The music industry has failed to keep up with the pace of technological advancements. There is no denying that the reaction of the music business to the digitalization of their industry was harmful. While the extent of this can be debated, what is clear is that the music industry suffered from a lack of engagement with new technology.

Thankfully, with the decrease of downloads and the rise of streaming, record companies are now beginning to place focus on big data, as a tool to drive the industry forward. As they increasingly look to incorporate technology-driven methods of analysis, we assess to what degree Artificial Intelligence poses risk to the field of A&R.

Streaming & Analytics

Data is refining the music industry in creative ways, with streaming now allowing companies access to a wide range of consumer statistics [1]. While record companies may have been late on the uptake, streaming services have advanced full-throttle. The belief of the latter, in the value of this data, is demonstrated by their acquisition and absorption decisions made in the last year.

In 2017 alone, three of the largest music streaming platforms in the world acquired companies in the business of music data. Firstly, January saw Apple Music purchase MusicMetric, as part of a wider deal. Then, just two months later, Spotify acquired the music data firm The Echo Nest. Shortly after, this was followed by Pandora absorbing popular music analytics service, Next Big Sound [2].

SoundCloud, a huge player in the streaming market has felt the force of the fast paced industry. Once thought to be valued at $1 billion, 2017 saw SoundCloud seek bids as low as $250 million, following a failed acquisition by Spotify [3]. There has also been significant personnel change. In July, 40% of SoundCloud staff were laid off in an attempt to cut costs [4]. Eventually, the service was saved by emergency funding [5], but such drastic changes to the workforce could perhaps be indicative of a trend. While the industry may once again have started to enjoy a period of re-growth, it is certainly not stable. As the influence of big data continues to penetrate record companies, there is speculation as to the impact this will have on their staffing structure.

If technology is increasingly able to move towards delivery of predictive analysis, what does this mean for the field of A&R? Can human prediction become redundant? Below we look at both sides of the argument, starting with the importance of AI and big data.

Machine: The Importance of Music Technology

Digital Interaction

The digital interaction between consumer and music is essential to understanding consumption, because the digital world is both where music is most often discovered and purchased. It then follows that the richest source of information would come from technology that can interpret this relationship, rather than a physical source (human observation).

If this information can be delivered, reliably, through new technology, it is inarguably an essential component of the new music world. Indeed, if we continue to trend in this direction, it is important for music companies to review their process for finding and fostering talent.

‘new technologies, big data, and audience insights are changing the game, and the teams with the best tools, curation, and expertise, win.’ - Hannah Chapple [6]

Affinity Data

Affinity data, which explores co-occurrence relationships, is one example of technology’s ability to dive deep on audience profiles. Data mining via the affinity method can find out exactly who an audience are and what they like [6]. Through this process, it is then possible to identify artists who are beginning to resonate with a specific crowd.

This is particularly valuable for the field of A&R, as it provides data about artists who are beginning to gain a following, within a subsection of the population, but are yet to achieve mainstream success. For record companies, this is the ideal time to make an investment, and pair the existing following with their marketing power, to profit. If record companies are able to leverage affinity data in a way that gives them almost immediate identification of rising talent, they are able to refine their discovery process and substantiality raise the probability of success. Essentially, this would mean that fans of music are finding the talent for them.

Such technology casts serious doubt over the value of dedicated label staff engaging in talent discovery. The question then becomes, not whether technology can screen and discover talent, but how reliably it can do so.

Spotify Algorithms

One indicator of the power of technology to reliably interpret big data comes from the algorithms used by Spotify to generate their ‘Discover Weekly’ playlists. The complex algorithm is based on a combination of factors, including user’s own music profiles and other people’s playlists, as well as which songs a user chooses to skip or play through [7]. Songs from the Discover Weekly playlists have been streamed over 1.7 billion times, and many have commented on how effectively it is able to generate playlists, that appeal to consumer interests.

The success of these machine-generated playlists demonstrates the ability of technology to accurately make prediction decisions, based on existing data. This then, gives indication that the role of talent discovery is facing further revolution at the hands of analytic-technology.

Predicting Hits

To go one step further, rather than just predict preferences based on affinity data, there is also evidence that machines can successfully predict songs that are likely to be hits, based solely on their auditory components.

Scientists at the University of Antwerp analysed electronic dance music songs on the basis of 139 auditory components, to impressive results [8]. Using US Billboard Dance Singles chart for 2015, the algorithm correctly predicted that all of the songs featured in the top ten had a 65%+ probability of success. For 6 songs within the top ten, the machine gave 70%+ probability. In further support of their findings, a similar study of EDM songs in the UK Singles chart for 2015 was conducted, and the algorithm gave each between a 68%-90% chance of being a hit.

The Holistic Machine

If we combine the abilities of affinity data to discover artists, algorithms to make preference choices, and auditory analyses to predict the commercial success of songs, it is evident that the future of music discovery lies within technology. Together, these functions are perhaps troubling for the future of record company staff that are working within talent acquisition. The second part of this article considers the value of the A&R manager in today’s industry and explores whether they are still able to add value in a technologically driven age.

Man: A&R Roles in a New Age

So far, the evidence would suggest that A&R may soon be made obsolete by technological advancements. However, to dismiss the value of the role, based solely on the use of data and statistics, would be naive.

Daniel Hall, Senior Director Insight, Sony Music International went on record at the 2016 MIDEM conference (originally reported by [9]), to say that while the music industry is working with analytics in a new and refreshing way, we have to understand that big data is not able to actually make the decisions. Despite this, he acknowledged that data can be good at unearthing hidden gems, or tracks that might have otherwise been missed.

Simon Wheeler, Digital Director of Beggar’s Group, is also clear in his opinion that A&R cannot be replaced by technology. At the same conference he stated that professionals want a rich, diverse world of talent and that a singular reliance on big data is insufficient in this regard. Wheeler continued:

‘we have the role of finding things that people don’t know they’re going to like… and data is not very good at doing that stuff. That’s why people who are great A&Rs can see down the line about what this artist is going to turn into. I’ve never seen any data that can do that.’

A year later, at a 2017 MIDEM conference [10], this sentiment was echoed by Daniel Miller, founder of Mute, an independent record label. Miller spoke staunchly in his belief that data is limited when it comes to A&R:

We don’t ignore it but it’s not a major part of the decision-making process. Gut reaction and emotional feeling are the most important things and you take it from there. When you find a new artist, there usually is no data – it’s only 10 people at a gig.”

It’s also important to remember that the discovery and acquisition of talent is only one aspect of a much larger A&R role. To suggest that A&R can be replaced by technology is to dismiss its value to other record label functions. Indeed, A&Rs also manage the development of talent, oversee the recording process and assist with the marketing and production of records [11]. Speaking to this point, in a guest post for Music Business Worldwide [12], former A&R Ben Wardle argues that the role of A&R goes far beyond abilities in music prediction and hit discovery. He highlighted the importance of the role across all label functions, such as facilitating relationships with producers, driving the direction of record labels and being a representative for artists throughout the creative process (including design and marketing - not just music making).

Seemingly, these tasks are fundamental to the field, and not ones that can be replaced with machines. For those working within the field, it is evident that despite massive changes, there remains a core value to A&R that is impenetrable from the growing scope of machine.

Concluding Remarks

Technological advancements are leading to novel measures of data capture and analysis, proven to provide real value in the fast growing world of music. A&R is one field that can benefit in particular from these new developments. Still, it is clear that while professionals are increasingly willing to integrate these measures to assist with their A&R function, they remain confident that technology cannot replace it.

Ultimately, it would appear that while music analytic software programs are informative tools, they not decision-makers. At least, not yet.