NETWORKING DATA A MULTIMETHOD EXAMINATION OF NETWORK PERSPECTIVE FOR ARTIST MANAGEMENT

(1)

NETWORKING DATA

A MULTIMETHOD EXAMINATION OF

NETWORK PERSPECTIVE FOR

ARTIST MANAGEMENT

S.J. DONKER

(2)

(3)

NETWORKING DATA: A MULTIMETHOD EXAMINATION OF NETWORK PERSPECTIVE FOR ARTIST MANAGEMENT

Master Thesis

Digital Humanities | Communication and Information Studies Rijksuniversiteit Groningen

In partial fulfillment of the requirements for the degree of

Master of Arts

By S.J. Donker s1815989 Supervisor R. Prey, Ph.D Second reader M. Esteve Del Valle, Ph.D

RIJKSUNIVERSITEIT GRONINGEN Groningen, Netherlands

(4)

Abstract

In the last few decades, the music industry has undergone huge changes caused by technological developments. Releasing and listening to music now mainly goes through online platforms. An important effect of this digitization has been the massive amounts of data that is produced in the process. This thesis proposes artist managers, as central figures, to approach the digital and datafied processes from a network perspective. The subject is investigated qualitatively and quantitatively. Incorporating a critical stance towards data and streaming platforms, network theory can help deal with the data deluge by teaching us how we can see things by their relations instead of as isolated events. The proposition is supported by a case study of Spotify's related artist network of Dutch drum and bass artist Noisia.

Keywords: digitization, datafication, music industry, Spotify, music streaming, network

(5)

Acknowledgements

I would like to express appreciation to my supervisor, Dr. Robert Prey, for introducing me to the subject and for his exceptional guidance and enthusiasm on the subject of this thesis. That last bit also applies to the second reader, Dr. Marc Esteve Del Valle, who I thank for his lessons on network theory that have been essential to complete this work. Meeting with either of them always left me inspired and eager to continue.

Special thanks goes out to Walter Flapper, for providing me with information necessary to complete this thesis as well as a pleasant working environment where I have learned a great deal about the music business.

(6)

1. Introduction

As Walter sits down at his desk, after the regular inbox-check, the first thing to do is browse through some statistics: how many likes did that post from yesterday get, how often was it shared? Followed by some phone calls and a quick look at performance data: have the followers grown, and how well does that new track do? Then, a Skype

conversation with a booking agent where it can be pointed out that maybe South America has potential for some shows, as according to the streaming platform a huge audience resides in that part of the world. After that, first a glance over some other data dashboards, to be armed with cold hard facts for the next call and the meetings later on.

Walter is an artist manager, and throughout the day, performance and tracking data are permeated in most of his activities. He manages a handful of EDM (electronic dance music) artists, of which the Groningen-based Noisia is his main client. Although the artists he manages are unique, there are many Walters in the music industry. It is true, in today’s fast-paced and always connected society, in the music industry the interaction with fans, users and other types of audiences nowadays largely goes through online platforms. Beside the social aspects of it, these platforms offer previously

unknown possibilities to their users— listeners, artists and their management alike. These possibilities stretch far beyond the straightforward ‘how many records did we sell?’ or ‘how do we reach as much people as possible?’ Fans serve themselves through a variety of platforms leaving traces of their activity. The systems behind streaming and social platforms are driven by the data traces the users leave behind and some of it is available to the management as well. In line with the way the platforms operate, we now find ourselves looking at things like ‘at what time does my audience listen to the music?’ or ‘in what kind of playlists is my music successful?’

There is access to quantitative information on very specific activity of users and the performance of products. The provided metrics play a huge role in the industry, influencing decisions for artists on every level; who to contract, where to book, who, what and when to include or exclude and so on. It becomes increasingly important to get a grip on what is happening online and in what kind of position the artists find

themselves. The abundance of data that is produced allows for an approach that’s both individualized as well as embedded in the greater structures of the industry. It has

(10)

created possibilities for insight of what happens with the music and how the audience behaves after an artist releases a track to the world. It is safe to say that the digital turn has had a massive impact on the music industry in a relatively short time. Its recent developments follow from a fast-paced technological advancement and is shaped by a complexity of social and cultural developments and practices that permeate the systems we’ve grown used to so fast. Having data gives a sense of having control, of knowing what is going on. Still, at the same time, the data only states certain facts or

measurements. The digital environment is relatively young and constantly changing, and there is still a lot of experiments occuring of which any user can be the subject of at any time. The degree of power and control one has is thereby relative. If we acknowledge that the current data deluge is new for everyone and while the platforms usually do lend to a great extent access to the data, in what way can we realistically gain insight?

If we take Spotify as an example, managers are in the first place users, clients, of the music streaming service. Although artists and their management might have

different interests than regular ‘listener-users’, importantly, all users depends on the provider (Spotify) for information: the users only see what the provider releases to them. With the information coming from a data-driven commercial business, it's safe to assume the mediated selection of information for users would be at least somewhat, but possibly heavily filtered in their advance. Management users get access to Spotify’s Spotify for Artists, that tells them their artist has, for example, ‘377,2k monthly listeners’ or ‘71% of your fans listened in the last 28 days’ . Also the less obvious, and visible to any user of 1

Spotify, lists of recommendations in the form of ‘popular’ tracks that appear at the top of an artist’s profile, together with related artists you can find on the platform. Here as well it is by sophisticated algorithms that are learning from previous and real-time use and click behaviour, that determine what is presented. Some of this information is visible for everyone, some only for artists and their managers. While for regular listeners a

suggested playlist for dinnertime is nice and can lead to discovering new artists; for artists and management something as appearing in the ‘right’ playlist becomes of vital importance and can make all the difference in success or failure.

