Navigation – Plan du site

AccueilNuméros17DossierData Science Versus a Sense of Ta...

Dossier

Data Science Versus a Sense of Talent? The Rise of a New Division of Labour in Content Definition in Hollywood

La science des données contre le sens du talent ? L’essor d’une nouvelle division du travail de définition des contenus à Hollywood
¿La ciencia de los datos se opone al sentido del talento? El auge de una nueva división del trabajo de definición de los contenidos en Hollywood
Datenwissenschaft gegen den Sinn für Talent? Der Aufstieg einer neuen Arbeitsteilung bei der Definition von Inhalten in Hollywood
Violaine Roussel
Traduction de Hayley Wood
Cet article est une traduction de :
La science des données contre le sens du talent ? L’essor d’une nouvelle division du travail de définition des contenus à Hollywood [fr]

Résumés

Cet article prend pour objet l’émergence rapide, dans la dernière décennie, de différents types de spécialistes des données intervenant dans les studios et services de vidéo en streaming d’Hollywood, du côté de la production. Ces professionnel·les sont désormais étroitement impliqué·es dans l’évaluation et la sélection des projets artistiques et dans la définition des stratégies de contenus. Cette contribution, issue d’une enquête par entretiens et observations, examine d’abord l’ascension de ce nouveau groupe, indissociable de transformations des organisations d’Hollywood et des arrangements de relations entre différents métiers et rôles en leur sein. Les cercles de relations dans lesquels se joue la définition des choix de production sont progressivement recomposés en conséquence. On fait ensuite la lumière sur les logiques qui opposent ces spécialistes des données aux professionnel·les traditionnel·les de la production : leurs différents profils et parcours déterminent des modes de définition professionnelle et des rapports à la création et aux processus de production qui contrastent avec ceux des producteurs et productrices. On met enfin en évidence l’existence de mécanismes de convergence des perceptions et de conversion mutuelle qui s’esquissent néanmoins entre ces groupes en tension.

Haut de page

Texte intégral

Figure 1

Figure 1

Credit: Pixabay, Violaine Roussel.

  • 1 Netflix numbers for the second quarter of 2023. See Julia Stoll, “Number of Netflix paid subscriber (...)
  • 2 The figures for Amazon Prime Video vary depending on the source and, most importantly, how they are (...)
  • 3 Max was created in May 2023 from the merger of the HBO Max and Discovery+ platforms, under the Warn (...)

1In recent years, video streaming services have spread rapidly around the world. In 2023, the leading streamers Netflix and Amazon Prime Video hit global subscriber figures of 238 million1 and 117 million,2 respectively. They now face competition from the numerous streaming services that have been launched more recently by the traditional studios and channels, such as Disney+ and Max,3 along with smaller competitors such as Paramount+ and AppleTV+. In the space of just a few years, streaming has become the main distribution channel for television products, if not for entertainment in general, and this trend has been further accelerated and intensified by the COVID-19 pandemic. Since 2012–2013 the big streamers have also begun to produce their own original content, alongside their distribution activities. They have invested heavily in this original content: between 2013 and 2021, Netflix alone produced around 1,630 titles, and became Hollywood’s leading film production studio. Streaming companies have also received artistic recognition from the authorities responsible for aesthetic judgment in the industry, including in the form of awards such as the Oscars. They have thus acquired a central position in the Hollywood game that was formerly occupied by the big studios, to the extent that they can be seen as the new “majors.” Meanwhile, the traditional US studios have pivoted to producing big-budget superhero films based on franchises, such as those in the Marvel Universe, allowing the streamers to position themselves as the go-to contacts for all other types of projects.

2It was within streamers, these new central organizations in Hollywood production, that various types of data specialists first began to emerge. These professionals are responsible for collecting and processing data on the behaviour of platform users, and comparing it to the types of content consumed in order to predict the success of potential projects. Here, the purpose of data analysis is therefore to inform decisions about what to produce or acquire, and to determine what types of content a particular streaming platform should make or distribute. These specialists are thus closely involved in the selection and making of content. It is this, rather than the fixation on predicting the success of film or television projects, that is new to Hollywood, since until very recently the authority to decide what would work and what should see the light of day was largely monopolized by the producers and studio heads who steered the greenlight process. This is now changing as data specialists establish a central role in the content definition process, and stake a claim to be able to influence what should be produced and offered to consumers, on the basis of different logics and registers of expertise.

3In the past decade, these specialists have flooded into Hollywood organizations, where they are now employed in their thousands. In 2013, when Netflix began producing its original titles, it had just two data scientists working on content. By the end of 2018, this team had grown to 50 people, and three years later, it had 150 data specialists dedicated solely to working on content. By the summer of 2022, this group, now merged with the Consumer Insights data specialist team under the authority of a new vice president, had a total of 400 employees. At the same time, and interdependently, the company’s Content Planning and Analysis team, whose data analysts interface directly with the more traditional production professionals, has also grown. When Netflix launched its new production business, this was a team of around 15 people, but it expanded rapidly, to over 90 employees in summer 2019, and a remarkable 500 by spring 2022. In the space of a decade, the company therefore went from having a handful of data specialists to employing 900 data experts working on content. These organizational changes are so significant that over this period of time, Netflix might even be considered to have been two different companies under the same name. A similar shift has also taken place, albeit to a slightly lesser extent, at Netflix’s competitors, notably Amazon Prime Video, as a result of the competition between companies in the sector.

4Although the presence of these new professional groups in Hollywood has been noted, these specialists have not hitherto been the subject of systematic sociological analysis. This article aims to fill this gap by analysing the shake-up in power relations and the reorganization of the division of labour associated with the rise of these new specialists involved in creative decision-making. The emergence and establishment of this new group on the content side of the industry is inextricably linked to changes in Hollywood organizations and the configurations of relationships between the different occupations and roles within them. This article aims to show how they have gradually led to the emergence of a new “negotiated order” (Strauss 1978). It reveals the tensions and struggles that have accompanied the rise of these new groups within the Hollywood professional space, but also explores the mechanisms by which they have gradually established themselves, acquiring forms of legitimacy to influence creative decisions, and developing increasingly routine ways of cooperating with traditional production professionals.

  • 4 See, among many others, Jeanpierre & Roueff (2014); and Lizé, Naudier & Sofio (eds.) (2014). For a (...)

5This analysis of the emergence of data specialists at the heart of the Hollywood “dream factory,” and the way in which this is affecting production decisions, draws on both the sociology of intermediaries in the cultural industries, which has developed in recent years to analyse the work that takes place behind the scenes in these spaces,4 and more broadly on the sociology of professional groups, a field that as Didier Demazière and Charles Gadea (2009) observe is once again growing in France after decades of neglect. In particular, this study follows on from work focusing on statistical or technical professionals, which has highlighted the transformative effects of their rise within institutions and organizations that, in various ways, govern our behaviour and shape our representations (Bigo 2020; Desrosières 2002 [1993]; Dodier & Barbot 2016).

6I approach the changes in professions in film and television related to the development of digital technology and streaming not with a view of technology itself as the mechanical determinant of change, but instead focusing on the professional practices that shape technological changes (Roussel 2023). Different types of technology specialists operate in the film and television industries, and have done so for a long time, particularly in Hollywood. While their key role in transforming creative processes has been examined (see, for example, Marzola [2021] on the role of technicians in developing the studio system), within Hollywood organizations they have always been kept strictly separate from production professionals, who decide which projects should be made. Their story is the story of Hollywood’s “technical” personnel, a category constructed in opposition to the sphere of artistic decision-making. In the field research explored in this article, however, “technical” activities take place within the organization, and influence production decisions.

