How does predictive modeling affect music consumption? Music has been present in people’s lives throughout history. It is a concept that goes beyond the tangible. Through music, emotions, feelings and experiences are shared universally. From the moment you are born, you are exposed to melodies, sounds and songs that accompany you throughout your life. Many of those songs are related to people, moments, memories and experiences. Music is linked to neuroscience since it gives us pleasure, and excites us or generates melancholy and sadness.
It is an art that also affects the nervous system, since through it, the brain secretes dopamine, a substance that generates from pleasant feelings to fear. Music is capable of bringing people together and stimulating neural synopses in the same way, leading to a common sentimental experience or connection.
Predictive modeling is the new revolution in the music consumption universe.
Internet, a new universe for music
The appearance of the Internet and the subsequent digital revolution have caused a great earthquake in the music industry. In the different areas of this musical ecosystem, great changes have been appreciated. All thanks to the transgressive injection of new technologies, where reality becomes more immediate for consumption through new digital streaming channels and online platforms such as YouTube, Spotify and SoundCloud. In addition, the appearance of artists who have promoted their own independent career outside of traditional labels is encouraged. This new digital scenario has also promoted total transparency in the industry, applying new methodologies such as blockchain, through data fragmentation and the extreme connection of information.
With rapid digitization, the music industry has moved towards a new way of measuring success, which is through data.
Algorithms and predictive modeling in music platforms
Digitization in the music industry has led to the creation of selected and segmented playlists using algorithms. Years ago, all playlists were programmed by human beings. However, today due to digitization there are other methods of selecting playlists that do not require any human involvement. These are automatically programmed playlists based on the large amount of data generated by the monitoring control mechanisms of the platforms. These automatically scheduled playlists are created through artificial intelligence-based algorithms that use big data.
For example, Spotify offers a series of personalized playlists selected through algorithms, as can be seen in the sections on “Weekly Discovery” or in “Your news radar”. These playlists consist of thirty songs updated weekly that are based on Spotify algorithms for audio analysis combined with the history of individual music listeners, connections with other users, number of likes, shares, and more. Automatic recommendation systems are by no means a novelty in the digital economy, and it has been confirmed that these mechanisms can have a high impact on the behavior of audiences. What sets Spotify apart from the competition is this powerful recommendation service, as it is managed through six companies based on machine learning applied to the digital music ecosystem.
Music platforms obviously have that access and have concluded that the ability to custom algorithmically curated music experiences for their listeners is key to achieving sustainable competitive advantage. Most access-based music platforms have invested heavily in playlist selection capabilities and there is a high demand for experts in music data analysis.
A hotly debated question is how algorithmically selected playlists affect listeners’ relationships with artists and music in general.
These curated playlists and personalized recommendations can reduce the cognitive load on music listeners by essentially eliminating the need to develop artist-fan relationships, which is why, once again, digitization is removing middlemen in the music industry
Algorithms effectively supplant the role of artist-based branding, and the public doesn’t need to remember artists for a satisfying music experience. Also, Spotify applies a natural language processing method, a key aspect in the development of the Semantic Web, since the algorithm analyzes what language users use to describe songs and relates them to other songs that are discussed in a similar way.
It also employs a convolutional neural network system, where Spotify uses a CNN-based model to analyze raw audio data with respect to the song’s BPM, key, volume, and other parameters. Spotify then searches for songs with similar parameters and recommends them to users. This model has proven to be exceptionally effective in discovering quality music that is yet to be recognized by the masses.
In 2018, Warner Music Group acquired Sodatone, a service that feeds streaming, social media and touring data into machine learning algorithms to identify which artists have the most potential for future success.
The beat of the future
The tool of predictive modeling algorithms can analyze the most popular songs in any region and then compare their attributes to any newly released song to identify potential for “hit”. This can become a secret weapon for today’s producers, songwriters, labels, and publishers as they can now tailor the sound to appeal to a specific target audience and fit in the new trends and sounds.
Through Artificial Intelligence and algorithms applied to consumption based on preferences, a “consumer tyranny” can be seen, a promising trend, as numerous companies such as BMG, Sony, Universal and Warner are producing content chosen and financed by fans.
The digital transformation has created a new scenario and as a consequence, new social behaviors adapted to it. It is a way of connecting with the world and understanding each other without the need to use other resources.