‘Insight’ does not logically or by definition follow from statements such as an audience being 75% male, that fans listen through their mobile devices more often than

(11)

on their laptops or that your artist is associated with a certain artist and not another. User-friendly dashboards with visually appealing presentations of often real time generated metrics tempt us to think ‘the numbers speak for themselves’ but there are several issues with this idea. When for example Spotify, Youtube or Facebook provides some access in the underlying metrics of their platforms, it’s easily said that data on audiences helps you find trends and patterns to predict behaviour and convert that into social or economic value. Surely the presented data can have huge economic and social value, but there are always many uncertainties. Did the manager check the right

websites, connected the most important events and maybe that new app can better accumulate all the information than the other one... Beside being overwhelmed by the abundance of data, concerns about accuracy can arise, or about the source. The

platforms that make artists able to be seen and heard simultaneously control the way it can be done and what data is accessible, with algorithms mechanically determining what is processed in the first place. These issues play a part in the background, and the daily reality does not leave much space for evaluation; we are given the facts, that should be enough. But we can not assume that merely having data gives a manager power. With today’s massive amounts of data, often generated real-time, the question of how to make sense of it beyond mere state-of-fact remains problematic. Considering this, the

knowledge and power data can give depends on the meaningful context or paradigm we place it. What follows is a proposition of one such paradigm as a way of approaching the issue of dealing with data and online platforms: through networks.

No man is an island entire in itself, as John Donne declared almost 400 years ago. Just as human beings shape and are being shaped by their environment −connecting to and relying on what surrounds us− the same can be said of non-human elements. From a network point-of-view, all elements −in network terms: actors− play a part as something that can ‘bend space around itself’, or in other words: have an effect on (and is affected by) their environment (Callon and Latour, 286). It is of no use to isolate any actor when meaning comes from relations and patterns in relations. From network perspective everything is embedded in structures of relationships, and network awareness “provides new insights into the structure and functioning of our societies, and how we should operate in them” (Rainie and Wellman 56). Getting away from the traditional

(12)

provide insight in the world as a changing place, as a “more diversified, more complex and more interesting place” (ibid).

We are hardly strangers anymore to jargon that was previously typically

network-scientific: words like hubs, clusters, social ties and of course the word network itself have been included in our regular language, probably simultaneously with the technological developments of the last few decades and most notably the internet revolution that started in the seventies. With the internet we find ourselves in a more decentralized, asynchronized world that is more open to individual choice, and that allows us to be more networked. But while ‘networking’ is a well known term, the fact that networks follow internal rules and patterns that are quantifiable is less well known. More often than not we rely on concepts that simplify to such an extent that the results can not sufficiently explain the reality we experience. Many explanations try to reduce a subject to an autonomous affair, while network explanations...

...do not assume that environments, attributes or circumstances affect actors

independently. Moreover they do not assume the existence of uniformly cohesive and discretely bounded groups (Marin and Wellman 12).

It is the difference between claiming that an artist has become successful because she has a talent and worked really hard to succeed by herself, and acknowledging that she reached a certain point because of a collective effort, made possible by the relations of her particular environment and having or creating access to tools that have influenced the outcome. Thinking in networks thereby offers an opportunity of doing more justice to the complexity of a networked reality than approaches focussed on the units

themselves.

Rainie and Wellman (2012) explain how today’s society has changed from large hierarchical bureaucracies and small, densely knit groups (like households and

communities) to a society made of what they call networked individuals. Through technology we changed the way we interact and relate to each other, enforcing greater social flexibility and more personalization (34). In this ‘world of networked individuals’ technology makes it easy to interact with each other as connected individuals, instead of embedded as groups members (ibid 12) . They call networked individualism an

(13)

communicate, and exchange information”, and also use the phrase because it

“underlines the fact that societies — like computer systems — have networked structures that provide opportunities and constraints, rules and procedures” (ibid). If we want to approach performance and tracking data with a concept of reality that fits our time and means, we can benefit from using network theory as a perspective to grasp what is happening. Although the idea of networks is not new, the way we are networked today is, and we can use new strategies and skills to address these networked structures that underlie our relations.

1.1 Research question

In this thesis, I would argue that adopting network perspective and methods of analysis can help artist managers in this increasingly fragmented and networked society to understand the current situation of the digitized music industry and deal with

information from online platforms or guide decisions in regard to what they offer. The subject will be addressed by a theoretical exploration of digitization of the music industry and network theory, as well as an empirical case study to substantiate the hypothesis of why network theory and methods can be a valuable asset. The case study regards a network of three Dutch drum and bass musicians known as Noisia, managed by Walter’s Flapper Management (‘Flapper’). Their case offers a suitable example, as they have a rather large online fan-base spread across the world which makes their data reliable to work with. Flapper is a small, but successful and self-made agency. They have been operating for about ten years, which means they have been building their business from just about the time the music industry started to adapt to the digitized landscape (see chapter 1). Flapper has offered permission to use their data, which was necessary to address the questions thoroughly.

From the idea that applying a network perspective can help artist management gain insights in data and online platforms, the general research question is as follows:

- What possibilities does network theory provide artist management?

With, for the study of a case, the subquestions: - What is Noisia’s related artists network?

(14)

- What insights can be provided by an analysis of this network?

In the next chapter will first provide full context of the topic by addressing the

digitization of the music industry and the increasingly important role data became to play. This background will relate the topic to a general context. Chapter three provides the theoretical framework where I will go deeper into network theory and different ways of using it, which provides background and basis for the research conducted. Chapter four will give a straightforward explanation of the methods used to do the network analysis for the case study, including the way the data was collected, processed and more details on the sample. The fifth chapter contains the study of Noisia’s case, where results for their related artists network are presented, preceded with some more information on the artists. Chapter six contains a discussion with a reflection on the results and answers to the research questions, concluding with some limitations and ideas for future research. The conclusion summarizes the thesis, going over the most important findings.

(15)

2. Data and the music industry

This chapter presents background and a literature review. Our world has undergone huge changes with the coming of the internet and digital platforms. I will discuss previous literature related to the digitization in the music industry and the effects on a large scale leading to today’s streaming practices (2.1). The second section (2.2) concentrates on explaining two sides of datafication: dataism and scepticism. The third section (2.3) briefly explains why the artist manager holds an important position, to justify why their point of view is central in the thesis. The last part points out the gap in the literature in order to situate this research (2.4).

2.1 Digitization of the music industry

For over 130 years, the technological development of music reproduction has been constantly active and changing. The development of music recording and playback from the last century has seen “Thomas Edison's 1877 tin cylinder speaking-recording tubular phonograph, Emil Berliner's 1887 flat disk playback-only gramophone, the Victrola, 33 1/3, 45, 78 shellac then vinyl records, reel-to-reel, 8-track, and cassette tapes, compact disk, DAT, and MP3” (Goodrich et al. 2011). Although a fascinating history, the focus here will lie on digital technologies that since the late 90’s have become increasingly

prominent, starting at the MP3. This is when the digitization of music took off, and in 20 years time changed at least as much as the 130 years before that. Digitization has had implications on a whole different scale than before, in terms of “how music is produced (and to some extent thereby how it sounds), how it circulates, and how people access music, engage with it and ‘socialise’ (through) it” (Nowak and Whelan 2016, 2).