  • 5 For a more detailed overview of this work, see the introduction to this issue.

7Finally, this article engages with a number of studies exploring how streaming has transformed the distribution, production, and consumption of content, which have been published over the last ten years in the fields of sociology, film and media studies, and communication.5 Netflix has attracted particular attention (Jenner 2018; Lobato 2019; Lobato & Lotz 2020). However, this work has mainly focused on recommendation algorithms and how they are used by audiences (Delaporte & Mazel 2021; Drumond, Coutand & Millerand 2018; Frey 2021). This research forms part of a broader body of work exploring the contemporary development of the power of metrics (Beer 2018; Denis 2018), the issues raised by the massive collection of personal data (Lyon 2019), and the ways in which algorithms govern behaviour, and resistance to them (Ananny 2016; Christin 2017). Yet despite calls to take “data work and data workers” seriously (Bastard et al. 2014), very few studies have focused on the activity of the professionals behind the power of data and algorithms—the people who shape these tools and frame how they are used by others. This work remains largely invisible and, as Rémi Rouge (2023) notes in relation to the professionals involved in digital memory apps, its invisibility lends credence to the idea of the algorithms and technologies used by the companies operating in these sectors as having their own power: it thus helps make data seem magic.

Methods

This article is based on 75 in-depth interviews conducted between 2017 and 2023, either in person in Los Angeles or by videoconference (during the COVID-19 pandemic). They involved various groups of data specialists—people who were working or had worked for streamers, studios, independent production companies, companies specializing in data analysis for film and television content, and talent agencies—along with some of their direct interlocutors on the production side of the industry. The data specialists I interviewed held a range of positions and functions in Hollywood organizations, with most of them working as data scientists, engineers, or technicians within data science teams, or as content analysts and strategists on the content side, and a smaller number working as specialists in marketing and research or consumer insights departments. Below, I provide more detail about how roles are divided between these teams and positions, in relation to the social characteristics of the interviewees.

I often spoke with these professionals on several occasions, for periods ranging from one to over three hours. The relationships I developed over time with several of the interviewees enabled me to track changes in their activities, perceptions, and working environment, and I often kept in touch with them as they changed jobs and sometimes employers. The relationships of trust that I built up and maintained in this way gave me access to confidential information, and to spaces that are not generally accessible to researchers.

As a result, I was able to conduct two observations in situ in Los Angeles: one in March–April 2022, within a team using data analysis to evaluate projects (mainly television series) and recommend changes to their content, and the other with a data scientist working for one of the big streaming services, who allowed me to shadow him on several occasions between spring 2021 and summer 2022. During the pandemic, since face-to-face meetings were impossible and professionals could not work on company premises, I conducted interviews in the form of a “calendar review,” which involved asking my interviewees to look back at their work calendars and go over everything they had done during the previous week, and discussing these interactions and experiences in detail. These interviews were often very long, and gave me greater insight into both the organizational and relational dimensions of the rapidly changing space in which my interviewees were working.

8I begin by exploring the effects of the emergence of data specialists in content-related roles within streamers on the structural configurations of these organizations, but also on the circles of relationships in which production decisions are made. I then examine the logics that pit these specialists against traditional production professionals: their different profiles and career paths determine modes of professional definition and relationships to content, the creative and production processes, art, and artists that differ from those of producers. Having demonstrated the conflicts over legitimacy that pit these different categories of Hollywood professionals against one another, in the final part of this article I reveal the mechanisms of converging perceptions and mutual conversion that are nevertheless taking shape between these groups in tension.

1. Data Specialists on the Content Side: A Redistribution of Roles

  • 6 This followed the decision by studios to pull most of the content they had initially licensed to st (...)

9The commitment of the original streamers to producing their own content, around 2013,6 led to the creation of the first data specialist positions within production and content teams, in particular at Netflix and Amazon Prime Video. For the executives at these companies, who came from Silicon Valley and the tech world, it was common to include data specialists in the decision-making process, but for Hollywood professionals, who associate creative decisions with the almost sacred dimension of artistic creation, it was perceived as a takeover.

1. 1. The Shifting Contours of Evaluation Communities

10From the outset, the organizational structure of streamers originating in the tech world has therefore differed from that of the traditional studios in terms of the positions held by data specialists in the chain of project decision-making, and in their economic and symbolic hierarchies. In these companies, data specialists are involved from a very early stage in the production process, that of deciding which programmes, series, or films should be bought or produced. They are key interlocutors on a day-to-day basis for those responsible for acquiring content, namely the traditional producers, who were originally often poached by streamers from the major studios or from successful independent companies.

11Data specialists occupy two different types of role, which across all of the streamers have gradually been structured into two distinct but interdependent departments: first, the data scientists who develop the algorithmic models and tools that are used to define the criteria and general framework for gathering metrics; and second, the specialists who work more directly on developing content strategies and assessing the economic value of projects, in close collaboration with the traditional Hollywood producers. This second type of role, which has various titles—including ‘content strategist,’ ‘content (finance) analyst,’ and ‘content planning specialist’—is a key intermediary position between the data scientists and traditional producers. One streaming executive described his work and its scope as follows:

We work very closely with [content executives] in helping them make decisions on how to programme the service and which content to select for either production or acquisition or license. So, we are between the more hardcore scientists and the end users of the data that are the content team. So, basically, we help [the data science team] build the models with an input on signals and how to model something, etc. They are the ones that are mathematicians and statisticians, and then, once they’ve got these models, we use those models to then translate that into a business advice. So, the typical process in a show would be: we get a pitch idea for a show, the creative executives get excited about the idea, they think it is a great idea, they send it to us, we run the models, we try to estimate how big the show could be, and then, based on that data plus other strategy considerations, we tell the executives: “Hey, we think it is a good idea” or “not a good idea.” And we work with them in basically the decision of: do we make it or not, and what’s the price that would be best for it? (Content analyst, Netflix, May 2019)

12The high degree of interdependence between the data science and content analysis departments is thus central to the production decision-making system. As noted in the introduction, these departments also grew very rapidly in the decade after the early streamers committed to making their own original content.

  • 7 With the notable exception of Sony.
  • 8 21st Century Fox and NBCUniversal, plus TimeWarner and Disney, before Disney became the sole shareh (...)

13Following the COVID-19 pandemic, the distinction between studios and streamers quickly blurred, as the streamers first became studios by investing massively in original content, and the studios then turned into streamers7 by systematically developing their own video-on-demand services (such as Disney+, HBO Max, Discovery+, Peacock, and Paramount+). By launching their own streaming services, the studios entered into direct-to-consumer sales, enabling them to capture the associated user data. But despite this new ability to seize data and use it to transform decision-making processes, the studios did not immediately adopt practices or organizational changes making them more like the original streamers of Netflix or Amazon. Hulu, the streaming service launched in 2008 by a coalition of studios,8 provides an example of the initial resistance of these organizations to using data for content definition. It took longer for a data science team to emerge at Hulu, and for a long time its scope for intervention in creative decision-making remained more limited.

