Is There a Formula to Making a Popular Song?
What makes a song popular? How much do beats per minute, rhythm, and stability contribute to how a song is received? I would assume that songs with a higher energy and that are easier to dance to would be more popular songs. This would prove that there is a specific checklist that songwriters could follow in order to ensure that the song would be more popular. Looking at the most played 100 songs from Spotify from 2017 and 2018 can help answer some of these questions.
Both of the data sets were found on kaggle.com and were created by Nadin Tamer. The data looks at thirteen different aspects of the top 100 songs from 2017 and 2018. In this analysis we focus on danceability, energy, and tempo. Danceability “describes how suitable a track is for dancing based off of a combination of musical elements.” (Tamer) Energy is a measure of “intensity and activity” where tracks with high energy “feel fast, loud, and noisy.” (Tamer) Tempo is the measure of beats per minute in the track and is taken into account when calculating the energy of a song. All of these are measured on a scale from 0 to 1 with 0 being low and 1 being high.
In 2017, the average danceability of a song was 0.6968 with an average energy of 0.6607. From 2017 to 2018, the average danceability increased to 0.7165 while the average energy decreased to 0.6591. When it comes to dancing, there appears to be a slight negative correlation as the danceability level decreases with the rank of the song.
This correlation may only appear this way due to the outliers of the dataset. There are a couple songs with a very low danceability score that are ranked closer to the bottom 100 songs which may be causing the negative correlation. The following charts show the top 10 danceable songs from 2017 and 2018 along with their rank and the energy of the song.
The same type of situation appears when we look at energy correlated to rank as well. However, energy appears to have a much wider random distribution than danceability.
The line of best fit for these two graphs is formed from the outliers so we cannot draw a conclusion between the amount of energy a song has and how popular the song becomes. Energy is much closer correlated with the loudness of a song than it is with the tempo of the song. If a song is more energetic and loud, this doesn’t necessarily mean the song has a higher BpM. The following graph shows the difference in correlation between loudness and energy and tempo and energy.
What happens to the graphs when we specifically look at the tempo of a song? Tempo is used when calculating energy, but so is loudness and noisiness and it is possible that those extra measures are spreading the data out. Solely looking at the beats per minute of the top 100 songs may give us an answer to the formula for a perfect song.
The average tempo stayed almost exactly the same between 2017 and 2018, hovering at about 119 beats per minute. Even though the average bpm stayed the same, the correlation between the popularity of the song and the tempo has gotten stronger between 2017 and 2018
In 2017 the bpm seemed more random and from 2017 to 2018 it appears to be more grouped. Granted, the distribution is still pretty dispersed but it seems like more popular songs tend to have a BpM that averages between 110 and 120.
Due to the lack of apparent correlation, I ran a dimensionality reduction on the data to try and reduce the amount of random variables to see some similarities in the categories. To accomplish this, I combined the two data sets into one and ran the reduction in two dimensions which resulted in the graph below.
Most of the songs are clumped in this reduction but the really interesting aspects lie in the outliers of this graph. How can these songs be so different from the others but still be in the 100 most played songs from the last two years? Some of these outliers include songs such as Lovely by Billie Eilish, Perfect Duet by Ed Sheeran, and How Far I’ll Go by Alessia Cara. It does not seem that any of these categories have a strong correlation to each other.
Going into this, I expected a strong correlation between danceability, energy, and tempo. I was quite surprised to see that there was basically no correlation, not only between these categories but also between these and the ranks of songs. One thing the data does not look at is the popularity of the artist before the year. The top played song of 2018 was God’s Plan by Drake who also appears three other times in the top 100 songs of the year. Drake has been famous for years so when he releases new music, he doesn’t need to try and find people to listen to his album. His notoriety helps his music become popular faster. Maybe this conclusion is for the best, if there is not a specific formula that artists could follow then they have to continue to make music for themselves. If this perfect-song formula were to actually exist, we would lose out on so many incredible pieces of music because musicians and record labels would want to produce what could be one of the top songs of the year. Popularity does not always matter when it comes to producing music as an art and the lack of structure for this particular art form helps make it that much more impressive.