The software to encode audio files that became known as MP3 in 1995, provided small, compressed, user-friendly files that were adapted for widespread use. When broadband internet was introduced, MP3 files could be shared easily and efficiently. Even though at first not meant to be used for illegal sharing, that is what it became known for. With technology advancing and the users of it spreading, it was of course only a matter of time before it seeped outside of the walls of firms and universities where the software originally was created. For those who loved music, it offered great

(16)

possibilities and it did not take long before the technology got hijacked and was made commonly available (Rose and Ganz 2011) giving rise to ‘music piracy’. Building on these developments, in 1999 the infamous Napster Inc. was the first to launch a large-scale peer-to-peer file sharing network with at its height over 36 million users (ibid). This sort of distribution was unauthorized, and Napster was forced to shut down two years after founding. Some others like Kazaa and Morpheus and later on ‘torrent’ services came with more distributed models for downloading music illegally, making it harder or impossible to trace a single source of distribution (David 54). Ways of dealing with intellectual property, copyright and control became things that needed fundamental reorganization from the traditional manners if the original players wanted to compete. The shift away from the physical product raised “fundamental questions about models of production, distribution and consumption of music which, up to that point, had come to appear permanent and unassailable” (Nowak and Whelan 2015). With their efforts first aimed at intellectual property law and trade agreements, the music industry adapted slowly to the ‘networked individualism’, as Rainie and Wellman refer to the way people now tend to function: “more as connected individuals and less as embedded group

members” (Rainie and Wellman, 12). The changing digital landscape needed new ways of legal distribution and monetisation strategies to deal with decreasing physical sales (as is visible in figure 2.1) but the music business ultimately started to find alternative ways of distributing music.

Figure 2.1. Global recorded Music industry Revenues 1999-2016. (IFPI 2017).

(17)

Digital distribution in a legal way initially came through a pay-per-track model of which Apple’s iTunes music player was one of the first (McLeod 526), but the introduction of streaming music was the real game-changer since around 2008:

The shift to streaming can be seen as an attempt to regain control of digital music, which was lost in the post-Napster file-sharing era. The recording industry has largely conceded that if music can no longer be exchanged as a commodity, then the focus should shift towards commodifying the very spaces of music consumption (Prey 2015, 3).

Mobile devices got more sophisticated and internet connections faster and omnipresent. In this post-download era the most popular ‘space of music consumption’ have become the on-demand music streaming platforms. Streaming music from ‘the cloud’— large data centres comprising networked servers connected to the internet— makes listening to music available without the need to download any files. The content is instead experienced in real time as continuous streams of data, but listening can also happen offline when downloaded (Hagen 228). Music streaming platforms are the fastest growing and at the moment seem to be the future of digital distribution and

consumption of music. The digital segment in general now accounts for half of all the recorded music revenues (US$7.8 billion in 2016), of which streaming makes up the majority (56%) with Spotify as the leading platform, followed by Apple Music, Google Play and Deezer. After years of decreasing revenues, since 2015 the industry sees growth again, which has undisputedly been attributed to streaming services (IFPI 2017) (figure 2).

(18)

Figure 2. Streaming growth year on year: 2012-2016 (IFPI 2017).

Online music streaming platforms cater to the changed (social) environment we find ourselves in today: personal and connected, asynchronous and constantly evolving. The audience made up of the loyal and active music consumers that had been casted as thieves in the download-era, put in a different perspective now become

hyper-consumers and subjects whose activities could be monitored, influenced and monetized.

The great difference of services like Spotify with previous forms of music

consumption and distribution is unmistakingly the data feedback loop they generate in real time: “On contemporary music streaming services all listening time is

data-generating time.” (Prey 2015, 9). Data on online behaviour is being collected and analyzed and influences the decision-making greatly. For example the things users ‘like’; actively by clicking a like-button, saving a track or following an artist— or passively by disliking, skipping or removing a track. The possibilities of tracking subsequently have resulted in a ‘datafication of listening’ (Prey 2016, 32), meaning that social action —in this case the actions taken on streaming platforms— is transformed into data that is

quantified, making it possible to track behaviour through metrics and allowing for predictive analysis (Dijck 198).

In the past, knowledge on music consumption was minimal. There were of course numbers on for example record and concert sales, but when people actually listened to their records, how often, where and how was beyond sight. Today this is not the case

(19)

anymore, and every act on the platform teaches the algorithm behind it how and when we listen to a track, an album or a genre. How often we skip, save or share something and how this fits their bigger picture of users. All our actions:

… work towards building a profile. Without knowledge of how these processes take place, we are being classified and categorised, perhaps as a ‘Jetsetter’, maybe as a ‘Gamer’, and as either a ‘low-value listener’ or a ‘high-value listener’. (Prey 2016, 41)

For the platforms, all this adds up to a broad insight of behaviour from the platform users. It is by using the predictive qualities of massive amounts of data points that the streaming services have been able to draw so many users. Analysing behaviour and linking that to characteristics of their content, they accommodate listeners with a highly personalized listening experience, with an extensive library supplemented by

recommendations and autoplay functions that are supposed suit one's taste relatively well. By your profile and behaviour they aim to make a prediction of what you as a listener should like to be served. Using the networking possibilities provided by the internet, supported by the now portable devices, streaming and online services present a popular place for fans and artists that connects everyone, everywhere and in a legal way where artists are able to get a share of the revenues. As artist manager Walter expressed, their market is very scattered; their artists have a very small market share in many countries, and together this is interesting. Without the digital opportunities, chances are they would never be able to reach their audience (Walter Flapper 2017). The resulting performance and tracking data from the platforms is partially released to the artist (‘s management), offering them their own data pool on digital music performance. But even though it did become essential to the daily reality, can we actually say that the data ‘speaks for itself’?

