Visualizing Artist Collaborations in the Billboard Top 10 Songs

Preface: I highly recommend taking a look at the HTML renderings of the charts contained in this post. You can also see all the charts in one place if you don’t want to read. For the force-directed graph especially, you can examine the relationships in the data more closely. You can see them here: http://steventang.github.io/billboard-top-collaborations/

I took a glance at the Billboard Top 10 charts recently and was surprised to find so many songs where artists had collaborated, i.e. where an artist features one or more other artists (e.g. Lean On by Major Lazer & DJ Snake Featuring M0). I became curious as to whether artists have started collaborating more in recent years so I set out to find the answer. Along the way, I took a look at a few other aspects of the dataset for fun and drew some charts for your viewing pleasure.

Let’s kick it off with a simple one before getting into any analysis:

This one is pretty straightforward, the proportion of spots in the Top 10 every year that are taken by collaborations.

The Dataset

Billboard’s datasets aren’t publicly available. Billboard offers “Research Services” for $200/hr if you need to get neatly presented data delivered to you. That seemed a bit steep. Luckily, I discovered that Billboard keeps a complete history of weekly Hot 100 charts at a distinct, easily guessable URL (for example, check out http://www.billboard.com/charts/hot-100/1990-01-06). I wrote a simple web parser to scrape the data from Billboard’s site, holding true to my philosophy that data collection should always be performed frugally.

For this analysis, I decided to use only the Top 10 songs every week, starting from January 6, 1990 and ending on August 22, 2015. I only wanted to examine collaborations in the most popular songs. With this in mind, I won’t make any sweeping generalizations about collaborations as a whole, limiting our scope to the most popular collaborations.

It wasn’t trivial to determine whether or not a song should count as collaboration; checking over 13,000 songs requires a bit of work. I decided to split each whole artist string by key symbols such as “&”, “,”, and “Featuring”, then did a few edits for when this heuristic failed. I’m fairly confident that I accurately determined whether a song is collaborative for the vast majority of cases, but I would be surprised if nothing slipped through at all. Feel free to check the work here.

An Upward Trend

Back to the chart above — the percent of collaborative songs per year has shown an upward trend since 1990. The boom in collaborations starts in 2001 and reaches its peak at 2003 before leveling off for the next decade. So no, artists haven’t been collaborating more than they have been in the recent past. Collaborations have been in vogue for the last 10 years. In fact, 2015 so far has seen a sharp dip in the proportion of collaborations in the Top 10.

Things aren’t too different for Billboard’s weekly #1 songs:

A scatterplot of the number of collaborative songs every week reveals a few interesting tidbits:

Most notably, there’s an entire decade, from 2002–2012, in which not a single week went by without a collaborative song in the Top 10.

Fact:

Throughout the period of Jan 1990 — Aug 2015, 23.59% of all unique songs in the Top 10 were collaborations.

That isn’t to say 23.59% of all spots in the weekly Top 10 are taken by collaborative songs, but rather that amongst all of the distinct songs to reach the Top 10 in that period, 23.59% were collaborative.

Here’s a more difficult question to answer: why did collaborations increase in the early 2000's? We can’t say that it’s because artists realized collaborations are valuable, or because listener’s shifted their tastes. Perhaps the artists who rose to popularity in the 2000's come from genres that tend to favor collaborations. Much more data is required to explain confounding variables here, but my best guess is that the rise of Hip Hop contributed greatly to the rise in collaborations. Hip Hop is a genre that makes heavy use of collaborations and it rose into the pop music mainstream in the late 1990's and early 2000's.

The next chart somewhat affirms this belief:

The Most Collaborative Artists

The chart above shows the artists who’ve spent the most weeks in the Top 10 with collaborative songs. I’ve also delineated how many weeks each artist had a song in which they’re credited as the lead artist. Boss Lady J.Lo ain’t getting featured on nobody’s tunes.

Similarly, we can examine how many unique songs each artist has placed in the Top 10:

The top collaborators are largely Hip Hop and R&B artists. Did the success of these artists give rise to the popularity of collaborations or did they just follow the secular rise in multi-artist hits? Unfortunately I don’t think I have enough data to answer this right now.

How artists are connected

I’ll define an artist as “connected” to another if they’ve collaborated on a song together. The graph below shows how artists who have been a part of the most distinct collaborative songs in the Top 10 are connected. The first graph shows the connections between the 100 top artists (For a more fun version of this graph, click here).

The width of the lines represents how many distinct songs each pair of artists has collaborated on. Lil’ Wayne and Drake, with 5 songs in the Top 10, are the two most connected artists.

Admittedly, the graph above contains perhaps an excess of information. Let’s shrink it down to the top 20 artists.

You can find the interactive HTML versions of these graphs here.

You can find the data here. If you have any ideas about what interesting insights could be pulled from this dataset, let me know or use it yourself!