Song Quality vs. Popularity

Mohamed Kamal

CEO and Cofounder of Unbias

In our last post, I shared with an experiment conducted by Yahoo’s MusicLab about the complexities of success in music.

Today, I will delve into a follow-up study on the MusicLab conducted by MIT researchers in 2012, that investigated the dynamics of song popularity and quality. The researchers wanted to answer a simple question:

Can we separate popularity (influenced by social signals) and quality (inherent appeal) of a song to better understand what ultimately contributes to its success?

To answer this question they created a simple model to disentangle popularity from quality using a technique called the Polya-Urn process. This technique mimics the way people make choices based on the choices of others. For example, balls of different colors are placed in an urn, and the probability of drawing a specific color is proportional to the number of balls of that color already in the urn. This process mirrors the way individuals in the MusicLab experiment chose songs based on the download counts of those songs.

To disentangle the effects of quality and popularity on song downloads, the researchers assigned each song an “appeal” value, which represented the song’s inherent quality based on ranking, which can be subjective. In the model, the appeal of a song influenced its initial probability of being drawn from the urn. As people listened to and downloaded songs, the download counts (popularity) also affected the probability of future users choosing the songs.

Now let’s examine the case of “She Said” by Parker Theory. In the control group of the MusicLab experiment (no social influence), the song was the top choice. However, in the socially influenced groups (where people were able to view download counts for each song), it initially ranked lower but climbed up the charts over time.

The researchers’ model gave an explanation into why. The appeal of “She Said” contributed to its initial probability of being chosen in the Polya-Urn process. While the song’s quality played a role in its eventual success, the influence of social factors (download counts) also came into play. As more people downloaded the song, its visibility and popularity grew, eventually creating a snowball effect. It took a while, but it got there.

One of the most interesting findings from this follow-up study is the idea that social influence primarily affects what songs people choose to listen to first, not their decision to download. In today’s world it could be the decision for a consumer to listen or save a song to a playlist on Spotify.

This means, while people may listen to a song due to its popularity, their decision to save the song is primarily based on its quality. In other words, social influence can help a song gain visibility, but ultimately, its quality determines its long-term success.

Another insight from the 2012 study is the importance of song availability and positioning. The researchers found that the way songs were arranged on the platform, such as in a grid or a column, significantly affected the impact of social influence. Today’s world, this suggests that song position in user or editorial playlists really matters (we kind of knew that).

My key takeaways are: A) It’s essential to remember that quality remains a crucial determinant of long-term success, even if social factors can reduce initial exposure. B) Song/product availability and positioning is important to maximize the chances of success. Out of sight, out of mind holds true.

Unbias is building an AI product with marketing capabilities. Join our waitlist here to get early access.

recent posts