It’s the fifth edition of Laughing Stock. Sam Hartford (@samhartford) takes a look at how streaming service machine learning is becoming the irreversible new face of music consumption and why we should care.
“Just as divine authority was legitimised by religious mythologies, and human authority was legitimised by humanist ideologies, so high-tech gurus and Silicon Valley prophets are creating a new universal narrative that legitimises the authority of algorithms and Big Data”.
So writes best-selling author and proclaimed guru of popular science Noah Nuval Harrari. These hyper prevalent mathematical equations are increasingly fundamental to the running of our daily lives, from social media or purchasing habits to medical diagnosis and government policy. Their grip (or that of those controlling them) and reach will only become stronger as humanity wilfully nose dives into a new age of dataism. You only have to watch something like The Social Dilemma to see the warning signs, as the Frankensteins of big tech tell us the monster is under their control. It’s not an entirely fatal path; on a personal level algorithms can be a force for discovery and refinement but they also cast an increasingly cold and calculated cloud over our cultural choices. Yet convenience often trumps concern and cognitive dissidence reigns supreme. How this exchange is deployed in our audio consumption has big ramifications for the future of music. It forces us to ask the question, should we embrace or decry it’s evolution?
Spotify will be our case study. Their 345 million worldwide user base, 155m of which are paid subscribers, dwarfs that of it’s direct competitors at Apple, Amazon or Deezer and the company has just posted total revenues of $2.17bn, up 17% year-on-year. It’s algorithm driven discovery sections and personalised end of year Wrapped experience also represents the most comprehensive and fertile ground to inspect. Those of us who use it will be familiar with the Discover Weekly & Release Radar functions that drop every week, with a plethora of unearthed or new tracks and artists tailored to your taste. Your playlists are also backed up by a renewing selection of recommended songs based on it’s contents for infinite further listening. Play, skip, save, repeat. There are a few methods to how this all works. First is ‘collaborative filtering’, which focuses on the collective listening and data of all users to determine the songs you’re served. Put simply, if an album or song is gaining high listening traction in the playlists of users who have a similar artist or genre profile to you and you haven’t yet heard it, it’ll likely end up your way. It’s like a musical echo chamber and the reason why you and your mates might end up with the odd match, but the algorithm is sophisticated enough to diversify what it’s offering you. Shakin Stevens isn’t gonna be hanging around after December.
The second starts to look like that maths whiteboard that bored you to death in school. Audio analysis is where the raw merits of a song (it’s tempo, style, key arrangements) are analysed to determine what you might like. This takes place prior to the collaborative filtering process warming up. Lastly, ‘web crawling’ (unconfirmed whether this is Spiderman homage) searches the streams of information about a song or artist via reviews, blog posts and headlines from across the online music industry to discover patterns and attribute these to new recommendations and sub-genres. Also informing all this is the minute behavioural profiling we’re accustomed to on social media, whereby Spotify will track all elements of your screen time. This includes save and listen rates, frequency of plays, artist profile visits, time spent in playlists, time of day, location and so on. That’s the creepy business and it’s how Spotify profits off us.
On a personal level it seems like a fairly benign exchange of information versus reward, when you consider how much we surrender ourselves for more mundane conveniences; endless streams of new music at your fingertips seems like one of the more decent pay offs. Sure you might scoff or be indifferent to some of what Discover Weekly or Release Radar throws up but when it lands right, you’ve instantly got access to that artist's entire back catalogue and the voyage begins. Losing yourself in a Spotify wormhole thereafter is not mindless scrolling or validation seeking, it’s musically enhancing. That’s valuable in an era where we are devoid of venues, record shops and even personal contact to enlighten us. Hachiku was the support artist I’d swayed along to approvingly on Snail Mail’s UK tour in 2018, one I’d noted to follow up on but who’s performance got lost in the haze of Lyndsay Jordan’s set and has long since been forgotten. Her top song has amassed no more than 75,000 streams yet it’s appearance on my Discover Weekly last month has led to her recent album becoming one of my favourite ‘new’ finds. These sections are also currently firewalled from labels or artists lobbying to appear in, making them one of most authentic aspects of Spotify’s fallible business model.
It's in that very business model where the darkside of this data driven discovery lies. It’s synonymous with Spotify’s wider ‘market share’ method of revenue distribution, which serves to shaft artists and reward labels with an unfair weight. Market share essentially means that Spotify’s pool of revenues from subscriptions and advertising is distributed on the occurrence of streams worldwide rather than centric to the artists you listen to via your subscription. Many could go the year without actively listening to Justin Bieber or Lady Gaga once, but they (or more precisely their record labels) could benefit from 2-3% of the annual streaming revenue. They are of course universally popular regardless but algorithms placing their albums at the forefront of playlists only serves to further the elite. It equates to a pitiful 0.004p per stream and is why a critically acclaimed artist like Nadine Shah has felt compelled to call out the model when revealing she’s struggling to pay rent. Another failure of trickle-down economics folks.Furthermore, there is an acknowledgement that the current system prohibits the wide scale emergence of sub-genres. Spotify might have introduced you to seemingly batshit terms like ‘chamber-psyche’ but pioneers of these new genres are still falling through the cracks when it comes to appearing in your recommendations.
