Publicis Music Project
In a company-wide survey, employees submitted their top four favorites bands or musicians. Results were revealed in a company-wide presentation advocating for Publicis to embrace network science.
WHAT AM I LOOKING AT?
A few weeks ago, the Search & Data Science team at Publicis conducted a company-wide survey that asked Publicis employees to list their top four favorite bands or musicians. Over 160 people completed the survey. We’ve compiled the results and visualized them in a network.
Why a network?
Music preferences are not random. Visualized as a network, the responses from the survey show that similar bands and musicians are clustered together. This shouldn’t be too surprising. “Homophily” (love of the same) is common in networks. Find a band or musician in the network that you like, look at everybody who is connected to it, and check out other bands or musicians those people are connected to. There’s a better-than-random chance that you will like those bands or musicians even if you have never heard of them. (Online music streaming services like Pandora and Spotify rely on this principle to make recommendations.)
One of the most interesting features of social networks is that people influence each other at a distance. In Connected, social scientists Nicholas Christakis and James Fowler show that a good predictor of your happiness is how happy your friends are and even how happy your friends’ friends are. They also found that if your friends’ friend is obese, there is a 25% chance that you’ll be obese. Our music preferences are similarly influenced not only by our closest social ties but also by people we haven’t even met.
Social networks can also help us identify the most connected nodes, or the nodes that would disrupt the network the most if they were removed from it. For this network, the person with the highest “betweenness centrality” score, which measures how central a node is to a network, is Zack Kurland. The opposite of Zack are people who are not connected to the network. If you look around the edges you’ll see a handful of people who do not share their top bands and musicians with anybody else. They either have unique taste or, due to the limited sample (n=164), happen to enjoy popular musicians that nobody else in the survey listed. For example, Katelyn Diekhaus is the only person who listed Carly Rae Jepson, one of the bestselling musicians of the past few years.
The goal of this project is to show how Publicis can use the science of networks to understand brands. Like music preferences, brand preferences are influenced by people in our immediate social circle but also by millions of people we’ve never met. We should think about our preferences not as a function of individual taste but as nodes embedded in a broader set of connections. The exciting implication is that network science, combined with data from surveys like this one, can give us new insights into which brands people prefer and why, perhaps revealing competitors and consumer profiles that traditional research could not detect. Let’s use this information to develop better strategy. Imagine a constantly updating database that uses network science to reveal brand preference—for example, the likelihood a person eats at Chipotle if they shop at The Gap, where they buy their clothes if they can’t get to The Gap, and all the competitors within any category.
Remember, the context in which people complete surveys is important. As you interpret these results, keep in mind the difference between what people report on a survey (stated preference) and what they actually listen to (revealed preference). It’s an important behavioral science principle that the Search & Data Science team relies on to build our tools, develop our methodologies, and write our decks.
One last thing. We’ve forwarded these results to the folks in charge of the awesome playlists that you hear on the 26th floor. Happy listening!