2.2 Dataism and data scepticism

Some researches and articles address the changes in the field of music with regard to the digitization of the industry in general and how companies implement and use

techniques. Still, this is not extensive and the fact that it is very much in development means change and new information is added continuously. Strikingly often the

(20)

widespread belief in the objective quantification and potential tracking of all kinds of human behavior and sociality through online media technologies” (Dijck 198). In this line of thought, most writing focuses either on how data is an exciting new tool, adhering to the belief in data (the dataist approach), or is otherwise counter-dataism and emphasizes how biased and uncritically used the data is (the sceptical approach).

From the viewpoint of the first, the endless flow of data that is generated can be put to use by artists and their managers. There are the full-blown ‘dataist’ examples, propagating the idea that data can explain the world better than we ever could by

ourselves because they represent cold facts and are not biased by human interpretations: more data is more knowledge and thus more power. Next Big Sound is a prime example of a platform that collects and presents data on social on streaming data for artists in appealingly visualised figures, tables and graphs. Their promise is to make insight easy, and to provide “smart and timely insights for the music industry” (Next Big Sound) by offering reports that throw together statistics from artists pages like Facebook, Twitter, Youtube etcetera. The subscriber can obtain a weekly report on an artist by choice, which contains information such as seen in image 2.1.

Image 2.1. Take-out from the Next Big Sound Profile Report for the artist Noisia (Next Big Sound,

(21)

From their (the dataists) side, the general tone towards data is celebratory, and attention mainly goes to possibilities and promises of using the data given on platforms. The same tone commonly returns in (serious) blog-type articles that often miss thorough deepening of the subject. A lack of crediting and sound argumentation makes their standpoints hard to validate (and by that, to dispute), and sometimes a certain commercial interest biases the message. The stance is mainly positive and expresses or reflects the marketing message of the subject. Spotify itself for example promises advertising brands “overall understanding of your audience, so you can reach them in the right context, with the right message” based on listening behaviour in their post ”You are what you Stream” (Spotify), and this tone and ideas are reproduced elsewhere, like IFPI’s (International Federation of the Phonographic Industry) reports. Titlow, for example, refers to one of Spotify’s new strategy of targeting potential audience by listener location:

For promoters and managers, this level of insight is incredibly valuable. The data,

combined with Spotify’s massive reach (it has 100 million listeners), offers something that no stakeholder in the old music industry–not managers, big labels, concert venues, or anybody else–could have ever dreamed of providing: a real-time, direct, and reciprocal fan-to-artist relationship that can yield some demonstrably meaningful results (Titlow, 2017).

There is a trust in the data because it is seen as ‘fact’, and it promises us to tell what the audience is, does and wants. Of course, to a certain point statistics do reflect what is going on, but the numbers leave out as least as much as they show. Unreflectively, this leads to the idea that data merely has to be put into use (calling it a ‘treasure trove’ (Mombert 2015)) and does not give too much attention to the fact that the data usually is packaged to be used or sold to make profit, or the mediated way the data is being

processed.

On the other hand we find the sceptics. They take an explicitly critical stance towards technology and emphasize its generative characteristics. This ‘camp’ questions the truth value of data and what follows from it. Dystopian books like Schneier’s ‘Data and Goliath’ (2015) and Harari’s ‘Homo Deus’ (2015) received much praise. Schneier is

(22)

asking the (quite rhetorical) question if we are giving up more than we gain by leading such data-driven lives, and Harari’s work is an intelligent, but highly speculative book that takes dataism as the starting point for a future where humans triumph over nature by using complex networks based on data points, leading to the creation of something bigger than ourselves; an emergence of intelligence that makes us redundant– or at least lesser beings. This still more or less accepts data to be a reflection of something true beyond ourselves. Important, however, is what several scholars point out. That,

… however magnificent it may seem to have so much data available and to be able to mobilize that material in different ways, the promises of big data are a mixture of real potential with uncritical faith in numbers and hype about what those numbers can explain (Boyd and Crawford, as cited in Baym 2012).

What those numbers can explain is, quite simply, whatever a human designs it to explain. With many platforms offering ‘user-friendly dashboards’ that are “simple enough for non-analysts” (as offered for example by Grow, a platform to collect data and visualise it from over a 100 different websites and tools). Webster et al. (2016) explain how such systems are in fact ‘sociotechnical systems’. Although Webster et al. refer specifically to recommender systems, this idea applies to a wider range of data-driven platforms. While the software is ‘cold’ –not living, a tool– it is more accurate to see the software (and the systems arising from it) as a collective effort with a mix of human and non-human components. Built and formed by humans, the systems are a product of both human and algorithmic effort, and they become ‘cultural intermediaries’ . Algorithms 2

have to be told exactly what to do and how to do it, by humans. As such cultural intermediaries,

machines have been delegated ethics, values and duties and these are relentlessly, due to their mechanistic qualities, and silently prescribed back to the human,

and by that

2_{The term ‘cultural intermediaries’ was originally introduced by Pierre Bourdieu, but did not include}

(23)

regulates the cultivation of knowledge and the accumulation of cultural capital, which in turn, affects how individuals perform as cultural intermediaries (Webster et al. 2016).

This concept of how specific ideas about how people function and act are being built into algorithms and used as filters has been pointed out more often. Seaver notes that in this respect, “[...] social theories become increasingly performative: the models of social science may come to shape the phenomena they were meant to describe” (2012). Not seldom, the sceptics express a fear and tend to get caught up in dark envisions of how this will make us lose our humanity or how we almost passively get shaped by using technology. On the other side, the dataists are busy with crunching numbers, often seemingly naive, without real time for philosophical reflection. Too often there is a gap separating the two, while they’re dealing with much the same substance. To connect these worlds might add some complexity, but it will probably be more realistic and pragmatic. A manager in particular might gain from ways to do so, as much of their decisions are influenced by online data and platforms.