14This is less due to organizational resistance to change than to shared beliefs in Hollywood about the creative process and the professionals who should drive this forward, in which the producer or studio executive is central to taking decisions about what should be made. In this model, these professionals are responsible for deciding what viewers will want to watch, and what should therefore be offered to them, based on their relationships with artists, who are considered to have an aesthetic and emotional connection with their audience. The projects that see the light of day are negotiated with talent agents who, like production professionals, claim to have a “sense of talent” and an intuitive ability to identify “good projects.” Other work has shed light on the activity of evaluation communities, formed of agents, artists, and production professionals, that define “what will work,” meaning both the artistic and economic value of “talent” and projects (Roussel 2017). The processes of change examined in this article represent a radical reconfiguration of these evaluation communities, with new groups of participants playing an increasingly central role.

15It is therefore not streaming as a technology that has automatically and naturally resulted in the centrality of data science in content definition, but rather the power relations established between different professional groups within Hollywood organizations, which take different forms depending on the history of these companies and the professional representations that prevail within them. The traditional studios and their streaming services are thus characterized by a different relationship to the creative process, production, and talent which has meant that data specialists have played a more limited role on the content side, and that statistical tools for rationalizing production decisions have been used in a different way (at least in the initial phase after the launch of these streaming services in the early 2020s). Conversely, in companies originating in the tech sector, such as Amazon Prime Video, AppleTV+, and Netflix, different relationships to data, organizational structures, and professional cultures have shaped different sociotechnical arrangements, and have given data specialists a key role from the outset in making decisions about what to produce or acquire.

  • 9 In this sense, the professional group of data scientists is much broader than that of Hollywood dat (...)

16The conditions for the emergence of data specialists in Hollywood, and for their influence on production decisions, thus differ based on the organizations in which this rise has taken place. However, the interdependencies related to competition between content-producing companies and the movement of professionals between these companies have gradually generated dynamics of convergence leading the specialists examined here to play an increasingly decisive role in Hollywood organizations as a whole. The logics of relationships and competition between the different categories of professionals involved are crucial to understanding the emergence of this composite group. As we shall see, the data specialists who have carved out a place for themselves in Hollywood often defected from other industries. While they have drawn on existing forms of professional recognition and skills attached to the broader category of “data scientists,”9 their work in Hollywood has also gradually defined the contours of a new professional category, specific to the film and television space. They have established themselves on the content side via confrontations and jurisdictional conflicts with the categories that had previously held sway over creative decisions. However, this was not—to use the concepts developed by Andrew Abbott (1988)—the result of a situation of “vacancy chain,” in which one group takes up a task left vacant by others, nor did it derive from a more offensive “bump chain” strategy, in which it actively disputes this territory with others. Instead, data specialists appeared to be shaping a new space of intervention, associated with equally new decision-making logics and tools, within the chain of activities involved in content creation, and thus producing a reorganization of the division of labour. Nevertheless, this has redefined the boundaries delimiting the jurisdiction over content definition in the particular art world that is Hollywood (Dubois 2021).

1. 2. Acting as the Voice of Users

17The traditional studios were not, of course, completely immune to the logics of using numbers to rationalize decision-making. The demand for “digital transformation” has shaped these organizations for many years, and to keep their jobs, the studio heads had to prove to the major communications and media groups to whom the studios now belong that they were capable of making the technological and economic shift into streaming. They initially tried to do this by appointing technology or digital consultants to work alongside them, but this was more of a symbolic gesture than a genuine change in how the production process was organized. Although the greenlight process remained largely in the same hands, data specialists were not absent from the studios, but were—and to some extent still are—confined to positions in the marketing and research departments. Their involvement thus only began at the end of the process, once a film had been produced, in the context of developing a promotional strategy. Excluded from the greenlight decision itself, they occupied symbolically dominated (because considered less creative) positions in the studio hierarchy. This was reflected in my interview with a producer who held a senior position in a studio:

For [us], where data is relevant is in marketing, we use data a lot to market and to reach our consumers, right? Because advertising is much more digitally driven now and so we can be more efficient with respect to our marketing spends. So that’s the place in which we use data. We don’t use algorithmic data to decide what movies to make. Because, without sounding too pretentious, I’m still trying to make art. And art is not the product of an algorithm. (Studio executive, April 2021)

18From the late 2010s, executives and department heads in these marketing and research teams began to take advantage of the demands for digital transformation within the studios to present themselves as leading the way in this area. Since they worked in the only departments within the studios whose expertise was precisely in collecting and processing data, these specialists monopolized the ability to speak the language of data, and had everything to gain from showcasing this expertise by promoting its use, and acting as the voice of subscribers in the same way as their counterparts at the streamers. By capitalizing on their know-how in producing and processing data and algorithmic models, they reinvented their own role. Some used these strategies to rise to a higher position in the studio hierarchy, including by calling for the creation of new roles, and securing their own appointment to them. These professionals thus often found themselves at the forefront of managing the new streaming services that were launched or developed by the studios in the early 2020s.

19Within the evaluation communities in which the fate of film and television projects is decided, different categories of data specialists are thus coming into competition with those who previously monopolized the authority to speak on behalf of audiences. What is at stake here is who (how, and with what tools) can legitimately and effectively say what the consumer wants: is it the artist, in connection with the agent and producer, or the data analyst and those who mobilize the authority of numbers and algorithms to influence what is produced and distributed? The claim to speak on behalf of the public is evident in the words of the following data scientist, who was working in a senior role at one of the leading streamers:

We can look at the data and see what our members are actually doing on the service and then come up with insights that guide the creators on what to do next, and so on and so forth, or even what’s happening right now, on that particular show. So, we definitely think what we do is lending a voice to the member, and what the members are doing, it is definitely one way to think about consumers coming to the table, as part of the decision-making process. (Data scientist at one of the big streamers, December 2021)

20Central to these claims of legitimacy to intervene in creative processes based on different principles and tools from those of traditional decision-makers, and more generally, central to the transformations of the Hollywood game, have been phenomena of convergence and objective alliance between different categories of data specialists: those working for traditional studios, and those working for the original streamers. Over the last few years, these professionals have moved between the positions available within these different organizations, gradually forming an interknowledge group linked by common points of reference, know-how, standards, and guidelines, which are disseminated when these specialists move on to continue their career in a new organization.

2. Career Paths and Tensions Between Professional Definitions

21The mechanisms examined above are gradually shaping the contours of a new group in Hollywood. Examining the characteristics and career paths of these new participants in the production process sheds light on their perceptions of their profession and of the worlds of film and television: their backgrounds, perceptions, and interests are very different from those of the traditional Hollywood professionals. This affects the way in which artistic projects are evaluated and selected, how content strategies are developed, and how content is ultimately made.

22Within Hollywood organizations, tensions can thus be observed between the old ways of defining oneself as a producer and the claims of these new data specialists, who now have a say in what is produced or acquired. These two categories of professionals have pursued separate career paths, and their professional socialization has taken place in different environments. This has given them distinct modes of professional definition, and practices and ethos governed by very different norms and routines.

2. 1. Who are the Hollywood Data Specialists?

23Any attempt to collect systematic data on the profiles and career paths of Hollywood data specialists, let alone compare them with those of the other groups of content professionals with whom they interact, is fraught with obstacles. These data are not publicly available, and the activity of data-related teams is considered sensitive and protected from external eyes. Based on my interviews, however, I have been able to collect enough information to sketch out the contours of a composite group about which little was previously known. I have notably drawn on my previous work on Hollywood talent agencies to compare these profiles and career paths with those of talent agents and producers, who often start out in agencies (Roussel 2017). This information helps to shed light on and compare and contrast the perceptions and beliefs of these individuals, along with a number of their practices.

