September 19, 2017

How the music broadcasting industry uses predictive analytics to spot rising and falling stars

Our client, a market leading music entertainment business,wanted help to know which tracks to present to its listeners, to ensure the maximum engagement,

With so many tracks to choose from, there are several factors that come into play, such as:

  • How popular the artist is with the audience, and how is that popularity trending?
  • Internal music experts’ judgement on how popular a particular track will be
  • Previous record sales

To aid our data science team we had qualitative and quantitative data made available in the form of surveys, and real listener opinions recorded on an app.

The survey data was rich, and gave a real insight into what people think about a track, but it’s expensive, and often not collected frequently enough to give the information needed to be able to act on, when tastes are changing so quickly.

The quantitative data came in the form of an online music app, where users’ real actions can be measured and, with statistical modelling, be used to predict the outcomes of the survey. The benefits of this are:

  • The ranking can be updated much more frequently to make the most of up-to-date listener behaviour
  • Tracking the ranking over a short period of time, you can see whether a track is becoming more or less popular, thus helping discover rising and falling stars.

The deliverable for our music broadcasting client has been an ongoing predictive algorithm, that spots every week trends in terms of which of some 300 tracks are becoming more or less popular.

This is then used by the D.J.s to help guide their selections, bring on new artists, and start to lay off those at the top as soon as the audience interest is starting to flag.

In this fictitious example, the current Taylor Swift track is one of the most popular (ranked number 3) but people are starting to turn it off – it could be time to give Louis Tomlinson a bit more air time?