2.3 The artist manager

The music industry entails many different positions and professions, such as the artists, producers, publishers, booking agents, the record labels, music venues and so on. They all play some part in the scene and it would be impossible to regard all players and every aspect of it in one paper. Although investigating any player and their relation to data and online platforms could possibly give interesting findings, a central figure in the whole is the artist manager. Looking at data and online platforms, the manager relies on the information that the provider of the data offers; the platforms shape and influence the work processes and guides decisions in the workplace with promises to provide insight to performance. While the same probably holds true for a music venue or a label company, the manager fulfills a different role. As the person between the artistic production and the economic side, they have the position of an intermediary on many levels. They have to balance between efforts of artistic integrity for a creative product (with its intouchable intrinsic value) and economic interests (with its planning,

marketing, profits etc.)(Van Maanen 2009). Hracs explains how managers play multiple intermediary roles: they are consultants, curators, connectors, coordinators,

(24)

skill deficiencies of their musical clients and then broker appropriate skill and service ‘matches’ with members of the local creative community or broader freelance market” (Hracs 471). Managers are essential in “directing the traffic of ideas and resources, and by ‘matching’ ideas, individuals and organizational tasks” (Bilton and Leary 62). As such, they are gatekeepers, or brokers, mediating between artists, audiences and professionals. The manager’s position has become even more important since the record industry underwent the ‘crisis’ of digitization mentioned at the beginning of this chapter. This has created a do-it-yourself ethos and made musicians and management more independent to choose their own path (Hracs 466, Hesmondhalgh and Baker 93). The newfound possibilities that emerged through technological advancement give managers different, and probably more, options for control over the organisation and strategy of their work. The manager needs to gather and ‘translate’ data to his own artists, as well as third parties such as booking agencies, into something that has meaning: valuable

connections, well-performing collaborations, status of global coverage and so on. The artist manager will be central figure exactly because of this. It is interesting to research the topic of data and streaming platforms from their point of view, because the way their business acumen and strategic thinking is applied to uphold both the integrity of the creative product ánd the economic interests of their clients impacts the whole business.

2.4 Relevance of this research

The question of how to deal with digital data and online platforms is a complex one, touching on several issues, such as control, power and surveillance. The platforms are still relatively young and in full development- as are the ones that are using it. The situation is in progress the data that is accessible today, was not available a few years, weeks or even days ago and by any means will likely change in a short time again.

Standing in the shoes of the artists’ manager, neither of the two approaches worked out in section 2.2 are very fruitful. There is either too little regard for the origins and effects of data, or for the daily reality where we can hardly work without it anymore. We need to bridge the two ‘sides’ in an approach that can incorporate the important philosophical reflections of theory, and at the same offers time tools to deal with it, instead of casting it aside.

(25)

To date, there has been insufficient research that takes a critical stance towards data and online activity (without dismissing it) while still providing a long-term

perspective that incorporates methods and tools to deal with it. In this thesis I try to find a way of filling this gap by proposing to use a perspective that fits the rapidly changing (digital) landscape: network theory. Network theory has a long history of theoretical development, but is deeply grounded in empirical practice as well. As such, on the one hand it is useful for the long term as a perspective to contextualize our environment, and at the same time offers tools and methods to analyse the subject.

As a ‘digital humanities’ thesis, an important aspect of it is the combination of approaches that cross over traditional lines of inquiry. It incorporates the humanities ways, with high regard for qualitative properties, such as critical and contextual

reflection. From the information sciences we find a more quantitative focus, from where we can use the skill and application of digital tools and analysis. It is a combination of paradigms that do not always combine seamlessly, but I deem desirable to regard the subject in a meaningful way.

The next chapter will dive deeper into network theory –the theoretical foundations of the thesis– and elaborate on some things already mentioned above. It is the theoretical basis for understanding the moving away from ‘group-think’ and towards a more holistic view of our world, inhabited by networked individuals, with special attention to certain

(26)

3. Theoretical framework: networks

This chapter deals with the theoretical foundation of the thesis, by supplying background information and explaining some main principles of network theory._{First, a short}

introduction of the internet revolution and how that set the stage for network theory to become more relevant than before (3.1). Then, a general outline of network theory (3.2) is given, followed by an explanation of basic principles and visualisation of networks (3.3). Finally, an additional network approach is drawn in, which will supplement the general one (3.4).

3.1 The new social operating system

The story goes that in late 1969 the American student Charley Kline sent the first

information packet over a wire trying to login. He attempted to transmit the text “login” on a Sigma 7 mainframe with a computer at the Stanford Research Institute. They used the first link on the ARPANET (precursor to the modern Internet) between two

computers. After the letters “l” and “o” had been sent the system crashed, making the first message ever sent on the Internet a meaningless “lo” . About an hour later, after 3

recovering from the crash, the full text of “login” was successfully sent. It was done so on a room-sized machine with under-floor air conditioning, with just 128 Kb of memory and 24 Mb of disk space (Rainie and Wellman 59; McDowall 2015). How things have

changed…

In the two decades that followed, innovators in government, technology firms and super-geeks kept experimenting and they had the internet to themselves. When in the early 90’s not only sharing, but also displaying and searching for data became somewhat easier, internet usage took flight and became usable for a broader audience. Browsers became increasingly user friendly and the creation of HTML to display information through what came to be known as “web pages” made it possible for computers –and by

3_{Which isn’t all that meaningless: in an interview, Dr. Leonard Kleinrock, Kline’s supervising professor back in 1969}

says: “If you think about it, ‘L’ and ‘O’ is ‘hello,’ and a more succinct, more powerful, more prophetic message we couldn’t have wished for.” (McDowall, 2015).

(27)

that their users– to become connected on a great scale (Rainie and Wellman, 61). It only took two more decades to today’s society where we can hardly imagine not being

connected for a few hours through our pocket-sized smartphones.

There have been some factors at play that are important for the way the internet developed that defined the typical online open and sharing culture. Precursors in development were U.S. Government and tech-savvy individuals. The government supported internet growth and left control at a minimum, kept new taxes off and supported commercial development. Technology improved rapidly, computer power grew and importantly, the internet “did not balkanize into competing, mutually

unconnectable bits and pieces, but remained a network of networks — interconnecting internet service providers” (ibid. 63). The designers were not creating for a single client, but either for many or just for themselves. So even though it was made by humans, the internet was (and is) not centrally designed. It resembles an ecosystem rather than a swiss watch (Barabási, 145).

It is against this decentralized and open backdrop we can see how the internet quickly became a personal information device, for individual use and not bound by time and place (unlike other communication tools like TV or radio). Because the system is connected _{and asynchronous}, the internet allows its users to be more networked than before while being tuned to personal preferences.

The previous chapter addressed the digitization of the music industry and

specifically dealt with the important but inherently biased position of data. In the light of these recent technological changes, we will now built towards a way of approaching the current state of the digital music landscape with a perspective suitable for today’s society. A ‘new logic’ so to say, that fits this information age: the logic of networks.