  • 10 Data were available for 70 of my 75 interviewees.
  • 11 Master of Business Administration.
  • 12 Information taken from the CareerOneStop.org website, supported by the Employment and Training Admi (...)
  • 13 Bachelor of Arts.
  • 14 Bachelor of Science.

24First and foremost, data specialists form a group characterized by high levels of education.10 In many cases this cultural capital has served as a gateway to the United States (US) and Hollywood for these specialists, many of whom hold master’s degrees, MBAs,11 and even PhDs, often from prestigious American universities. 18.2% of my interviewees had a PhD (in neuroscience, industrial engineering, communications, IT, or economics), and this was particularly common for those in executive positions or heading up data science departments. In contrast, the Employment and Training Administration at the DOL estimates the proportion of PhD holders in the “Producers and Directors” category to be 2%,12 and the proportion with a master’s degree to be 15%, compared to 51.5% of my interviewees (more than half of them an MBA, the others being divided between master’s of science in economics, computer science/engineering science/computational information science, and master’s of arts in film or media studies). Only 30.3% of my interviewees had only a BA13 or BS,14 and none were self-taught, whereas the US Department of Labor reports 82% of “Producers and Directors” as having a BA or a lower qualification. In the US, the four-year BA course is typically generalist, covering a wide range of disciplines (history, psychology, political science, film, media studies, economics, marketing, and IT). Bachelor’s degrees in arts and humanities or communication and media were not particularly overrepresented among my interviewees.

  • 15 In the absence of other available data, these percentages are calculated based on my study populati (...)
  • 16 23.7% were from Europe (UK, Spain and France, although people from France may have been over-repres (...)

25The second key way in which the group of data specialists contrasts with the traditional Hollywood professionals who have come up through talent agencies is that it includes a large proportion of racialized people (44.7% of my respondents15), who are very much in the minority among agents, and professionals who have taken this traditional route. 55.3% of my interviewees were recent immigrants (first or second generation), and 39.5% were first-generation immigrants: the majority from Asia (57.1%), in particular India, followed by Europe.16 A significant proportion (36.8%) were also women. In terms of gender and race, the characteristics of this group are therefore usually those of people who occupy dominated positions within Hollywood.

  • 17 According to 2019 figures from the Silicon Valley Institute for Regional Studies, 64% of tech profe (...)
  • 18 For example, the most senior positions in the data science teams at Netflix and Disney (until the s (...)

26These two observations are connected: the acquisition of these advanced degrees, often from prestigious American universities, is what gave people who had recently arrived in the US an attractive profile in the eyes of their Hollywood recruiters. This enabled them to convert the cultural resources they had begun to accumulate in their countries of origin into “skills” that are recognized and sought after by streamers. Their profiles and backgrounds are thus far removed from those of the “click workers” studied by Antonio Casilli (2025 [2019]), but are more like those of the Silicon Valley professionals examined by Olivier Alexandre (2023).17 This group has its own internal hierarchies: senior positions such as vice president, senior vice president, and department head are mainly held by white Americans—often women18—or immigrants of European origin (though this is less the case in the traditional studios, where these positions are also less influential, since as discussed above, they do not really influence content definition).

27Finally, data specialists often have early professional experience, or even a longer early career, in a field outside the world of film and television that has therefore helped shape their initial professional socialization. This is especially true of people who were involved early in the process analysed here (from the mid-2010s onwards) and who now hold leadership roles in the departments concerned. Those with training as engineers or specialist technicians, or a PhD, often started out as consultants in IT transformation, or in developing data architecture models for different types of company, while those with a background in mathematical economics and statistics, and sometimes an MBA from the most prestigious programmes in the US, have built up professional experience in the worlds of trading, finance, or insurance. The first group joined the streamers’ data science teams, while the second group joined their content analysis and strategy departments.

2. 2. What it Means to Work in Hollywood

28The data specialists studied here thus have very different backgrounds from the classic Hollywood career path, which involves learning the ropes in the studios and major agencies. Their representations of themselves, of what is professionally possible and desirable, and of the world of film and television, thus stand in contrast with those of traditional Hollywood professionals.

29The latter are largely the product of a system of on-the-job training, in which they have had to prove themselves by working their way up from the lowest rungs of the ladder and, if possible, being mentored by more established professionals. The data specialists I interviewed for this study were not familiar with these modes of professional socialization, which were described by the talent agents I interviewed for my previous fieldwork (Roussel 2015, 2017). By learning the ropes from more senior professionals, trainee agents and producers were placed from the outset at the heart of networks in which the fate of Hollywood careers and projects is (or was) decided. This enabled them to gradually form ties with artists, and learn how to do the relationship work that would enable them to establish lasting links with their key contacts within the evaluation communities in their area of specialty, and steadily build up a reputation. Within these processes of professional socialization, they found their bearings in the Hollywood game, and developed an understanding, beyond the formal hierarchies and explicit rules of the game, of how forms of power are informed by the history of relationships and the past successes of the individuals involved. The professionals who came up this way have also assimilated the pivotal importance of art, aesthetic quality, and talent that is central to the “illusio” in this space (Bourdieu 1980 [1977], 1996 [1992]). This represents both a genuine shared belief that has often motivated the decision to work in Hollywood, and a way of justifying practices or decisions that are sometimes inspired by other factors.

  • 19 In this sense, their “individual experience” echoes that of people working in the employer groups s (...)

30This brief overview shows the gulf between these modes of socialization and professional representations, and those characterizing the data specialists who are the subject of this article. The latter, recruited by organizations originating in the tech sector that do not value experience in the world of film or television (this was especially true in the early days), often do not express a vocation or even a desire to work in a creative art world. Unlike Hollywood professionals, who are deeply attached to the idea that they work in a wholly unique industry characterized by the magic of art and creation, and in a sphere that demands unique abilities and complete commitment, data specialists value their skills as being transferable from one professional sphere to another, and separate the professional identity they ascribe to themselves from the sector of activity in which they operate:19

This domain of expertise is a very portable skill. For example, myself, in my prior job I was working on data for the telecommunication industry, with marketing data, the airline industry, and then, prior to that, I was doing brain data because I was a neuroscientist. I literally worked with human data that was recorded from human brains, recorded from monkey brains, from even locusts and insects and stuff like that. So, it’s a very portable skill to that extent. (Data scientist at one of the big streamers, November 2018)

31Similarly, the following content strategist saw his transition from working for a hedge fund to a position with one of the major streamers as essentially doing the same type of work in a new environment:

So it was really similar to what I was accustomed to, it was using all the data—which in the hedge fund world is the stock market, but in the streaming worlds, it’s the internal data—to try to understand what’s the value of different content, what content to invest in, how do you create the equivalent of a price-to-earnings multiple in the internal streaming world, to help decision making. (Content strategy executive, May 2021)

32Many of my interviewees thus emphasized that their work is and should be “industry-agnostic,” and that the level of the individual project (a particular film, television series, or programme) is irrelevant to them. The relationship work that is such a core component of the activity of traditional Hollywood players appears to be a foreign concept to them. They find it difficult to understand and do, and often refer to it in an uneasy or negative way as “Hollywood politics” or “theatrics.” The performative dimension associated with work in Hollywood—which one “super-agent” I interviewed for my previous work summed up by saying, “it’s a show, it’s show business” (Roussel 2017)—was thus portrayed by one data scientist as empty posturing:

You can’t stand up and make a pretty argument, […] get up in front of a crowd and smile and do a little bit of theatrics and, you know, do a polished presentation and convince people. That doesn’t work anymore. I think of a lot of executives who just one-hundred-percent ride on relationships and their ability to sway and influence… (Data scientist, HBO Max, April 2021)

