Behind the scenes there is a lot more going on than just songs getting shorter. I'll focus on Spotify as it's the firm I'm most familiar with. The article briefly touches on the hook being moved to the beginning, so I'll skip that part - much the same thing is happening on streaming platforms for movies/tv shows, where analytics are used to determine the ordering and features of a piece of content to maximise engagement/minimise churn - at the expense of directorial freedom.
There are 2 other significant factors at play. The first is that not all streams are created equal. It is relatively well known that the large studios have cut deals on payouts per stream, eg. $0.0029 vs $0.002, depending on the artist's following, age of the album, etc. But, a lesser known fact is that not all streams are created equal in a number of countries. Songs played is sequence on an album, or as part of a playlist, may attract a lesser premium than individual plays, and radio broadcast format pays less still. How it works depends on the jurisdiction/contract. With this in mind, consider how you listen to songs on Spotify. You may search for songs and albums, but a large proportion of the music you hear probably comes from autoplay, where Spotify matches what it 'thinks' you might like to hear.
With the above in mind, Spotify has been acquiring podcast companies and building them into their client for several years. There is undoubtedly an audience for podcasts, but the business case is that they do not have any of the licensing legacy issues that music has, where Spotify gets to keep a bigger piece of the pie.
The point in the article - that 'songs are getting shorter, albums are getting longer, and artists are collaborating across genres' is mostly a result of artists producing to satisfy a recommendation algorithm that balances retaining listeners, whilst minimising licensing costs.
The collaborating across genres is also explainable; several years ago, Spotify bought a company called Echonest (amongst others), which was essentially a large graph database mapping songs and their musical attributes (genres), onto which a recommendation engine was built - it's worth looking up their whitepapers on matrix factorisation at scale if they still exist. The corrolary though, is that similar listening habits could be used to cross-recommend songs with a good degree of accuracy.
There are also a smattering of other considerations to take into account - a lot of the large playlists not operated by Spotify are pay to play - if you're a new artist, you can pay to feature on playlists for a given genre. There is no official channel for doing this, but the practice is widespread. These playlists make more money if they have 50 x 3 minute songs, than 30 x 5 minute songs.
I'm sure you can see where this is going - a cost/engagement optimisation engine, along with a recommendation engine where artist remuneration is more or less directly proportional to the number of people that are happy to listen to it if it is played for them.
The artist is rewarded if they produce a larger number of songs, suitably homogenised to appeal to the mass market and sound familiar to what they already listen to, and to hit as many of the genres possible to maximise inclusion in recommendations. Unless you are one of the few exceptional artists, you are going to be punished for creativity vs. playing it safe.
There's no easy fix for this, assuming it needs fixing. All media streaming platforms in the long run chase the marginal user, who by and large wants average content almost by definition.
Tweaking the remuneration model would be a good starting point, where instead of Spotify taking eg. 30% of the subscription, and paying $0.0018-0.0032 per stream depending on artist, if instead the model worked as Spotify taking 30%, and each artist payout for a user's streams was calculated as ($10 * 0.7) * (number of plays by user of artist songs / total plays in given period), one would likely see artists serving niches or esoteric music to be better remunerated, albeit at the expense of the studios managing household name artists.
There are 2 other significant factors at play. The first is that not all streams are created equal. It is relatively well known that the large studios have cut deals on payouts per stream, eg. $0.0029 vs $0.002, depending on the artist's following, age of the album, etc. But, a lesser known fact is that not all streams are created equal in a number of countries. Songs played is sequence on an album, or as part of a playlist, may attract a lesser premium than individual plays, and radio broadcast format pays less still. How it works depends on the jurisdiction/contract. With this in mind, consider how you listen to songs on Spotify. You may search for songs and albums, but a large proportion of the music you hear probably comes from autoplay, where Spotify matches what it 'thinks' you might like to hear.
With the above in mind, Spotify has been acquiring podcast companies and building them into their client for several years. There is undoubtedly an audience for podcasts, but the business case is that they do not have any of the licensing legacy issues that music has, where Spotify gets to keep a bigger piece of the pie.
The point in the article - that 'songs are getting shorter, albums are getting longer, and artists are collaborating across genres' is mostly a result of artists producing to satisfy a recommendation algorithm that balances retaining listeners, whilst minimising licensing costs.
The collaborating across genres is also explainable; several years ago, Spotify bought a company called Echonest (amongst others), which was essentially a large graph database mapping songs and their musical attributes (genres), onto which a recommendation engine was built - it's worth looking up their whitepapers on matrix factorisation at scale if they still exist. The corrolary though, is that similar listening habits could be used to cross-recommend songs with a good degree of accuracy.
There are also a smattering of other considerations to take into account - a lot of the large playlists not operated by Spotify are pay to play - if you're a new artist, you can pay to feature on playlists for a given genre. There is no official channel for doing this, but the practice is widespread. These playlists make more money if they have 50 x 3 minute songs, than 30 x 5 minute songs.
I'm sure you can see where this is going - a cost/engagement optimisation engine, along with a recommendation engine where artist remuneration is more or less directly proportional to the number of people that are happy to listen to it if it is played for them.
The artist is rewarded if they produce a larger number of songs, suitably homogenised to appeal to the mass market and sound familiar to what they already listen to, and to hit as many of the genres possible to maximise inclusion in recommendations. Unless you are one of the few exceptional artists, you are going to be punished for creativity vs. playing it safe.
There's no easy fix for this, assuming it needs fixing. All media streaming platforms in the long run chase the marginal user, who by and large wants average content almost by definition.
Tweaking the remuneration model would be a good starting point, where instead of Spotify taking eg. 30% of the subscription, and paying $0.0018-0.0032 per stream depending on artist, if instead the model worked as Spotify taking 30%, and each artist payout for a user's streams was calculated as ($10 * 0.7) * (number of plays by user of artist songs / total plays in given period), one would likely see artists serving niches or esoteric music to be better remunerated, albeit at the expense of the studios managing household name artists.