3.2 The logic of networks

Taking things apart and investigating the components has long been the dominant way to try and understand a whole. We looked at atoms to understand the universe,

molecules for life, genes for human behaviour and prophets for religion. This can be seen as a reductionist perception, with the underlying idea of ‘once we know the parts, we know the whole’. This point of view has been showing flaws and is increasingly being

(28)

challenged. Undoubtedly, we have learned much about the parts, but it has been found difficult to reassemble the parts back into a whole: the functioning of parts appear too complex by themselves and according to some leave us in the dark about understanding the whole. In addressing this problem, another way of understanding is taking root, one that accounts for the complexity of parts and how to relate those to a bigger picture. As stated by Barabási in 2002:

... we increasingly recognize that nothing happens in isolation. Most events and

phenomena are connected, caused by, and interacting with a huge number of other pieces of a complex universal puzzle. We have come to see that we live in a small world, where everything is linked to everything else (Barabási 6).

This is the the basic thought that underlies network theory. Sixteen years further our connectedness and the awareness of it has undoubtedly grown, although the casual and common usage of the word ‘network’ does not mean we have shaken off the reductionist vision. Maybe it is because the idea of being able to explain the whole by its parts feels so comfortable and is relatively simple next to the more organic and flexible network

perspective. But in fact, in the ‘architecture of complexity’ the linked components (in social theory these components are usually called actors, and their links ties) show patterns of connections that are crucial to the behaviour of a system and can indeed be investigated, by representing them in networks.

At first network models didn’t explain reality that well. Faced with complexity, the first abstract mathematical model (Erdős and Rényi, 1959) assumed randomness in ties amongst actors. Yet when we look at actual social networks, random distribution would mean that almost all actors (persons) would have approximately the same number of ties (relations): this kind of randomness is seldom –if ever– the case. Later on, new models and theory emerged to fit research results that showed specific non-random structures: highly connected subgroups within a network, causing clusters (Granovetter 1986, Watts and Strogatz 1998) and after that, actors with an anomalously large number of ties, called hubs or connectors. With ties to an unusually large number of actors, hubs create short paths between any two actors in the system. The internet is a typical example, with the Web showing a complete lack of democracy when it comes to visibility of web pages.

(29)

Studies of the structure of the internet shows how it is being dominated by a few highly connected hubs such as Google or Amazon.com. It forms “the strongest argument against the utopian vision of an egalitarian cyberspace”; those hubs dominate the structure of the network in which they appear (Barabási 64). These same structures tend to show up in other types of networks.

Findings like these point towards certain actors to be more ‘central’ than other. Central actors are often perceived as prominent, influential, as leaders or gatekeepers, or as having great autonomy, control, visibility involvement, prestige, power and so on (Borgatti et al. 164). For example, when talking about human relations, heavily

connected people in the core of a network would be most apt to spread things, like news, gossip or even disease, and those on the periphery have little impact within the network, but they are often important for sending and receiving things to or from other social milieus (Rainie and Wellman 47).

What characterizes network perspective and distinguishes it from other perspectives, is that the focus lies on relations and the way they function instead of determining outcomes. The premise of social network analysis is that social life is created ‘primarily and most importantly’ by relations and patterns in relations (Marin and Wellman 11). Distinctively, network explanations do not assume that environments, attributes or circumstances affect actors independently. Moreover, the network

approach does not assume the existence of uniformly cohesive and discretely bounded groups. This does not mean they can not exist (at any point, many groups do ‘exist’), but an analysis is not dependent of it. Finally, network explanations take context so seriously that relations themselves are often analysed in the context of other relations (ibid 11-13).

The reason why we would look into patterns of connections or interactions between parts, is because it can teach us something about the structure of these networks and it could help us understand processes. Physicist and network specialist Newman explains:

... the structure of such networks, the particular pattern of interactions, can have a big effect on the behavior of the system. The pattern of connections between computers on the Internet, for instance, affects the routes that data take over the network and the efficiency with which the network transports those data. The connections in a social

(30)

network affect how people learn, form opinions, and gather news, as well as affecting other less obvious phenomena, such as the spread of disease. Unless we know something about the structure of these networks, we cannot hope to understand fully how the corresponding systems work (Newman 2).

So when we analyse a (social) network, we investigate how an actor is embedded in structures of relationships that provide opportunities, constraints, coalitions, and workarounds. The visualisation of a network (a type of graph), can uncover limitations and possibilities within that network, as these properties are built into their construction (Barabási 12, Newman 2).

3.3 Basic principles and visualisation of networks

A basic network graph consists of actors (sometimes called nodes, or vertices) and of ties (or links or edges) that connect them, usually visualised as dots and lines (figure 3.1). The actors, or network members, are the units that are connected by the relations

represented as ties whose patterns we can study (Newman 1).

Figure 3.1. Example of a small network of eight actors and 11 ties.

The number of direct connections an actor has, is called its degree. For the actor that appears in the top-right in image 3.1, the degree is 2, since it has two ties to other actors. For the one in the mid-right position, the degree is 5, and so on.

Ties in a network can be directed or undirected. If we take the artist Noisia as an example, a (piece of) network of their collaborations with other artists could look like this:

(31)

Figure 3.2. Example of an undirected network of three actors and two ties.

Because a collaboration between artists goes both ways, a connection from Noisia to The Upbeats means a reciprocal connection from The Upbeats to Noisia: the tie is undirected.

If we now study remixes instead of collaborations, this might not be the case. It is possible that Noisia remixed a track from The Upbeats, but that does not necessarily mean The Upbeats remixed a track by Noisia. In a network graph we can show that through directed ties, such as in figure 3.3. For directed ties, degrees are more complicated. There is a distinction of in-degree and out-degree. The in-degree is the number of ingoing ties connected to an actor and the out-degree is the number of

outgoing ties (Newman 30). In figure 3.3, Noisia has an out-degree of 2 (has remixed two other artists) and an in-degree of 0 (has been remixed by none). The two others have an out-degree of 0 (have remixed no other artist) and an in-degree of 1 (have been remixed by one other artist). Values of in- and out-degree have distinctive characteristics and potentials: “actors who have unusually high out-degree are actors who are able to exchange with many others, or make many others aware of their views. Actors who display high out-degree centrality are often said to be influential actors” (Hanneman 2005). In general, degree can be an effective measure for centrality and power-potential of an actor. More on degree and other centrality measures can be found in Chapter 4.