33Conversely, for the traditional Hollywood producers, being part of an art world, making films, and maintaining relationships with artists is neither trivial nor comparable to working in any other economic sector. Their value in their own eyes lies precisely in the fact that what they do is not “industry-agnostic.” They see the logics that govern the selection of projects as linked to experience and knowledge of reputations built up over a long period of time, and to the relationships forged with artists and other Hollywood professionals. Interviewees presented these skills as reflecting their artistic sensibility and their own talent, positioning their “eye for quality” in opposition to using data to decide what should be produced. The skills they value are not based on standardized metrics, but rather on the education of taste over time, via relationships and transactions, as made manifest in the elusive form of intuition:

[In the traditional studios], there’s a few individuals with a track record working there who can say… whose job is to judge quality. From my viewer’s point, the vast majority of what Netflix does […] is just really low quality. They’ve got a tremendous quality problem. And they can keep pointing to metrics and say, “Hundreds of people will watch it,” I think that’s a little bit thin… […] When you talk traditional studio, the big question they always have is, is it theatrical? Can it get people to the theatre? And then, they also think about, is it really good? And Netflix, it doesn’t feel like they think about is it really good. (Independent producer, April 2019)

34These production professionals speak in terms of art and their relationships with artists, in opposition to the commercial strategies of the streamers, which prioritise “volume over quality,” in a mirror image of the data specialists who present themselves as the voice of consumers:

The streamers’ choices are very data driven, […] their choices are much more data driven than creative choices have ever been in the past. By the way, I’m not sure that’s a great thing, I’m not sure it’s the best for art that the lowest common denominator of an algorithm is making a decision about what you watch. You got, you know, twenty Adam Sandler movies on Netflix. But, for me, as a movie studio, it’s kind of good, because they’re mediocrity. Streamers are full of mediocrity and competing with mediocrity, and I’m in the quality business. (Studio executive, April 2021)

35This emphasis on quality and the relationship with art does not, of course, prevent commercial logics from prevailing among producers or studio heads, but rather indicates two contrasting repertoires of perception, professional definition, and legitimization. For the professionals from traditional backgrounds, this is sometimes compounded by the feeling that their positions are vulnerable, and that they are on shaky ground. Since their skills are subjective and relational, they fear that the rationalization tools used by data specialists will make them vulnerable or even “obsolete”:

I think what happens is that people in the marketing departments love that, but creatives think that it’s the end of the business because the robots are gonna be, you know, running the place, so... (Head of an online intellectual property and production sales company, May 2021)

36Some fear the devaluing of their particular resources, and above all the loss of value of the relationships with artists from which they have previously derived their respect and power. They worry that their ability to influence what is produced will now be subordinate to data, i.e. the data specialists, and that the latter will be able to expose (through metrics) their possible failures or partial failures to their superiors or colleagues in other departments. Within Hollywood organizations, tensions therefore exist between professionals with contrasting backgrounds, socialization pathways, norms, and skills, who are engaged in conflicts over (loss of) legitimacy.

3. The Drivers of the Stabilization of New Configurations

37The coexistence of professionals with such different backgrounds and representations within the teams selecting, producing, and programming content initially resulted in mutual incomprehension, as the individuals involved often reported in my interviews with them at the turn of the 2020s:

The problem in Hollywood is that Hollywood people don’t understand technology, really, aside from visual effects. And technology people don’t understand storytelling. (Product designer, Hulu, July 2019)

38Despite this, Hollywood organizations have also been the site of mechanisms of mutual conversion and influence that are gradually bringing together the two groups contrasted above: traditional producers and data specialists. The stabilization of new configurations can therefore be seen across all Hollywood organisations involved in production, from the original streamers to the studios and their more recently launched platforms.

3. 1. The Translation Work of Intermediaries

39Traditional producers with no training in, or particular affinity for, data analysis found it difficult to interact directly with the engineers, technicians, and scientists working on algorithmic models, since understanding their language and way of thinking did not come naturally to them. As a result, one of the key jobs of department heads was to identify what one interviewee called “ambassadors,” who were tasked with maintaining dialogue and communication between what those involved often described as two opposing camps: the production heads, in contact with the artists, and the “nerds,” responsible for developing algorithmic tools and using these models and instruments to guide content decisions.

I knew how to work with data and I knew how to work with analysts and I knew how to communicate to executives, so I kind of played that middle role and, when they hired me, they said: “Well, look, we want someone who can come in here and basically sit in between the nerds and the executives, because right now, the nerds and the executives, they don’t get along very well, like, the executives will say one thing and the nerds will spend two days doing something and just… They didn’t answer the question, so, we want you to play that ambassador role.” (Data scientist, HBO Max, April 2021)

  • 20 To develop these abilities, the early data science teams first had to learn and adapt to situations (...)
  • 21 Rather than considering algorithms or data to be the “actors” in this process, my focus here is on (...)

40The ability (acquired through specific career paths20) to speak these different languages and translate the logics of these different groups of employees has thus become a strategic skill that enables those who have it to rise within the organizational hierarchy. What we see here, embedded in organizational logics, are processes of translation, in a similar sense to that described by Michel Callon (1984), whereby actors who are initially unable to communicate gradually engage in transactions (or even share common visions of the world around them, and of the importance of what they are helping to produce), as a result of activity and specific work intended to stimulate cooperation.21 The head of one data science team described this translation work and the value of the associated know-how to me as follows:

The level of data science [here] is pretty sophisticated and so it’s hard to understand for people who don’t have that background, […] it[’s] just sort of impenetrable for a lot of people, it feels intimidating to that group of people, and in the average conversation with a data scientist, it’s like they’re speaking two foreign languages to each other. So, a lot of my role and the role of the leaders on my team is to help translate and help coach the teams to translate. So, it’s like, “we build this very complex machine learning model, but here’s the only thing you need to know about it: it helps to predict whether this title’s going to be more successful than this title, or actor A is worth spending money on vs. actor B, and these are the types of decisions the work informs.” So, we have to build enough trust in the model by talking about a little bit how it works, what information we take into account, how confident are we in the predictions, but that’s all in language and translation […] so, in some ways, we have to recruit a special type of person to the teams who work with content folks, because you have to do that translation really well. (Head of a data science team, July 2022)

41These specialists, who are positioned at the interface between technology and creative decision-making, are often part of the “intermediary” teams liaising between the data science departments and the teams responsible for content acquisition or original productions: the content strategy, planning and analysis departments, which have a major influence on production decisions and on determining the value and cost of projects. These employees typically have a background in economics and statistics, and some familiarity with analysing algorithmic data based on their previous roles. This enables them to talk to the other group of data specialists discussed above: those with a background in digital science and technology who populate streamers’ data science departments and work across artificial intelligence, machine learning, and the development of algorithmic models (designed to determine content strategies, or even the details of projects developed by a streaming platform, but which are also behind the operation of content recommendation systems or the user interface). The specialists who do the translation work described above also have specific backgrounds: they draw on the basic knowledges of coding and data manipulation that they have acquired in previous positions (often as consultants) in their dealings with data scientists, but also make use of their (early career) experience in television or film production in their relationships with traditional production professionals. It is this specific combination of experience that provides them with the resources to carry out these translation activities, which they are also able to do effectively because of their intermediary positions within streaming companies.

  • 22 Content analyst, Amazon, May 2019.
  • 23 Content analyst, HBO Max, June 2021.
  • 24 Content analyst, Disney, November 2021.

