The issue with interpreting streaming analytics
Nowadays the streaming stats are the primarily data source that record labels use. It allows them to make inferences about the popularity of a specific genre of artists or music and make informed decisions with regard to what type of artists / music they should focus on. However, interpreting streaming analytics is challenging because of two reasons:
1. Streaming data only provides insight into behaviour and has a low explanatory power
Streaming data does provide some insight into the popularity of particular genres of music or types of artists on the basis of demographic information. They use metrics like number of times listened, popular countries, other artists the listener listened to. The data just provides insight into the popularity of particular music or artists (based on the sound of the music, song structures or lyrics) in certain audiences. However, this is a restricted view, because it does not answer the true why question. There may be specific underlying deep-seated motivations (e.g. values, lifestyle, attitudes, experiences) why something or someone is popular in a particular audience. Targeting that specific audience by just focusing on the sound of the music, song structures and lyrics leads to a limited understanding of what makes an artist or specific type of music appealing to a target group. By using these insights as a guideline, creative direction is restricted and it may lead to an investment in only particular types of artists. However, other creative choices or artists might also have been appealing to the target audience(s) and even be more profitable, when there was a better understanding of the target audience(s) beyond demographic data provided by streaming analytics.
2. Streaming data have issues associated with validity and reliability
Even if labels take the low explanatory power for granted, the demographic data that it provides is also blurred because of the ability to listen to music through curated playlists (such as the new releases playlist and the big hits playlist). This hinders obtaining a clear picture of the true popularity of a specific genre of artists or music and thus the collection of usable and informative data. A song of artist can be popular in a hit playlist, but actually be less popular than it appears to be. Characteristic for playlists is that people tend to not skip songs. As a result, in extreme cases the popularity may be primarily based on the songs/artists that come before and after in the playlist rather than the song/artist itself. Moreover, due to playlists, the demographic stats that are supposed to give more insight into the characteristics of the audience in which the song/artist is popular, are also blurred. This is because they reflect a lot of mainstream listeners that got to the artist/song through the playlist, listeners that actually do not belong to the core audience(s). As a result no clear insight can be given in the core audience(s), which are actually the most relevant for labels in terms of revenue.
Both the limited explanatory power of streaming metrics as well as the contextual effects of curated playlists restrict the collection of accurate data and as a result the extraction of insightful findings. Labels are now chasing the streaming metrics and this is leading to a data feedback loop. By looking at the hit playlists and seeing what is doing best, labels push more of that into the marketplace. The streaming insights influence the creative development of artists as well as the types of new artists signed, and in the end also the marketing campaigns and communication strategies. This in turn creates the illusion of that genre of artists / music becoming even more prevalent and the process starts over again. The fact that in recent years hit playlists have become more and more homogenous in sound proves the occurance of a data feedback loop. However, the data feedback loop due to a focus on ‘tracks’ poses several negative consequences for the music industry.