning and neural networks: Spotify uses neural networks to identify complex patterns in user data, creating personalized playlists like Discover Weekly. Recurrent Neural Networks (RNNs) focus on sequential recommendations, where they predict based on history what the user might want to listen to next. These networks are constantly evolving, so sometimes the algorithm may not always be to your liking.

Another aspect of recommendations is function Discovery Modewhich allows artists and labels to highlight certain songs in recommendations. These songs are only favored if they are in line with the user’s taste profile, so if a song doesn’t get enough response, its priority goes down. At the same time, Spotify customizes search results based on how the user interacts with the recommended songs.

You might be interested in:

We can influence our recommendations, for example, by marking songs as uninteresting or excluding specific playlists. We believe that the platform will bring you good enough songs based on your preferences in the future. If not, you can still exchange your subscription here, for a competing platform like Apple Music or Tidal.