42These intermediaries therefore know “how to talk with creatives”22 and how to “packag[e] the insights”23 in a way that is palatable and effective for producers. This involves not only adopting language that is familiar to Hollywood ears, but also framing the data with narratives that “speak” to production professionals. In this way, data specialists work to “come up with a narrative”24 that makes sense to them. This storytelling work is not just a roundabout way of gaining acceptance for new modes of decision-making, but also an indicator of the continued importance of traditional Hollywood stakeholders, as well as their beliefs, values, and frame of reference (or illusio), which have long characterized this space. Ultimately, these efforts to be heard and to influence creative decisions do not simply reflect recognition of the skills specific to producers, but also lead data specialists to incorporate elements of these frameworks of judgement and perception into their own work. Quietly, these frameworks are merging, and mutual conversion between these two categories of professionals is taking place.

3. 2. Dynamics of Mutual Conversion

  • 25 Head of the data science department at a streamer, July 2022.

43What data specialists are gradually learning to do, initially within the organization that employs them, is therefore a form of relationship work designed to establish trust, which involves adopting traditional Hollywood categories of thought and action, such as the central importance of intuition. Another aspect of this translation work consists of identifying allies among the professionals “on the content side” who seem most open to using data and who can become, as the head of one data science department put it, “evangelists because they speak both languages, from their side.”25 The clearly strategic dimension of this work was demonstrated by another data analyst, who during our interview described his tactics for “educating” producers and converting them to using data:

I think you can educate people to data science in Hollywood, and it can grow gradually, but it’s hard, it doesn’t always work, some people are just not interested. And so, the tactic then is: work for production executives who are more data savvy or more data oriented, and start with them, and if possible, try to enhance their skills, that’s something that always works. Because it’s much easier to get your message through if it’s coming from somebody who speaks their language. So, others will be: “ah, ok, so maybe I should…” They see other people do that. It’s a patience game. (Data scientist, studio, April 2021)

44The goal here is nothing less than to secure a definitive place for one’s team at the heart of ‘greenlight’ processes, or in other words to find a role within evaluation communities that have now been expanded to include data specialists. This objective was clearly expressed by the head of a data analysis team at a major studio that has launched its own streaming service:

The ultimate goal is for my team to have a seat at the table with the creators, with the studio partners, and letting them feel like they get additional value from our team to help them take those creative decisions. (Head of a content analysis team, November 2021)

  • 26 Head of the data science department at a streamer, December 2018.
  • 27 Head of the content strategy/planning and analysis team at a streamer, November 2021.
  • 28 Head of the data science department at a streamer, December 2018.
  • 29 Head of the content strategy/planning and analysis team at a streamer, June 2021.

45It is often the professionals who are able to embrace the Hollywood discourse about the creative magic of the industry and the unique nature of what it produces who manage to get into these decision-making circles, but also to establish a career and move up within Hollywood organizations. Their success is based on their adoption of some of the shared beliefs in this space about the value of creation, talent and intuition, and relationships—all things that can never be fully rationalized and quantified. They thus recognize the importance of relational work, and demonstrate their appreciation of Hollywood conventions and norms, notably in their interactions with their colleagues on the content side. While they may have initially found the world of Hollywood creatives as unreal as that of fictional characters, and their behaviour as verging on the caricatural—“they’re, like, always very busy, they’re always on their phone, they’re always out at sets”26—the data specialists who are able to establish themselves in Hollywood are precisely those who come to “embrace [production professionals’] qualitative, creative sensibility,”27 and who define the value of their own work based on the relationships they have developed with these individuals, and their membership of the same art world: “part of what makes my job so fulfilling is interacting with the creative side of it, and seeing our world through their eyes.”28 The head of one content analysis team described to me her pride in feeling “creative adjacent,”29 while a data scientist described his work as contributing to the creative process itself, if not directly, then at least in a no less real and decisive way:

You know, even though we’re not the creative guiding the shows, I think we have a lot to do with what shows are made, what gaps should be filled, and, even during execution, we actually help them quite a bit. So, yeah, we do take a lot of pride in the shows that come out and I take the opportunity to go on set whenever I can, and that’s a great learning opportunity, but it’s also very inspirational (Head of a data science department, August 2021)

46Genuine mechanisms of conversion to the Hollywood frame of reference are at work here, as opposed to situations in which the value of a form of expertise is described by the professionals who hold it as lying entirely in the technique itself (Lee 2015; Polletta, DoCarmo & Wald 2025). The data specialists who successfully stay in the Hollywood game are thus increasingly distancing themselves from the “industry-agnostic” visions of their profession to which they initially adhered. For those who head up a department, this leads them to recruit more and more people with similarly hybrid profiles, in a self-reinforcing process. Since the process examined in this article is still ongoing, it is too early as yet to conclude what determines the success of these conversions, but there does not appear to be a systematic correlation with social origin characteristics or previous career paths that might predispose them to adopt such relationships to art and the creative process. The relational dynamics that have developed within the reconfigured chains of activity in Hollywood organizations thus seem to have played a decisive role in this gradual transformation of modes of professional definition.

  • 30 Content analyst, HBO Max, April 2021.

47In parallel, traditional producers are also increasingly converting to the discourse of data as king. It is often the case that they believe less in data than in the idea that they can no longer do without it, i.e., that they can no longer ignore data specialists. It has become almost unsayable to openly express scepticism about data and the intention to free oneself from them in this space: in the 2010s, I regularly heard the disdainful “I don’t believe in data,” but now, as one of my interviewees put it, “it's blasphemy to say that you don't believe in data.”30 Over and above the question of the sayable, a new horizon of possibilities has emerged, along with new ways of conceiving of the profession of producer or studio executive, which now include paying attention to data and its spokespeople. Within this process, the system of beliefs and roles that for many decades governed the Hollywood game (and prescribed who can legitimately contribute to making production decisions, the value and quality of projects, according to what methods and criteria, and in what interactions) is gradually being called into question. This is also quietly redrawing the previously accepted dividing lines separating the fundamental tasks, on the creative side, from those of support personnel, who are denied the dignity of creatives (Becker 1982).

Conclusion

48This study reveals the effects of the rise of a new category of professionals—equipped with new representations, armed with new tools, and part of new professional configurations—on the transformation of production decisions, and the ways in which artistic value is defined in film and television. The establishment of this new group corresponds to shifts in the structure of Hollywood organizations, the division of labour, and the professional hierarchies within this space. The question of the scope of relevance of these observations, and their broader validity for analysing the logics and the effects of the emergence of a new group within an existing work organization, merits consideration. The apparently paradoxical coexistence of tensions and jurisdictional conflicts with translation practices and conversion dynamics is undoubtedly not unique to the case examined here. Exploring what enables these mutual adaptation processes to take place, and the profiles of the professionals who are driving them, is a research strategy that could therefore bear fruit in other fields. The recent emergence of data specialists across a whole series of professional spheres might prompt exploration of whether there are any similarities in the processes by which these professionals establish themselves and their ways of understanding and measuring reality.

