Rutgers students Shaun Ellis and Thomas Engelhardt tried to discover the secrets sauce behind a “hit” song by analyzing 4,200+ songs that made it to the top ten of Billboard’s Hot-100 chart. Using the echonest api, they took a look at tempo, duration, time signature, key, and abstracts like “energy” and “danceability”.  The main results are laid out here.

The fun part is, they made the whole data set available in Tableau for us to play with (download Tableau reader (free) and the dataset). Using the filters, you can answer bizarre questions such as how many hits in 1979 were on the charts more than 20 weeks that were recorded in the key of C (answer: 7). Or you can look at the data over time to discover all kinds of interesting long-term trends:

A scatter of all of the songs illustrates that the average tempo is 120 BPM.

Finally proven mathematically, songs of the 1980s consistently had the highest “danceability” (suck it, 90s!):

Hits are getting longer in length:

You can document the much maligned increase in loudness after the introduction of the CD, though it is also part of a longer-term trend.

If you have the full version of Tableau, you can design your own charts (but you probably don’t, because Tableau is too damn expensive).

Update: There are a few additional Tableau visualizations of this data available (that don’t require you to install anything) that are also quite interesting.