We should all feel for the artists caught under the boot of the behemoths and user centric alternatives must sit atop the manifesto for change. Bandcamp, which has surged in popularity during the pandemic, is the most transparent way to get your money to artists and has the potential for scale. The statement of direct financial support, sense of community and personal curation you get from the site are a welcome middle finger to our streaming overlords. Exclusive endorsements from artists is a nice touch too, like when the incredible Phoebe Bridgers dropped her fascist bashing inspired cover of Iris on the platform. The announcement from Soundcloud that they’ll be shifting to a ‘fan powered’ royalty system should also be welcomed. More of the ‘big streamer’ heavyweights need to make moves like this because pay-per-click services like Sonstream, as noble as they are, lack the wealth of licensing. Who are we to cast as the villains holding this back then? The major labels of course. Much like dairy or oil companies when faced with the prospect of plant based or renewable solutions, the labels stand to relinquish a lot, £700,000 per hour worth of a lot. That’s what the market share streaming model earns them, it’s algorithm driven methods ensuring the continued lining of their pockets. Some even retain stock market shares in Spotify, it’s the kind of vested interest we’ve become accustomed to in politics the world over. Speaking of which, Universal Music UK CEO David Joseph recently professed the need for a less algorithmic and more curated system. Yet that is at odds with his counterparts and palatable words are easy to find when under the scrutiny of an MPs select committee.
It’s probably nothing new to us that streaming is doing a disservice to musicians. Yet those algorithms, utilised by the likes of Spotify to find new and deepers ways to entice or retain us, are a big factor in facilitating it. Be that through discovery, release, playlists, stations and mixes or the hyper personalised Wrapped campaign, which along with being a fun reward for our data is an extremely savvy marketing tool. Spotify users will continue to grow and it’s data profiling capabilities will submerge us further into the music matrix. The more the algorithm knows the more refined a service it provides and the more we come to view it’s function as reliable, dependable even. Our emotional connection to music entrenches this dependency further.
Now then, the very use of the algorithms and the hyper deterministic way they allow for the discovery of music isn’t the apocalypse in of itself. As mentioned, personal experience has made me grateful for the countless ways this can lead to enriching and expansive music consumption. Yet the way the algorithms are being deployed creates a pretty centralised pattern of that consumption and props up an imbalanced system of financial distribution. By feeding the algorithm, we are starving the industry.
The impact of this pandemic and our haphazard departure from the EU could well result in a generational black hole for music. Not just artists and writers but all the gig economy professions that both enable and rely upon the revenues they generate. Ultimately, big changes from the fat cats in companies and government is what’s required to avoid this, but we can do our part too. The purist view may well be to cancel your subscription to a Spotfy or Apple Music immediately, but only you lose out there. Sacrificing some convenience is a more realistic start. Subscribing to alternative platforms or paying for content, merchandise and ticketed live streams, when you can, are all rewarding and tangible ways to know you’re helping musicians. Broadening the sites in which you browse for new releases and dedicating more of your time to live radio, commercial or niche, are other ways to decouple from the streamers algorithm. The extent to which we do this will depend on your own level of affinity to the music you consume or whether you have the luxury of individual spending power. Yet just diversifying how we discover music can be a personally empowering approach to keep our reliance on the machines at bay.
Every Friday I get a list in my inbox of 20 songs to listen to. This isn’t devised on a streaming service or even recommended via a music publication, it’s curated by five of my work friends in a spreadsheet and spans three generations of musical tastes. The following Friday, the list renews along with a summary from each of us on last week's tracks. Sometimes it’s glowing, sometimes it’s dubious, but it’s introduced me to all sorts I wouldn’t have otherwise heard. A recent submission of some Japanese electro punk can testify to that. It’s hardly sticking it to the man, we all divert some of the listening to our streamer of choice. But until we can fling ourselves back into sweaty clubs, nodding in appreciation as the reverb of a set sends the lager in hands flatter still, it’s a neat way to not be so dependent on computer says: listen.
Each week we will share some tracks that the contributers to Laughing Stock currently have on heavy rotation. You can follow the rolling playlist on Apple or Spotify.
ICYMI: More from Laughing Stock
Edition 4 - @dickiewalker on how amazing organ music is
Edtion 3 - @williampalmer on Music as Muscle Memory
Edition 2 - @connahr on Arctic Monkeys and Quantitative Easing
Edition 1 - @dickiewalker on Dua Lipa and the new pop vanguard
If you are interested in contributing to future editions of Laughing Stock, please DM either @dickiewalker or @connahr