49The emergence of these specialists who now have influence over the production process has challenged existing forms of creative control in Hollywood. Whereas, until recently, the creator-producer team monopolized the legitimacy over deciding what the public wants to and should see, and therefore what should be produced by a studio or production company, this (artist-centred) narrative is now competing with a new production narrative in which data specialists, equipped with new rationalization tools, are best placed to understand what consumers want and to speak on their behalf, and therefore to decide what stories, formats, genres, and stars should be produced or bought. It is therefore clear that the contours of production roles are being redrawn, at the same time as “audiences” and the right way to capture them are being reinvented. These changes also extend far beyond Hollywood, where subscription video-on-demand (SVOD) services were first developed. The business models, technical arrangements, professional roles, and organizational changes that originated in Hollywood are circulating and being appropriated elsewhere (with varying degrees of local resistance and reinvention), notably via the international subsidiaries that the big US streamers have set up in numerous territories. This research thus helps provide insight into the transformation of production roles beyond the US. Comparisons might also be drawn with the effects of the emergence and establishment of data specialists, algorithms, and artificial intelligence in other social fields, so as to gain broader insight into how the uses of these technologies are affecting creative practices, and the definition of creativity and creation themselves.

Haut de page

Bibliographie

Abbott Andrew (1988). The System of Professions: An Essay on the Division of Expert Labor. Chicago, The University of Chicago Press.

Akrich Madeleine, Callon Michel & Latour Bruno (2006). Sociologie de la traduction. Textes fondateurs. Paris, Presses des Mines.

Alexandre Olivier (2023). La Tech. Quand la Silicon Valley refait le monde. Paris, Le Seuil.

Ananny Mike (2016). “Toward an Ethics of Algorithms. Convening, Observation, Probability, and Timeliness.” Science, Technology and Human Values, 41(1): 93–117.

Bastard Irène, Cardon Dominique, Fouetillou Guilhem, Prieur Christophe & Raux Stéphane (2014). “Travail et travailleurs de la donnée.” In Calderan Lisette, Laurent Pascale, Lowinger Hélène & Millet Jacques (eds.). Big Data. Nouvelles partitions de l’information. Louvain-la-Neuve, De Boeck: 133–148.

Becker Howard S. (1982). Art Worlds. Berkeley, University of California Press.

Beer David (2018). “Envisioning the Power of Data Analytics.” Information, Communication & Society, 21(3): 465–479.

Bigo Didier (2020). “The Socio-Genesis of a Guild of ‘Digital Technologies’ Justifying Transnational Interoperable Databases in the Name of Security and Border Purposes: A Reframing of the Field of Security Professionals?.” International Journal of Migration and Border Studies, 6(1–2): 74–92.

Bourdieu Pierre (1980). “The Production of Belief: Contribution to an Economy of Symbolic Goods.” Translated from French by Richard Nice. Media, Culture & Society, 3(2). [“La production de la croyance : contribution à une économie des biens symboliques.” Actes de la recherche en sciences sociales, 13: 3–43, 1977.]

Bourdieu Pierre (1996). The Rules of Art. Genesis and Structure of the Literary Field. Translated from French by Susan Emanuel. Cambridge, Polity Press [Les Règles de l’art. Genèse et structure du champ littéraire. Paris, Le Seuil, 1992].

Callon Michel (1984). “Some Elements of a Sociology of Translation: Domestication of the Scallops and the Fishermen of St Brieuc Bay.” The Sociological Review, 32(1): 196–233.

Casilli Antonio (2025). Waiting for Robots. The Hired Hands of Automation. Translated from French by Saskia Brown. Chicago, The University of Chicago Press [En attendant les robots. Enquête sur le travail du clic. Paris, Le Seuil, 2019].

Christin Angèle (2017).“Algorithms in Practice. Comparing Web Journalism and Criminal Justice.” Big Data and Society, 4(2): 1–14.

Christopherson Susan & Storper Michael (1989). “The Effects of Flexible Specialization on Industrial Politics and the Labor Market: The Motion Picture Industry.” Industrial & Labor Relations Review, 42(3): 331–347.

Delaporte Chloé & Mazel Quentin (2021). “Des publics ‘à la demande’ : penser les usages des plateformes audiovisuelles.” Communiquer, 31: 1–11.

Demazière Didier & Gadéa Charles (2009). “Introduction.” In Id. (eds.). Sociologie des groupes professionnels. Acquis récents et nouveaux défis. Paris, La Découverte: 13–24.

Denis Jérôme (2018). Le Travail invisible des données. Éléments pour une sociologie des infrastructures scripturales. Paris, Presses des Mines.

Desrosières Alain (2002). The Politics of Large Numbers. A History of Statistical Reasoning. Translated from French by Camille Naish. Cambridge, Harvard University Press [La Politique des grands nombres. Histoire de la raison statistique. Paris, La Découverte, 1993].

Dodier Nicolas & Barbot Janine (2016). “La force des dispositifs.” Annales. Histoire, Sciences Sociales, 2: 421–450.

Drumond Gabrielle Silva Mota, Coutant Alexandre & Millerand Florence (2018). “La production de l’usager par les algorithmes de Netflix.” Les Enjeux de l’information et de la communication, 19(2): 29–44.

Dubois Brianne (2021). “Redéfinir une juridiction professionnelle. Artistes et assistantes de production dans les projets d’art contemporain.” Sociologie du travail, 63(3).

Faulkner Robert R. & Anderson Andy B. (1987). “Short-Term Projects and Emergent Careers: Evidence from Hollywood.” American Journal of Sociology, 92(4): 879–909.

Freidson Eliot (1986). Professional Powers. A Study of the Institutionalization of Formal Knowledge. Chicago, The University of Chicago Press.

Frey Mattias (2021). Netflix Recommends. Algorithms, Film Choice, and the History of Taste. Oakland, University of California Press.

Jeanpierre Laurent & Roueff Olivier (eds.). (2014). La Culture et ses intermédiaires. Dans les arts, le numérique et les industries créatives. Paris, Éditions des Archives Contemporaines.

Jenner Marieke (2018). Netflix and the Re-Invention of Television. New York, Springer.

Lee Caroline W (2015). Do-It-Yourself Democracy. The Rise of the Public Engagement Industry. Oxford, Oxford University Press.

Lizé Wenceslas, Naudier Delphine & Sofio Séverine (eds.) (2014). Les Stratèges de la notoriété. Intermédiaires et consécration dans les univers artistiques. Paris, Éditions des Archives contemporaines.

Lobato Ramon (2019). Netflix Nations: The Geography of Digital Distribution. New York, New York University Press.

Lobato Ramon & Lotz Amanda D. (2020). “Imagining Global Video: The Challenge of Netflix.” Journal of Cinema and Media Studies, 59(3): 132–136.

Lyon David (2019). “Surveillance Capitalism, Surveillance Culture and Data Politics.” In Bigo Didier, Isin Engin & Ruppert Evelyn (eds.). Data Politics Worlds, Subjects, Rights. London, Routledge: 64–77.

Marzola Luci (2021). Engineering Hollywood: Technology, Technicians, and the Science of Building the Studio System. Oxford, Oxford University Press.

Menger Pierre-Michel (2002). Portrait de l'artiste en travailleur. Métamorphoses du capitalisme. Paris, Le Seuil.

Polletta Francesca, DoCarmo Tania & Ward Kelly (2025). “Good Storytelling and the Fraught Problem of Intimacy.” In Fleetwood Jennifer, Presser Lois & Sandberg Sveinung (eds.) Narrating Justice and Hope. How Good Stories Counter Crime and Harm. New York, New York University Press.

Rouge Rémi (2023). Le Temps retrouvé, la fabrique des souvenirs numériques saisie par ses professionnels. PhD thesis in political sciences. Paris, Université Paris 8.

