AI is already changing the music industry – but those changes are not being driven by fakes and cloned voices says Duetti CEO and co-founder Lior Tibon. The real impacts are coming as turbo-charged AI assisted algorithm-driven personalised discovery changes the way people find and consume music.
Duetti’s platform helps artists receive up to $400,000 for their tracks, utilizing data-driven prices for established master tracks, allowing artists to sell individual tracks or even parts thereof, while Duetti then markets those tracks using unique ROI-focused techniques. In 2024, Duetti announced $90M in new funding, bringing its total raised to over $120M in just 18 months since inception.
Previously, Lior was the COO of TIDAL, the global music streaming platform. In January 2015, Lior joined the core team that launched TIDAL by Shawn “JAY-Z” Carter and its original artist owners. TIDAL was acquired in April 2021 by Block (formerly known as Square). As COO, Lior oversaw all of TIDAL’s strategic business and marketing initiatives. Prior to TIDAL, Lior worked in the Corporate Finance division of Deutsche Bank in London from 2009 to 2015.
According to proprietary research conducted by catalogue acquisition company Duetti, which recently raised $90 million of new funding, show that over the last six months algorithmically originated streaming on Spotify grew by over 10%, while editorial driven streams are down by over 20%.
For over a year the music industry has been engaged in a very public, vocal and often polarising debate on the implications AI might have on music. In April 2023, a “fake Drake” track started trending around the world, becoming headline news before being swiftly taken down by streaming services. The focus of that debate was multifaceted and complex, encompassing copyright infringement, the moral ethical and legal implications of cloning an artist’s voice and musical style, and more. While all of these topics are clearly important, the current debate is not focused on the more important and immediate impacts of AI on music.
In fact, AI’s most significant impact on music in the foreseeable future will be less in terms of changing the fundamentals of music creation. Instead – AI will hyper-accelerate the ongoing gradual shift to more personalised discovery of music by fans, by making the existing discovery algorithms of the various platforms much more sophisticated and much more customised to the tastes of individual listeners.
“The core challenge in music today is not creation, but discovery…”
As this hyper-personalised AI-powered discovery and listening drives more streams the power of editorial playlists will lessen. In turn, this will continue to drive a greater fragmentation of music consumption, skewing away from high visibility playlists populated by big name artists. As a consequence, listening habits will shift to support a broader array of “middle class” artists and creators.
The implications of this AI-powered shift will have significant ramifications for the current commercial set-up and focus for artists, their teams and other constituents in the music industry. Instead of a number of big “superstar” releases with mass market appeal being amplified through editorial playlisting, AI will support fans and listeners in the discovery of new music and artists at all stages.
In this new world of AI-led discovery, the Grammys will feel less and less relevant, out of tune with the desires of listeners and fans as they flock to a broad diversity of discovery catering to their specific tastes. Blockbuster hits will still have a place for some – but their domination of streaming market share will shift significantly. Robert Kyncl, Warner Music’s CEO, reportedly correctly identified this meta-trend in his annual note to employees in January – whereby he pointed out the need for Warner to build future-proof digital infrastructure which will serve many more artists in much more sophisticated ways.
Recent industry stats give a forensic illustration of this trend. The core challenge in music today is not creation, but discovery. With streaming platforms adding tens of millions of tracks a year, there is a huge volume of music that is either being newly created, or older content being digitised and ingested for streaming for the first time.
With that increase in quantity comes a corresponding increased need for better discoverability – both for artists and fans, but also for streaming platforms themselves. Discovery is a core driver of retention for streaming – listeners and fans want to find new music, and discover music from the artists they care about. If those fans and listeners feel overwhelmed by the volume of music available, and feel that streaming platforms are missing the stuff they care about, or pushing too much stuff they don’t care about then the whole ecosystem suffers. Luminate data for 2023 shows that out of the 184 million tracks live on DSPs, 83% generated fewer than 1000 streams.
“…we are already detecting gradual ongoing shifts in Spotify’s streaming activity, with fewer streams being driven by editorial playlists and more coming from recommendation algorithms…”
In this environment AI is inevitably going to have a tremendous impact on improving personal discoverability by enabling more sophisticated recommendations for fans based on the attributes of the music itself, or various consumption markers.
This is not a shift, but an evolution – an extension and enhancement via AI of the existing machine learning algorithms that power discovery. But it means one thing: algorithms will become even more important. If we see an exponential increase in the quality of algorithmic recommendation powered by sophisticated AI driving enhanced “matches” between artists and relevant listeners and fans, then “top down” manually curated playlists will inevitably decrease in importance.
At Duetti, our focus on developing sophisticated predictive technology and pricing models means that we are closely monitoring these underlying trends. Coupled with our innovative data-driven track management and marketing tactics, this enables us to purchase older catalogues of tracks from artists of all sizes – knowing that we can help these tracks perform in that new model of discovery. This in turn expands the financing options which are available to artists.
