The Future of Music: APIs
Music is, without a doubt, an indispensable part of culture. Put differently, music serves as a renewable resource. For this reason, the music industry has become a behemoth, a giant indeed. However, giants are slow, groggy, and often sick—hence David’s triumph.
Last year, I woke up from a dream with a strange idea: create a data mining solution for musicians. I have posted the narration I made as soon as I woke up.
As you’ll hear, I wanted to create a place for musicians to receive insights about the popularity of their music, and the nature of their fan base. As a musician myself, I understood that musicians struggled with connecting with their fans, and making sense of the deluge of random events that occur along the journey.
The meta-narrative goes something like this: one second you’re jamming in your bedroom, and two weeks later you’re at Madison Square Garden. However, musicians know that the journey is much longer, less certain, and littered with surprises and mishaps. Fans shine brightly on the path, as gems that you discover—or better yet, discover you.
Two months ago, I revisited the dream/idea with my brother and close friend while in Miami. I found it surreal that we were discussing the dream I’d had. They both had an interesting insight: building a data mining solution would not be enough.
We would have to build a new way for fans to discover and support artists.
When I returned for my last semester in college, I began digging around for academic research around data mining and music sales. My thought was: can we predict how many albums an artist will sell based on her followers on Twitter? I chose this research area because I had spent the last few years doing machine learning research, most recently using Twitter data to gain insights about brand popularity. This line of thinking was also heavily influenced by the woefully incorrect predictions about J Cole’s 2014 Forest Hills Drive. It struck me that something was very wrong with sales predictions and, more broadly, decision-making in the music industry.
My first research foray brought me to a research paper by Professors Dhar and Chang, from NYU and USF respectively. Their paper “Does Chatter Matter? the Impact of User-Generated Content on Music Sales” found that “the volume of blog posts about an album is positively correlated with future sales”. I found this interestingly related to my earlier reading in Eric Siegel’s Predictive Analytics. Following this, I found a host of other papers assessing the possibility of predicting album sales based on social media data.
My second research foray was to find out if there were any other companies attempting to solve this problem. At this point, I was truly amazed. My first search directed me to Nielsen SoundScan, although I already knew about them. Nielsen is the main data provider for the music industry. Like the industry, it is a giant. Its main clients include Billboard magazine, major and independent labels, distribution companies, as well as artist managers. Nielsen primarily provides point-of-sale data from local stores.
More interestingly though, I discovered a company called Next Big Sound (NBS) which “provides a dashboard, charts, and reports to monitor popularity, activity, and metrics for musicians across social media, sales and events.” NBS seemed to provide everything I could have imagined. Looking at their team and technology stack revealed that they were deeply immersed in data gathering and analysis. They have positioned the company as “the #1 provider of data-driven artist recommendations for brands.” Indeed, I think helping brands is an interesting market segment to choose. Nevertheless, I believe that the data insights they are collecting are tremendously useful to artists, especially indie artists. Assuming NBS places its attention on brands, I believe that a gap will still exist, an undeserved market segment, so to speak. In short, there can and should be multiple players in this space, providing differentiated services to artists of all disciplines.
To be honest, I considered scrapping my research, hunkering down, and applying for a job at NBS. I still might…
The next company I discovered rocked my world. EchoNest is a Boston company that synthesizes two fundamental principles: 1) learning about music from its sonic properties 2) learning about music from peoples’ conversation about it. This allows Echo Nest to do what Pandora does in significantly less time. One of the company’s insights is this: “[T]he more you know about a community, the more you understand peoples’ preferences.” This echoed what Silvio Pietroluongo, VP of charts and data development at Billboard, said recently: “album sales … capture the initial impulse only, without indicating the depth of consumption thereafter.”
Two of EchoNest’s founders are MIT graduates with deep experience in music technology. The parallel between EchoNest and NBS is that they both have APIs (application program interfaces). In other words, they provide a way for other developers and companies to leverage the incredible amount of data they have developed to do compelling things. For instance, Paul Lamere built The Infinite Jukebox, which takes any song and plays it infinitely, after using EchoNest’s API to analyze the song.
After the last few weeks of market research, here is a summary of my current line of thought. The presence of APIs that make music streaming, analysis, and (re)synthesis possible means that really innovative products can be built. Paul Lamere has compiled a helpful list of the music APIs currently available.
To give you an example, building a music dashboard that allows a user to simply find a song by an artist, and play it infinitely is a really simple task. A developer would simply need to make use of APIs provided by EchoNest and Songkick. This approach to development interests me because it allows for quick, virtually free, and fascinating experimentation. Moreover, it also means that design becomes a major differentiating factor. I am borrowing here from Michael Tavani’s insight that the best companies will differentiate themselves based on the ability to craft a pleasant experience for users.
What’s next for me? My plan is to build a few experimental products with some of the APIs on Lamere’s list. In addition, I plan to collaborate with local musicians to determine what tools they need, and how their lives can be made easier by leveraging some of these novel technologies.