When “Funkytown” comes on at a wedding, you can’t help but dance, right? What about Nelly’s “Hot in Herre” or “Bad Girls” by Donna Summer?
If your answer to any of those is no, you have defied computer science. (Also, I don’t want you at my party.) Those songs are among the most “danceable” number-one hits in the history of pop, according to new research from Columbia Business School and French business school INSEAD, using data from Billboard and audio-tech company Echonest.
Developed by students at the MIT Media Lab and owned by Spotify, Echonest uses digital processing technology to identify attributes of songs, such as valence, instrumentation, and key signature (pdf). The company created a proprietary algorithm to determine the “danceability” of a song based on its tempo and beat regularity. The calculation emphasizes the ability to dance throughout the whole song, so a bridge that even briefly changes the mood is highly penalized.
Researchers at Columbia and INSEAD then combined Echonest features with US Billboard rankings going back to 1958. Although they were able to calculate danceability for more than 90% of Billboard-ranked songs, Taylor Swift’s album 1989 was not available from Echonest at the time. Thus we may never know whether “Bad Blood” is more danceable than 50 Cent’s “In Da Club,” though my own personal investigation suggests it’s a dead heat.
The Most Danceable US Billboard Number Ones since 1958
The purpose of the research—published in the American Sociological Review and beautifully explained here by data scientist Colin Morris—was not to rank the most danceable mega-hits; it was to identify song features that could be predictive of mega-hits. Researchers found that top-ranked songs tended to have more difference from past hits than lower-ranked songs, defying the trope that popular songs are just copies of other popular songs.
Still, the optimal pop song should be only slightly off the beaten bath. Take Snoop Dogg’s number-one hit “Drop It Like It’s Hot.” It’s quite similar to much else on the charts, except for its distinguishing slow tempo.
If you want to experience computer science’s vision of danceability for yourself, try the playlist below. And if you want to experience mine, hold out for that party invite.