Critically, we are already detecting gradual ongoing shifts in Spotify’s streaming activity, with fewer streams being driven by editorial playlists and more coming from recommendation algorithms.
Our data allows us to estimate that, on average, and in just the last six months, algorithmically derived streams grew by over 10% for the typical independent artist, l while editorial streams dropped by a whopping 20%. Part of this is due to a specific push by Spotify for artists to opt-in into their Discovery program, which promotes algorithmic discovery, but – more broadly – it represents a more fundamental increase in the impact and appeal of algorithmic discovery, in a world with unimaginably vast quantities of music.
“Algorithmic understanding will become key to driving music marketing success…”
Our research aligns with Bloomberg’s January reporting on the de-prioritization of editorial playlists and the increasing switch to algorithmic discovery by Spotify – for example, the report refers to declines of over 30% in streaming activity on Spotify’s flagship hip hop playlist Rap Caviar, and notes comments made by Spotify CEO Daniel Ek saying that users were increasingly opting for algorithmic suggestions and Spotify would be leaning into that trend.
All of this will only move faster this year, aided by AI adoption by streaming platforms. Labels and managers will need to consider how to adapt their marketing strategies and operations in 2024 and beyond:
Playlist pitching will become less relevant for many artists and songs.
The playlist universe is already showing the effects of diversification and atomisation. Chartmetric’s playlist data is currently tracking 20 million individual playlists on Spotify – a huge ~33% increase over just one year ago. The average followers per playlist is well under 100 people, and fewer than 1% of playlists are able to break above 10,000 followers – the kind of threshold that can drive meaningful scaled audience engagement.
Data-driven analysis of this playlist ecosystem will become increasingly important, with music marketers needing to target dense networks of playlists rather than just a few flagship editorial placements.
Algorithmic understanding will become key to driving music marketing success.
As the trend for sophisticated algorithmic discovery gathers pace, music marketers will need to develop a strong understanding of the underlying metrics and mechanics that drive the recommendation algorithms across the major platforms.
Understanding how songs are classified and viewed by the algorithms, with associated audience sizes and potential, will become more critical in assessing the marketability of each particular song and concentrating on the right marketing tactics.
“The biggest impact that AI is having on the music industry today is not gimmicky vocal clones of superstar artists, or huge volumes of low quality AI-generated tracks…”
There are a number of advanced data tools and platforms which support mapping out relationships across artists and across songs, as well as tracking consumption behaviours – all are critical inputs in “training” and enabling optimisation of the algorithms. According to the excellent EverNoise project, there are over 6000 genre codes within Spotify, which form a critical part in classifying music to different data “buckets” and therefore drive different algo recommendation patterns.
Different platforms use different algorithms – and understanding those differences is important.
While the vast majority of digital platforms are gradually incorporating more sophisticated AI into their recommendation algorithms to push and improve them, there are significant variations between the approaches of different platforms. This is not just the shape, construction and underlying technology powering those algorithms, but also the relative impact algorithms have on overall streaming volumes. For example, Duetti’s research indicates that YouTube’s algo-driven streams (as a percentage of total streams on the platform) are at least two times higher than Spotify, meaning algorithmic discovery is driving more consumption on YouTube than on Spotify – for the moment.
Where some platforms only take consumption markers into account when designing their algorithms, others assign great importance to intrinsic attributes of the music itself, such as genre, tempo, or proprietary metrics like Spotify’s “danceability” score.
While “gaming the algorithms” is extremely difficult, potentially counterproductive, and not at all advisable, music marketers should focus on ensuring that the music they are working with on is optimally positioned to get “picked up” by the recommendation engines across all the major consumption platforms, utilising a variety of ancillary assets and specific audience targeting to ensure algorithms are seeing the rights “signals” from tracks.
The biggest impact that AI is having on the music industry today is not gimmicky vocal clones of superstar artists, or huge volumes of low quality AI-generated tracks. Instead, it’s the “hidden” impact that AI-augmented discovery algorithms will have on discovery and consumption shifts.
As AI becomes ever more ingrained in the core operating mechanics of the existing streaming platforms this will enhance the quality and capabilities of recommendation algorithms. In turn, the power of “top down” curation and superstar power will be reduced as algorithmic discovery accelerates diversity and truly individualised discovery and consumption. Music marketers need to prepare and adapt, taking a more data-driven and nuanced approach to promotion.
Duetti stands ready to advise and support independent artists and their teams as they prepare to navigate this new ecosystem of discovery and consumption of music.
Leveraging their experience in streaming and support from music and tech’s most innovative investors including Viola Ventures and Roc Nation, Duetti’s music financing platform has helped over 60 artists receive up to $400,000 for their catalog and single-track masters.
The unique model provides data-driven prices for established tracks, allowing artists to sell individual tracks or even parts thereof, while Duetti then markets those tracks going forward using unique ROI-focused techniques.