Roussel Violaine (2015). “’It’s Not the Network, It’s the Relationship.’ The Relational Work of Hollywood Talent Agents.” In Roussel Violaine & Bielby Denise (eds.). Brokerage and Production in the American and French Entertainment Industries: Invisible Hands in Cultural Markets. Lanham, Lexington Books: 103–122.

Roussel Violaine (2017). Representing Talent. Hollywood Agents and the Making of Movies. Chicago, The University of Chicago Press.

Roussel Violaine (2023). “Faire sens de la transformation numérique d’Hollywood. Naissance et métamorphoses du digital agenting.” Sociologie du travail, 65(2).

Roussel Violaine (forthcoming). Data-Driven Hollywood. The New Data Professionals in the Age of Streaming. Princeton, Princeton University Press.

Strauss Anselm (1978). Negotiations: Varieties, Contexts, Processes and Social Order. San Francisco, Jossey-Bass Publishers.

Zimmermann Bénédicte (2011). Ce que travailler veut dire. Une sociologie des capacités et des parcours professionnels. Paris, Economica.

Haut de page

Notes

1 Netflix numbers for the second quarter of 2023. See Julia Stoll, “Number of Netflix paid subscribers worldwide from 1st quarter 2013 to 2nd quarter 2023,” Statista, 20 July 2023 (accessed on 31 July 2023).

2 The figures for Amazon Prime Video vary depending on the source and, most importantly, how they are counted (subscription, one-off payment, or free-to-view). See Julia Stoll, “VOD subscriber count worldwide 2020-2028, by service,” Statista, 15 May 2023 (accessed on 31 July 2023).

3 Max was created in May 2023 from the merger of the HBO Max and Discovery+ platforms, under the Warner Bros. Discovery banner; the streaming service is expected to remain part of Warner Bros. following its announced split from Discovery Global, set to take effect in 2026.

4 See, among many others, Jeanpierre & Roueff (2014); and Lizé, Naudier & Sofio (eds.) (2014). For a more in-depth discussion, see the introduction to this issue.

5 For a more detailed overview of this work, see the introduction to this issue.

6 This followed the decision by studios to pull most of the content they had initially licensed to streamers for distribution, as the growth of the latter increasingly led studios to view them as competition. The loss of this content drove the streamers to invest quickly and heavily in original content.

7 With the notable exception of Sony.

8 21st Century Fox and NBCUniversal, plus TimeWarner and Disney, before Disney became the sole shareholder in 2019.

9 In this sense, the professional group of data scientists is much broader than that of Hollywood data specialists, and is also structured differently. In this article, by examining the emergence of this new group of specialists involved in production decisions in film and television, I have deliberately chosen not to explore it from the angle of what would, or would not, make them a “real” professional group (notably the existence of a professional body or association, and forms of recognition by the state [Freidson 1986]), but rather to examine the rise of participants who perceive and refer to themselves within Hollywood as belonging to the same category of (data) specialists. This group does, however, include professionals with diverse backgrounds and training, some of whom consider themselves to be “data scientists” in the narrow sense (and who continue to frequent spaces within which professional norms and best practice are confirmed, such as conferences). This is also why I use the term data “specialists” (rather than “scientists”) to refer to the group as a whole.

10 Data were available for 70 of my 75 interviewees.

11 Master of Business Administration.

12 Information taken from the CareerOneStop.org website, supported by the Employment and Training Administration at the U.S. Department of Labor. See United States Department of Labor, Employment and Training Administration, data published on the website: CareerOneStop.org (accessed on 31 July 2023).

13 Bachelor of Arts.

14 Bachelor of Science.

15 In the absence of other available data, these percentages are calculated based on my study population, which gives them a limited representativeness. They do however offer particular insight into the breakdown of my interviewees across the spectrum of data specialist positions (across different roles, within different organizations, and the hierarchies that structure them), and in terms of the “generations” of specialists who have arrived at different points within the short timeframe of the process studied.

16 23.7% were from Europe (UK, Spain and France, although people from France may have been over-represented within my interviewees), while 4.9% were from Latin America.

17 According to 2019 figures from the Silicon Valley Institute for Regional Studies, 64% of tech professionals in Silicon Valley were born abroad. According to the same source, in 2021, people identifying as Asian or Asian-American constituted the largest group among the employees of the biggest tech companies. 37% of leadership roles in Silicon Valley are held by women, and 25% of Silicon Valley residents have a master’s degree or higher (compared with 13% in the US as a whole). See The Silicon Valley Institute for Regional Studies, “Silicon Valley Indicators,” 2019 and 2021 (accessed on 31 July 2023).

18 For example, the most senior positions in the data science teams at Netflix and Disney (until the summer of 2023) were held by women, and women had been appointed to these roles at these companies since they were created.

19 In this sense, their “individual experience” echoes that of people working in the employer groups studied by Bénédicte Zimmermann (2011). Their career trajectories are largely characterized by continuity, despite the variations that mark successive work configurations. I would also join with Zimmermann in emphasizing the importance of approaching experience by taking into account both the individual’s personal career and the collective, organizational framework within which their perceptions and abilities are formed.

20 To develop these abilities, the early data science teams first had to learn and adapt to situations, at a time when these teams consisted of just a handful of people and production within streamers was also in its infancy. Since then, this ability has increasingly been identified as a “skill” that is sought by the heads of these teams when recruiting, resulting in a gradual shift in the profiles of data specialists. I analyse these mechanisms in more detail in a forthcoming book to be published by Princeton University Press.

21 Rather than considering algorithms or data to be the “actors” in this process, my focus here is on the activity of the humans who shape the data and the technical tools surrounding them, and who use them in completely different ways at different times and in different contexts. Attributing agency to algorithms would run the risk of concealing the work of their developers and those who use them, leading to underestimation of the unstable nature and frequent reconfiguration of sociotechnical arrangements (but also of organizational frameworks) in this context and over the period of study.

22 Content analyst, Amazon, May 2019.

23 Content analyst, HBO Max, June 2021.

24 Content analyst, Disney, November 2021.

25 Head of the data science department at a streamer, July 2022.

26 Head of the data science department at a streamer, December 2018.

27 Head of the content strategy/planning and analysis team at a streamer, November 2021.

28 Head of the data science department at a streamer, December 2018.

29 Head of the content strategy/planning and analysis team at a streamer, June 2021.

30 Content analyst, HBO Max, April 2021.

Haut de page

Table des illustrations

Titre Figure 1
Crédits Credit: Pixabay, Violaine Roussel.
URL http://journals.openedition.org/bssg/docannexe/image/8025/img-1.png
Fichier image/png, 2,7M
Haut de page

Pour citer cet article

Référence électronique

Violaine Roussel, « Data Science Versus a Sense of Talent? The Rise of a New Division of Labour in Content Definition in Hollywood »Biens Symboliques / Symbolic Goods [En ligne], 17 | 2025, mis en ligne le 14 novembre 2025, consulté le 16 février 2026. URL : http://journals.openedition.org/bssg/8025 ; DOI : https://doi.org/10.4000/15bkh

Haut de page

Auteur

Violaine Roussel

Professor of Sociology, Université Paris 8, Centre de recherches sociologiques et politiques de Paris – Laboratoire des Théories du politique (Cresppa-LabTop, UMR7217), 59-61 rue Pouchet. ORCID: 0000-0003-2827-1518. vroussel[at]univ-paris8.fr

Articles du même auteur

Haut de page

Droits d’auteur

CC-BY-NC-ND-4.0

Le texte seul est utilisable sous licence CC BY-NC-ND 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont susceptibles d’être soumis à des autorisations d’usage spécifiques.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search