Figure 3.3. Example of a directed network of three actors.

(32)

Ties in a network can also be unweighted or weighted: if they are unweighted, every connection counts equally. Sometimes one connection is stronger than the other (for example because of frequent collaboration) and that can be represented by

assigning a value to the tie and/or make it appear thicker in the graph. The weighted collaboration network example of figure 3.4 shows a weak tie for one collaboration (Noisia-Black Sun Empire) and a strong tie for multiple collaborations (Noisia-The Upbeats). Weighted networks can of course also be directed according to the same principle: one remix for a weak tie, several remixes for a strong tie. Additionally, actors can also have a self-tie: if Noisia remixed their own track, a tie from themselves, to themselves exists.

Figure 3.4. Example of an undirected, weighted network of three actors.

The relations appearing in a network can tell us something about the overall structure and about the position of individual actors. The actors can additionally be given

attributes, which we can reflect in the visualisation as well. As an artist’s popularity ranking, reflected in the size of the actors for example, as we will see in Chapter 5.

Visualising data on actors and their relations in a network graph can not only give an alluring image, it can also be very useful, especially with larger databases. A network graph allows you to instantly see important structural features of the network, that would otherwise be difficult or impossible to distinguish from lists of numbers (Newman 10).

3.4 An additional network perspective

As described above and in generally in social network theory, social networks take humans as their actors: the social units that form relations with each other. But as

(33)

Newman points out, the classification of networks as social networks, information networks or other kind of network is fuzzy and many examples “straddle the boundaries” (64). For example, we can approach algorithm-based technology as

sociotechnical systems. But where does this fit in network-terms; would these belong to social networks, or information networks? Neither and both. For reasons explained in the previous chapter, we can assume that not only humans play a role in shaping the social. Things too, can express power relations or reinforce social inequalities (Latour 2005, 72). When we want to incorporate this idea in our network-thinking, we need to expand the theory. We need to be able to address the networks that are social and technical at the same time.

Within social network theory sub theories have been developed that build on the general concepts. One of these is actor-network-theory (ANT); a theoretical and

methodological approach described by Bruno Latour, scientist in the field of sociology and theoretical philosophy. ANT is helpful for our problem because it opens up the way to consider ‘things’ as social actors. Latour equalizes humans and nonhumans as both being actors without hierarchy a priori, for the role they play in the course of another actor’s possibilities. When we maintain a division between humans and things, Latour claims we obfuscate an exploration of “how a collective action is possible. [...] an action that collects different types of forces woven together because they are different” (Latour 2005, 75). ANT allows us to navigate through data and make no difference in

investigating a person, a place, an institution or an event. The usage and even mere existence of non-human ‘things’ (of which I consider data and software are also a part) play a great part on how lives and work are shaped. By recognizing this, ANT caters to a more realistic approach of social ties in a digitized landscape, because “...the continuity of any course of action will rarely consist of human-to-human connections [...] or of object-object connections, but will probably zigzag from one to the other” (ibid).

By allowing non-human elements in dealing with social networks, we are freed of concerns regarding ‘fuzzy borders’ of classification: it does not matter when a social network stops being social, and becomes an information network or vice versa. This could be something especially fuzzy when we incorporate sociotechnical systems into the research, as they incorporate human and machine elements by default. Algorithm or online-created networks are human and technical at the same time. Their particular automated constraints, rules and procedures of possible connections, influence what

(34)

these network can, or even should be. Adopting Latour’s ideas makes a lot of sense when we think about the influence ‘lifeless things’ have on ourselves and society. By doing so we can use network theory for more than just human centered social networks, and apply it on the sociotechnical networks such as those we encounter through streaming platforms.

In this paper’s arguments and analysis, Latour’s thinking will complement the more general social network theory that we find in Newman and Rainie & Wellman’s descriptions. We can regard the general network theory as the basis for network

perspective. Latour’s ANT builds on that theory with additional concepts, that are fit for investigating social networks from a particular standpoint: that non-human elements play an equal important part in shaping the social as human elements do. These network theories describe a certain idea of how our reality works, and based on this ‘model of reality’ (a networked society) particular methods have been developed to analyse the world in network terms. A combination of general network theory and ANT will allow for an analysis that incorporates quantitative methods and substantial theory, while staying open to the ambiguous matters (such as sociotechnical systems) that are so important for any study that deals with social matters.

The following chapter will explain the specific methods that are applied in the case study and will also provide the reader with information on data collection and processing.

(35)

4. Methods

This chapter explains the samples and methods used to analyse the case of Chapter 5. To create a coherent picture, several techniques have been combined. First, I will explain the rationale for a case study (4.1), and some straightforward details of the used sample. The second part (4.2) explains which network analysis methods were applied and why.

4.1 Case study: Spotify related artists network

Adapting a network perspective to approach activity and events can be meaningful on many levels, as broadly addressed in the previous chapter. From the manager's point of view, every artist is different and requires their own approach. For that reason it is valid to analyse a specific network by means of a case study of a particular artist. As

previously brought to attention, Noisia is the central artist for the analysis and Flapper, their manager, is the subject who’s possibilities are explored by doing the study . More 4

information on Noisia is provided in the next chapter. The results of the analysis will turn to network theory for explanations. Section 4.2 in this chapter will elaborate on the kind of network analysis that is done. Additionally, a critical assessment in light of datafication will guard against a too simplistic interpretation for all intents and

purposes, which can be found in the discussion (Chapter 6). Since both quantitative and qualitative methods are applied, it is a mixed-methods investigation of the subject.

The sample under analysis is Noisia’s network of related artists. The information comes from the music streaming platform Spotify. The findings from the case study are meant to substantiate an answer to the main question: What possibilities does network theory provide music management? For the case study, the attention will be directed towards the subquestions:

- What is Noisia’s related artists network?

- What insights can be provided by an analysis of this network?

4_{In an early stage of the research I interviewed Walter Flapper, manager of Noisia, to gain a general idea}

of what he thinks about data and online platforms. I refer to some of his remarks in the thesis and therefore the interview is included in the appendix.

