What can Data Analytics tell us about Flower Boy?

There are already countless blog posts and videos answering this question from a musical perspective and I wanted to add on to this topic by taking an analytical approach.

I think this is the right timing too as he will be releasing a new album tomorrow, and who knows maybe with a completely different style.

All the analysis was made on Python 3 and I will publish my code on GitHub once I can clean it.

Warning : some content in this post might not be appropriate for everyone especially in the text-analysis section.

Getting and cleaning the dataFor this project I combined data acquired through both Spotify and Genius.

Many thanks to the author of this post who shows an easy to understand and fast to implement way of calling the Spotify API and who saved me a huge amount of time.

Unfortunately as Bastard was not on Spotify, I could not get data about this album so I excluded it from the analysis.

For each songs, Spotify gives a fair amount of information and the one that interested me the most here were the audio features.

There are objects that define a song characteristics such as the acousticness, the energy, the danceability, the instrumentalness and so on.

I highly recommend you to read the Spotify documentation, that you can find here, to have a precise definition of each objects.

Genius allowed me to get the lyrics of each songs so that text analysis could be performed.

Spotify data were already in an easy to analyse format and I just had to convert them into a data frame.

Lyrics, though, needed to be cleaned.

To do so I removed all the punctuation and the stopwords with the NLTK library, put everything in lowercase and also removed expressions that are heavily used in Tyler’s lyrics without adding much meaning such as: “yeah”,”na” or “um”.

Original data from GeniusCleaned dataNow to the fun part !Analysing audio featuresI firstly had a look on the mean for each audio features and for each albums to have a general idea.

Among all available features, Flower Boy is clearly standing out from the other albums from one aspect : acousticness.

This is defined by Spotify as “A confidence measure from 0.

0 to 1.

0 of whether the track is acoustic.


0 represents high confidence the track is acoustic”.

The song that scored the highest on this scale is “Boredom” followed by “Where This Flower Blooms” and “Goblin”.

If we have a look on the distribution of this feature we can see that while other albums are skewed and have a long right tail, Flower Boy’s distribution is more even, showing a greater diversity.

To further test this hypothesis I ran a one-sample t-test with a 95% confidence level.

The null-hypothesis being here that Flower Boy’s mean for acousticness is similar to the population means.

The p-value found was 0.

039 which shows strong evidence that we can reject the null hypothesis.

However, due to the relative small size of the population we must be careful about this result.

In my opinion Flower Boy scores the highest in acousticness because of the use of: sampled acoustic drums (Boredom, Foreword), classical-like string arrangement (See You Again, Where This Flower Bloom), clean electrical guitar (Garden Shed, Boredom, Glitter) and piano (See you Again, Sometimes).

These elements were of course present in Tyler’s previous work but to a lesser extent.

He even made an acoustic version of some Flower Boy’s songs in his Tiny Desk session held in December 2017.

So treat yourself and listen to the following video while reading the rest of this article !Another comment often made in reviews has to do with the general coherence of Flower Boy.

Indeed, Tyler’s work has often been criticised for its lack of unity and for being “messy” as mentioned Pitchfork.

In the past, Tyler’s albums have been bloated and messy.

Flower Boy is 17-minutes shorter than the average Tyler album with more understated transitions and less disorder and chaos.

The composer and music theorist Arnold Schoenberg wrote that musical coherence is produced through the repeat use of a motive that is at the centre of a song and while keeping a relationship with other motives.

Unfortunately, Spotify’s data do not allow us to study the coherence within a song.

We can however have a look on the change in values for each features between each songs to have a general idea of coherence.

After having calculated the standard deviation for each features (which states how far a set of values is from the mean), I found that Flower Boy has the lowest values for both tempo and energy implying that the difference in tempo and energy between each songs and the mean is lower.

By making a graph line we can see that the change in tempo between Flower Boy’s song is less dramatic than others except for the last two songs.

Flower Boy seems therefore more constant in terms of tempo and energy change.

Text Data AnalysisDespite having the lowest average of words per song, Flower Boy has the highest lyrical richness defined here by the number of unique words to the total number of words, followed closely by Cherry Bomb.

If we have a look on the most used words, we can see that as opposed to Tyler ‘s previous album, Flower Boy’s most used word (which is “time”, used 71 times and 50 times in “Boredemon” alone) is not a swear word.

In fact, as Tyler grew older the share of swear words in his lyrics decreased from 10% in Goblin to 4% in Flower Boy.

The share of swear words are represented in brownThe wordcloud below is interesting to highlight the most prominent themes of Flower Boy.

In this album, Tyler talks a lot about his feeling, he gives us a glimpse into his daily daydreams and we can see it through some of his most used words (“feel”,”boredom”,”lonely”,”sick” etc.


The sharp contrast in lyrics is easy to spot when comparing Flower Boy’s wordcloud with one of his previous work.

Word cloud for Flower BoyWord cloud for GoblinFinally I used NLTK’s vader function to determine the general positiveness or negativeness of a song based on its lyrics.

In this scale, the scores are between -1, being the most negative, and 1 being the most positive.

Overall, Tyler’s songs tend to have a negative meaning and Flower Boy seems more positive.

Again, it is hard to state the significance of this number based on our small population.

The higher average score of Flower Boy might have to do with the fact that it contains less swear words and that some of the themes being addressed are less ‘dark’ than his previous work.

ConclusionBased on my analysis, Flower Boy differs from Tyler’s past albums by:being more “acoustic”being more coherent and constant in terms of tempo and energy change between each songshaving a higher level of lyrical richness and somewhat positivityWhether it is his best album is a subjective matter but I think it is his most distinct album so far and those are proof that he gained in maturity.

I hope you enjoyed reading this post as much as I enjoyed working on this little project.

I would love to have some feedbacks and get better at this so please don’t hesitate to comment and challenge my findings !.. More details

Leave a Reply