I took on an interesting project with a Pitchfork dataset I found on Kaggle. 

Pitchfork is one of the most popular music reviews sites in the world, they were founded in 1999 and I must confess, they are known for being a bit unconventional.

Anyways, the dataset came in form of six tables and in SQLite format having 18, 393 reviews in total. I’ll skip the rest of the technical stuff till the bottom of this post, so I don’t bore core music readers.

First thing of note is: I limited my analysis to Rap, Rock and Jazz, because I felt it would be clearer to compare 3 genres than, say, 7.

I attempted to answer a few questions, let’s dive right in.

1. How were reviews distributed across the 3 genres?

So, I plotted a pie chart to show how the reviews were distributed across the 3 genres.

We can see that Rock had the highest number of reviews, followed by Rap, and then Jazz. Are there more Rock albums released yearly than Rap and Jazz, or are Pitchfork writers simply more interested in Rock than the other two genres? At this point, we can’t really tell. Let’s move on.

2. How have the 3 genres performed when you compare ratings side-by-side?

Of course, we all want to know. So, I plotted a nice line graph to show the average rating for each genre per year, from the 1960s to 2016.

Now, you’re probably wondering how we have reviews from the 60s when Pitchfork started in 1999. Ah, right. They actually went back to review albums that dropped before they started.

So, yeah, the graph:

We can see a few things:

• The earliest reviewed genre was Jazz, the Rock, and Rap came much later (in 1982, to be precise).
• Generally, the 3 genres seemed to have gotten worse over the years, as we can see a downward trend in the average ratings.
• Rap started low, then peaked at a perfect score, before it started trending downwards like the others.

To add more context, I added the number of albums reviewed per year, per genre, to the graph:

We can see that much fewer albums were reviewed for the early years, and as more albums got reviewed, the averages began reducing. This is logical. We can also see that Rap seems to be the most commonly reviewed genre in recent years.

Sidenote: the earliest-reviewed Rap album was: The Message by Grandmaster Flash & The Furious Five.

3. What are the best performing albums, and artists for each genre?

Okay, this part is exciting!

Rap

Best performing artists:

So, there they are…

We can see that Kendrick Lamar and Kanye West stand out here, as they have the most albums reviewed in the top 15 (8 and 12 respectively). Can we infer that they’re the most consistent Rap artists of all time (at least, according to Pitchfork)?

Highest rated albums:

• We can see 8 Rap albums with perfect (10/10) scores.
• We can also see that Kanye West has 3 albums in the top 20. Legend!

Jazz

Best performing artists:

We can see Miles Davis standing out here, with the most reviewed albums by far and still landing in the top 15.

Highest rated albums:

• Miles Davis has 5 albums in the top 20. Wow. Is he that great, or someone at Pitchfork is just a stan? Lol.

• There are only 3 albums with perfect 10/10 ratings. Does this mean it’s harder for a Jazz album to be perfect, than it is for Rap?

Sidenote: I wondered why John Coltrane and Nina Simone’s top albums were so recent, so I asked Google and learnt that the albums were re-released recently. So, yeah, we’re good.

Rock

Best performing artists:

The Velvet Underground seems to be the standout band here, with 13 albums and still cracking the top 5.

Highest rated albums:

We can see something very interesting here! All the top 20 albums have a perfect rating of 10/10. What does this mean?

Let’s not forget: in terms of volume of reviewed albums, Rock was by far the highest with over 80%. This, of course, makes it logical to have a lot of perfect albums here.

Interesting stuff, right? I proceeded to ask some more interesting questions.

4. Is there a relationship between the length of a review and the final rating it gives?

So, I plotted the number of characters (letters, numbers, spaces, symbols) in each review against the average rating for that character length.

We can see that most reviews have character lengths between 2000 and 12 000. And, yes, as the character length increases, the review score tends to get higher.

The longest review was for Ice Cube’s Death Certificate with over 22 000 characters and a 9.5 rating.

I also found an album with a 0.0 rating (weird!): Liz Phair’s self-titled album with a little under 6000 characters. But a 0.0 rating… Seriously?!

5. Who are Pitchfork’s top reviewers, what are their roles, and what arethe average ratings they give?

We can see that Ian Cohen has reviewed the most Rap, Rock and Jazz albums, and he’s also the hardest critic to impress, as his average rating awarded is 6.3, the lowest among the top ten reviewers. There doesn’t really seem to be any definite connection between role, volume of reviews, and average score.

Well, after these, I had to get some sleep. I hope you guys enjoyed this little project. Analytics has the potential to open our eyes to insights that aren’t obvious on the surface, and I believe it will play a huge role in the future.

Technical stuff

The data came in 6 tables in SQLite so I used Alteryx to clean, understand the data and do some preliminary operations like counting the number of characters for each review, then I exported each table as sheets in one Excel File.

I did all filtering and visualizations in Tableau.

Please, feel free to leave your comments below, and clap for us!