How music platforms know our tastes

Have you ever wondered how music platforms know what we want and on a regular basis offer us new music that fits your taste?

Are you a regular stream user? Are you listening to Spotify, Apple Music, Google Play Music playlists? You’re in good company! Just to give you an example, in the third quarter of 2020, 320 million people used Spotify. Of these, 144 million were paying users.

The strength of these platforms is that they allow people to play an infinite amount of songs against a monthly subscription fee much lower than what would be necessary to purchase Cds, also offering a virtually perfect audio quality. But that’s not all, their strength lies also in using systems that can examine the signals that we send as we interact with the platform to be able to read better and better in our minds so as to understand our tastes and anticipate suggesting choices.

A speech that reconnects and deepens the concept of sensory marketing that we have already talked about

Playlists say a lot about us

Have you ever wondered how these platforms work? They are based on algorithms, a precise sequence of instructions, a system of calculations that through a certain number of steps collects and processes the information that we send to achieve a certain goal.

The strength of these platforms is the level of customization and the possibility offered to users to expand their musical knowledge taking into account their tastes.

For this purpose great importance is given to the creation and use of playlists. By listening to a playlist we communicate to the platform what are our preferences in terms of music and at the same time, we provide suggestions to create new playlists that meet the tastes of other users. Specific software filters the contents and creates personalized and specific suggestions for each user. On the platforms are loaded playlists created specifically for each individual user and based on his listening habits (what he shared until then, what he has shown to like, what he saved, and what he skipped) and on those of people with similar tastes.

For example, if some of my favorite songs appear in other playlists and these have in common another song that I do not listen to, this song will be suggested to me with good probability that it meets my taste. The same applies, of course, to all other users.

How the algorithm works

The more a person is a frequent user of the platform, the more they will have gathered information about his tastes, thus being able to create playlists of songs that meet their preferences. As the platform is used, user information is updated and the playlists that will be created will be in line with these new data.

The algorithm used is based on three systems that study the way users interact with the network:

  • Collaborative Filtering, that is the feedback provided by the users, the number of listenings, the listening time, the clicks on the artist page, how many “hearts” has received the track etc.
  • Natural Processing Language, used to probe blogs, authoritative websites, and other resources to understand what is being said about music trends and specific tracks;
  • Raw Audio, catalogs the tracks according to their characteristics.

The characteristics of our favorite songs

If we take Spotify into account, using this information, the platform develops weekly playlists that take into account the most loved songs by the user, songs that it imagines he may like, new releases in line with his tastes, and of songs played several times or of the past of which he repeats listening.

For example, every Monday midnight the music streaming giant processes over 100 million different playlists for its users, all fall under the name of New Discover Weekly but each is different from the others and is created specifically for the individual registered user to whom it is intended.

When we decide to listen to a particular song, we transmit through our choice of precise information, these are divided into characteristics associated with the songs that are taken into consideration in this way:

  • energy: measures how fast, loud, and noisy a track is;
  • acoustics: measures how much a track is acoustic compared to electronics;
  • instrumentality: measure how much sung and spoken is present inside the song (if the musical part exceeds 50%, the song will probably be classified as instrumental);
  • audience presence: the presence of the audience leads to classify the song as a live performance;
  • spoken: determines the distinction between a song and, for example, an audiobook; values between 33 and 66% are generally understood as indicators of the presence of both music and speech, values over 60-70% tend to be categorized as tracks that contain only spoken;
  • danceability: the predisposition to dance of a given song is evaluated on the basis of bpm (beats per minute) and rhythm;
  • valence: describes the positive attitudes (for example happiness) or negative (sadness) inherent in the song

The playlist we listen to is therefore the result of a very advanced technology, which is constantly updated and perfected and which requires huge investments in the face of, of course, very substantial revenues. And its results are extended not only to musical proposals but to a more general marketing speech that leads to create ad hoc commercials for each user depending on the profile that has been outlined through its online interactions.



Comments are closed.