Knows You Better Than Yourself: How Music Platforms Recommend Songs?
It’s no secret that over the years social media platforms in general have become more and more appealing, attractive, and overall doing everything it takes to lure us in. One quite obvious way is to send the users content the app can be certain the user would like. In the case of music platforms, it can be individually-tailored playlists, recommendation for next songs, and even find similarities between users' tastes.
I was surprised to learn how people are predictable when it comes to personal taste. Our tendencies show listening to similar music as our friends and family, and our individual taste is quite narrow. And unless your cousin offered you listened to ‘Style’ by Taylor Swift or it falls into your taste in music, you’d probably stick to Kanye West.
As an amateur data scientist, I did a project to predict the drinking habits of 60,000 OKcupid users with 72% accuracy only by looking at users profiles. All I had was the information users provided after downloading the app — their age, their occupation, some lifestyle habits… And remember, polling’s accuracy is oftentimes lower than realtime accuracy because of biases and users not willing to show the full truth in their profiles.
In another project, I predicted with 93% accuracy whether a person is obese, overweight, normal weight or underweight based on lifestyle — main mode of transformation, exercising, eating habits, etc. I also wrote about my findings here.
If humankind is so predictable, why not use it for business?
When a big corporation sends us adverts, be sure that this add passed through the eyes of many workers, aiming to make the ad reach the right audience, at the right time, with the optimal design, words, and colors needed to lure you to do one thing — pay the company for their product.
In the last quarter of 2023, Google alone had a record-breaking advertising revenue of 65.52 billion$. Their algorithms work, because you are much more predictable than you think.
Back to the music
Songs released nowadays get much higher popularity than older songs, probably for 2 reasons:
- Old music is not everyone’s taste.
- They don’t harness the qualities the new musicians found.
The new music industry has learned to create music like adds, entailored to their audience. According to my project — music released in the 2000s are more energetic than ever, louder, more danceable, and include more explicit content. Also, the acousticness is lower than ever and songs nowadays are about 25% shorter than songs from the 80s and 90s.
One interesting fact I learned is that the levels of speechiness is on the rise. With a constant decline from the 1930s to the 1960s, a minor rise from the 1980s and a reverse in the 2000s, speech and lyrics become more important over time. Even if your grandparants don’t think so.
As an amateur data scientist and amateur pianist, I liked seeing how there is not much change in the key of songs.
Predicting the future
Assuming the trends remain relevant in the near future, you will see more explicit content in shorter songs, with more place for words, energy and danceability. While it might seem narrower, shallow and might even be concerning for some, music shows changes and adaptation all the time.
Songs both nowadays and in pre-WW2 times were about as danceable, though a drop occured until the 60s. Also, songs today are about the same length as the ending of the 60s and the beginning and the 70s.
As long as music finds its audience, it would be appealing to listeners. Or else, you can always listen to past-decades music.