(36)

To investigate a particular artist, access to private information is not necessarily

indispensable, but for a more complete, nuanced case, it is quite substantial. To work on the present study, Flapper Management has agreed to provide the case. They recognize the importance of data for artists, but fail to feel in control over it, because there are so many connections that they “simply do not make”, stating: “It's just a big deal of data” (Walter Flapper 2017). Some information used for the case was already publicly

accessible, and additional information granted by Flapper Management has allowed me to access and situate data more thoroughly.

4.1.1 Data collection and processing

The principle source of information is the music streaming platform Spotify. Spotify provides the ‘related artists’ through their API. With the use of a crawler, the related artists, and artists related to those artists, are obtainable (crawler by Bernard Rieder 2017). The resulting file in .gph format contains data on the artists and their relations that makes up Noisia’s Spotify related artist network. The data is one-mode, because all the actors are of a single type –artists–, meaning every actor could conceivably be connected to any other actor. By default, the artists have a Spotify-given attribute ‘popularity’: a value between 0 and 100, with 100 being the most popular on the platform. Every artists on Spotify falls somewhere along that scale, and the value is calculated from the popularity of all the artist's tracks in relation to every other artists on the platform (Spotify for Developers). I have not removed this attribute, as it reveals something about Spotify’s whole artist network, where Noisia is embedded in and relates to, and I will refer back to it later.

The network analysis and visualization software package Gephi can process the datafile as a network. All network graphs in this thesis have been built in Gephi. The overall network is shaped by the Force Atlas2 algorithm of the program. Force Atlas2 uses a formula for repulsion and attraction: without links, the actors repulse each other and spread. The ties work as springs, that draw the actors together, aiming to produce a layout that shows visual densities that denote structural densities (Jacomy et al. 2014).

The Gephi software renders statistics through built-in features, and adds these measurements as attributes to the actors. Three of those have been used for the

(37)

centrality analyses. Since the methods of calculating centrality measures can vary a great deal (Freeman 1978, 217), it is worth to mention that the algorithms used in Gephi come from Ulrik Brandes (2001). In addition to the visual presentation rendered by changing settings in Gephi, I wrote a basic Python code to extract the top centralities and

descriptive statistics from the extensive network data file (see appendix for data and code).

4.2. Analysis: centrality measures

The network approach emphasizes that power is inherently relational (Hanneman 2005) and thinking in networks can contribute to important insights about social power or influence. A prominent way to to investigate power relations in a network is through the structural attribute of centrality, because it is so closely related to “other important group properties and processes” (Freeman 1978, 217). Centrality, though, can mean different things, just as power does not have one straightforward interpretation. To determine an actor’s centrality, we can refer to a handful of different concepts, of which three shall be treated in this thesis. The most popular centrality measurements are investigated, being degree centrality, betweenness centrality and closeness centrality. In the following sections the specific measurements and their value will be explained. For the reason that Noisia’s Related Artist Network from Spotify is an unweighted and undirected network, all explanations are limited to such simple networks (meaning that the values of the ties are all equal; if actor A is connected to B, B by definition is

connected to A. See also 3.5).

4.2.1 Degree centrality

The most plain measure of actor centrality is their degree. Degree defines the number of direct ties an actor has to other actors in the network. Actors with a high degree are highly visible, and tend to be seen as important, to have more influence, access to

information or prestige (Borgatti 166, Newman 169). High degree centrality means their position is advantageous in the exchange of information: more ties usually mean greater opportunities because the actor is believed to have more choices. It is a favored position, because it gives the actor autonomy and makes an them more independent of others (Hanneman 2005).

(38)

Consider the small network below:

Figure 4.1. A small undirected network with five actors and five ties.

The degree centrality for each would be as followed: Actor Degre e A 1 B 2 C 4 D 2 E 1

Table 4.1. Degree centrality for each actor in network from figure 4.1.

Actor C scores highest on degree centrality, reaching the maximum possible degree as it is connected with all others in the network. In terms of communication, we can

understand its position to be powerful or independent because it has a direct relation with (i.e. are known by, or have access to information of) the other actors. Actors A and E are only directly tied to C and thus rely on that one. Between them are B and D, who are not completely dependant on C for ties as they are connected to each other as well.

If these actors were people and if the tie would be a friendship bond, and

someone was in need of advice, we can see how it would be most beneficial to have the position of C over any other. That person could ask four others for their opinion and make a well informed decision, while A and E, with only one tie, could only consult one other: C. Beside C having the most choice (being most independent) in getting advice, this

(39)

actor also has the most influence on the others, since their opinion is for two others all they can access, and for the other two, C is half of who they can consult.

We can imagine the network representing other entities and flows than humans and information, like data packets on the internet or synapses and neurons. The degree of a point for any kind of network is seen as important as an index of its potential activity (Freeman 211) and for that an interesting metric.

4.2.2 Closeness centrality

Closeness centrality gives a very different value than degree, as it aims to define the most central actors in terms of the overall structure of the network. It measures the mean distance from an actor to all other actors. This would basically mean that a higher value means less close. Gephi uses inversed closeness, so that the highest values reflect the most central actors. As such, an actor that has a high closeness score is a short distance from most others. Having a high closeness, an actor will be able to obtain information (or whatever flows through the network) originating at a random actor potentially very quickly. If we have information flowing through the network, the diffusion process tends to introduce distortion as the information has to pass every actor. For that reason we expect information received by central actors to have higher fidelity on average. Thus, a high closeness would seem a significant advantage for an actor to the extent that it can avoid the control potential of others. Logically, shorter distances means fewer

transmissions and depending on the type of network, shorter times and lower costs, better access to information or more direct influence (Freeman 224, Borgatti et al 173, Newman 183).

To get the closeness value, we take the sum of geodesic (or: shortest) distances for a specific actor to all other actors. The higher the value, the more central in closeness terms. If we take the network from figure 4.1 again, actor C has the highest possible closeness, with a path length of {1} to all four other actors. Its closeness is 5/4 = 1.25. Compare that to B or D, with paths of length {2,1,1,2} giving a closeness of 5/6 = 0.833, and for A and E, with path lengths {2,1,2,2}, and closeness 5/7 = 0.714.

Noisia’s network is built from one starting actor (the ‘ego’ Noisia) and is therefore an egocentric network_{(as opposed to a whole, or sociocentric,} network) (Marin and