Understanding music sentiment by data mining YouTube

29 October 2013

Tom Jayne

Tom Jayne


In the last decade, the music industry has undergone far reaching and extremely public changes in how music reaches the consumer and, in turn, how revenue is generated. The success of Spotify and other subscription based music services highlight this change in revenue collection and represent a clear step forward in diversifying the business model of an industry which has been hugely affected by digitization.
However, more so then subscription streaming services, the biggest shift in music consumption has to be the rise of YouTube and what it offers – unobtrusive yet ad funded streaming of music – whether that be officially released music videos via VEVO or fan uploaded ‘audio only’ videos.

This represents a challenge for the music industry – in how to monetize this huge volume of YouTube streams (in the case of the infamous Gangnam Style, this number tops 1 billion). This remains a difficult question to answer but all these YouTube can be used for other means – and represents a new opportunity to explore, in real time, some of what these YouTube Views mean:

  • Comparing YouTube views with digital downloads. Both YouTube hits for a particular release and Digital Downloads will follow a sales curve in the time following their initial release. By comparing the curves of both, and overlaying the two into the same graphical representation, it will be possible to understand how the behaviour of the two differs.
  • Analyse true sentimentality via chatter versus hit ratio. YouTube views are an important measuring stick for understanding the total exposure any given artist or track is receiving at any time. However, they only tell on portion of the story – YouTube will also track two additional pieces of information – the ratio of likes/dislikes associated with any given video, and the total volume of comments received by any one video. This information can be used to provide a clearer picture on the true sentimentality towards any given video (likes as a % of hits), and of the total influence of the video (comments as a % of hits).
  • Trace geographical clustering of artist popularity. Whenever a video is viewed or liked, the geographical location of where this ‘like’ or view came from can be leveraged to build up a picture of where an artist is receiving the most exposure – which can have implications in how marketing spend is deployed, and in how tour dates and venues are determined. All of this information is already available when monitoring digital and physical sales, but the real time nature of YouTube views – and the ability to compare the distribution of YouTube activity and true sales – makes the information gleaned from YouTube more compelling then analysis of sales data alone.

Naturally, any combination of the above information can be compared and then visualised, offering a wealth of information to marketing and management teams which previously was either not available, or not available in a timely fashion. Any concerns with data privacy aside – I wonder – is anyone in the industry already doing this? And, if the answer is no, why not?

View comments


Blog post currently doesn't have any comments.

Bluefin and SAP S/4HANA - welcome to the one horse race

We use cookies to provide you with the best browsing experience. By continuing to use this site you agree to our use of cookies.