Algorithms sometimes know more about us than we do. This is particularly the case with streaming services. Therefore, we would like to explain to you how the Spotify algorithm works and why it changes the music industry in the long term.
Brief historical review: Spotify – better the predecessor Echo Nest – was originally developed from a doctoral thesis by two MIT graduates – namely Tristan Jehan and Brian Whitman. Their goal was to dissect and separate each piece of music in order to make it analyzable.
In order to obtain these small pieces of the puzzle, they evaluated several million songs using artificial intelligence (AI). The program managed to filter out conformity independently. These are fixed structures within different music genres.
Around 60 people are currently working on the Spotify algorithm, which has always been called Echo Nest , at the headquarters in Somerville, USA .
How does the Spotify algorithm work?
Every user will be able to underline that the Spotify algorithm is very successful in suggesting suitable pieces of music and attractive playlists. At the moment he works with three main components.
1. The Profile button feeds the Spotify algorithm
Spotify creates a so-called personal “button profile” for each user. This summarizes which songs, albums, artists, playlists and podcasts everyone hears.
The location, duration and frequency are also recorded. The Spotify algorithm summarizes this data in the Profiles button of the user and compares it with that of other subscribers.
So if there are frequent overlaps in the amount of data, songs of other users are shown in the “Discovery Weekly Playlist”. Accordingly, playlists that are heard by many followers also appear more frequently in the corresponding Discovery Weekly folders.
Spotify’s algorithm also examines individual parameters of the songs. The puzzle pieces mentioned at the beginning include the timbre of the song, the instruments used, the tempo as well as the highs and lows of the piece of music.
Since 2015, Spotify has been using an AI for the complex analysis of these parameters, which currently unraveled the songs in over 1,500 different genres. The Spotify algorithm also takes regional differences into account.
As a user, you may have already noticed this phenomenon: Sometimes Spotify spits out nonsensical song suggestions. This has to do with the genres of the pieces of music, which also seem quite irritating.
2. Emotions are part of the analysis
Another essential parameter that plays a role in the analysis are the emotions that the songs convey. So Spotify differentiates between happy and sad songs and those that sound melancholic. Many experts see this classification as critical.
Because every user most likely combines different feelings with a certain song. Associations, memories or the current, individual emotional state are also responsible for how we feel about a song.
Here too, the Spotify algorithm helps: if a certain song appears more often in a playlist spiked with love songs, the AI ??concludes that it is also a love song.
Isn’t this conclusion too short-sighted? In any case, it is certain that the algorithm gets better and better through machine learning.
3. The Spotify algorithm reads and listens everywhere
Not really surprising: The Spotify algorithm continuously scans the entire network for clues to be able to assess a particular song.
Everything revolves around the question of how a song is received. Spotify uses blog articles, comments on websites, Facebook or YouTube, as well as tweets. The algorithm analyzes each opinion, relates them and captures an exact opinion of each song.
Spotify will find out within a few days how a song is received by the internet community and whether it is recommended. This data then determines the Discover function and other recommendation formats from Spotify.
Spotify doesn’t just rely on AI-generated suggestions
But Spotify is not that easy. The platform is well aware that our taste in music cannot be determined entirely on the basis of certain parameters and collected data.
That’s why Spotify editors and music connoisseurs have been filling an expanded editorial playlist with fresh songs by new artists since 2019. The aim is to give previously unknown musicians a chance in the individual countries.
Conclusion: what is to come?
In the future, the Spotify algorithm will be able to propose pieces of music to subscribers according to our current daily situation.
If we sit on the train to work on time at 7 a.m., the platform could suggest songs that fit exactly. Or in the afternoon, when we are on our way to the gym: The Spotify algorithm can then remember the times and recommend appropriate music according to the schedule.
Uncertainties remain: Because we, as individuals, are too unpredictable, so Spotify can suggest the right music at any time of the day or night.
But one thing makes us shudder: Spotify always manages to better predict how individually tailored media usage can look in the coming years.
