Music Streaming Analytics — Can Artists Be Data-Driven?

When I first received an invite to join Spotify back in 2009 it did more than just open up a library of on-demand music. I could finally discover new music without hitting a payment barrier.
It was a revolution for the way I enjoyed music and quickly replaced iTunes as my everyday music player. When Spotify began to monetise the platform I was more than happy to pay a subscription.
What did this mean for artists? For the first time, data could be gathered from a larger set of users. Not just from fans who were willing to pay directly for their music but from listeners discovering their work.
Artists could see how they were performing and who was listening to their music.
Spotify (arguably) suggests that this data could be used to guide Artists.
I’m going to review some of the metrics that are (or could be) available to artists on Spotify. I'll ask what they could do with the data to what extent an artist could be ‘data-driven’.
What’s available to Artists?

1. Streams
First things first, we can see how many times a track is played.
For some music streaming services, this translates into monetary value for the artist and will be a key measure of return for their investments.
The amount paid per stream to the artist depends on their arrangements. I’ve heard it could be a penny or even a third of a penny.
On Spotify, for a stream to be counted — the listener must not skip before 30 seconds.
2. Saves
This is a count of how many listeners add your track to their library or their playlist. In Spotify, this is done by clicking on a star icon on each track/album.
Although it does do a good job of telling you how many listeners really like your track there could be a proportion of ‘casual’ users who don’t use playlists or build a library.
There could also be an instance where your track is saved to a playlist where it does not belong or you would want to it to be listed.
3. Listeners
The number of individuals that have streamed a track.
Individualism is somewhat flawed with music analytics. Some businesses use one account to play music to their teams. Others will carry Bluetooth speakers in public places.
So, we can’t use this as a true measure of audience size. With music, you never will.
Nevertheless, it’s the best we have. Taking this forward, we can use this metric to create performance ratios such as:
- Repeat Rate (Streams / Listeners) — is this track catching on? Is this track played more frequently per listener than others? Should this track be in my encore?
- Save Rate (Saves / Listeners) — is there a track making its way into a users library more frequently than others?
If a track or a set of tracks have a higher repeat rate then the artist could potentially choose to produce similar tracks.
By looking at save and repeat rates we could start to quantify listeners apart from fans.
4. Source
Or, ‘where your streams come from’ breaks down your streams by where the user played it from. We have these ‘sources’ to review on Spotify:
- The Artists profile and track-list
- The listeners own playlist or library
- Other listeners playlists
- Spotify algorithmic playlists
- Spotify curated playlists
- ‘Other’
We can use these sources to assume at what point a listener is in their ‘streaming’ journey. For instance, if they played the track from the artist profile, then they are likely to be familiar with you enough to check for the latest releases or to find their favourite tracks.
The count of streams from Spotify playlists will be important to gauge if your track is actively being promoted on editorial playlists or platform generated playlists.
As these playlists are a source of discovery for listeners, it would be useful to run an analysis on streams originating from these playlists to see at what rate they convert a listener into a fan.
Spotify pushes the importance of playlist presence on their own materials so this would be important to know more about.
5. Demographics
We can split listeners by age, gender, country, and city. In basic terms, having an understanding of your audience can be useful for planning promotions or social media activity.
Besides that, it’s more or less a trivial understanding of who your listeners are. There may be some pleasant surprises or some stereotypes broken.
When it comes to planning a tour or a set-list, location data will no doubt be useful — especially for emerging artists.
How could we expand on these insights?
We have a lot of potential with the metrics available on our Spotify dashboard. In theory, these are additional metrics that could be added to the mix.
1. Average drop-out time (track)
Wouldn’t it be good to understand at what point in a track a listener hits skip?
Perhaps a track concept was too left field for some listeners or the track was too long.
Looking at drop out time would be the first step in using quantitative data to assess track performance.
Of course, there would be a lot of distortion around this metric. Users can like a track but not be in the right mood or place for it. They may need to stop their music for work or conversation.
This metric could prove to be unreliable and perhaps ‘too much information’.
2. Number of completed playbacks
As we now know, streams are counted beyond 30 seconds of playback. This means that if a track is skipped at 31 seconds it is still counted as a stream.
Without a measure of completed playbacks, we can’t gauge how many listeners actually listened to the track in its entirety.
Many, if not some, of our favourite tracks, will build up towards a chorus after 30 seconds. The peak of the track may be at an arbitrary section — perhaps at the end such as on classics as Bohemian Rhapsody or Paranoid Android.
As streams are the only volume perception we have of playback count, the artist is missing insight into whether their music was given a chance to be fully appreciated from start to finish.
3. Repeat rates per demographic
Is there a group of listeners repeating a track more than others? Currently, we can only breakdown demographics by listeners and not streams or saves.
If this information was available we could see which groups of people are more engaged with the music.
What does this all mean?

Music streaming analytics can give artists valuable insights into how widely their music is being circulated.
It offers access to demographic data that will show — from a birds-eye view — who enjoys their music and where they’re enjoying it from. This could be used to help promotions and even to plan tours.
The question of whether this could be taken any further as to guide creativity or anything more is less clear.
The data available from Spotify seems somewhat incomplete. There could be more in the way of data to help find engagement patterns and track performance.
Is this a deliberate tactic from Spotify?
Besides, music streaming behaviour is a little bit trickier to contemplate than say, website user behaviour, where a person would be trying to get from A to B at most times. Music is inherently linked to emotion which makes it particularly tough to assess with quantitative data.
Nothing can be taken away from the creative input of music. The trends found in data will be difficult to translate into real-world success. The relationship between artists and their audience is far too complex.