Public Service Broadcasting and Data-Driven Personalization

A Shifting Structural Environment

Much has been written about what Nissen (2006) identifies as profound forces for PSB in the new media landscape: digital technology, changes in the market, regulatory shifts, and changes in both audience composition and its mechanics. The organizational landscape has been defined by domestic and transnational commercial media enterprises and by international institutions (e.g., the World Trade Organization). Throughout the European Union (EU), public value tests and market impact assessments have constituted regulatory interventions intended to maximize market efficacy (cf. Donders and Moe 2011; Suárez Candel 2012). At the same time, through the European Broadcasting Union (EBU), broadcasters themselves have influenced and pursued various forms of advocacy though PSM funding, governance, and trade policy.

Still, the language of adaptation to outside forces has seeped into PSB management discourse, as demonstrated in a recent reader by Głowacki and Jackson (2013). To this, Nissen (2013b, 74) has revisited a discourse of internality and externality: “The internal complex factory mode of production is increasingly inappropriate” for meeting the external demands from the media market and the “change in the basic, essential media paradigm (from collective/mass to individualized/personal).” Elsewhere, he has argued for PSBs to more thoroughly adapt editorial strategies to a nonlinear narrative paradigm (Nissen 2013a, 96).

PSBs have to balance concerns for their public service remit while trying to attract audiences (Tambini 2014, 24). A key question is how to define public service value and, increasingly, how to deliver it effectively in a landscape dominated by social media and its pervasive logic (van Dijck and Poell 2014). Helberger (2011, 2015) argued that algorithmic profiling and targeting are not the only means to show users more personally relevant advertising, but profiling and targeting also could be used proactively and ethically by a PSB to gently nudge the audience toward making more diverse media choices based on their personal preferences and habits.

Nordic PSBs have been characterized as “resourceful, resilient, and adaptable to changing circumstances” and based on “an adaptive approach to enlightenment” (Syvertsen et al. 2014, 129, 80). Despite their programming having been significantly modified in response to societal change, their perceived core obligations remain (comprehensive programming, public enlightenment, and an ethos of being important to society rather than to individual consumers). These characteristics—which have “a central position in society, substantial public funding, a comprehensive and varied output, and authority as the voice of the nation” (Syvertsen et al. 2014, 73)—have depended nevertheless on political and public support for welfare state solutions. Legitimization strategies thus become paramount. As Larsen (2010) pointed out, PSBs in Norway and Sweden seem to experience a strong Habermasian duty, where the remit perceived is to serve “the public” singular. This implies an egalitarian ideal, steeped in a social democratic tradition of civic empowerment and, equally, social cohesion and a strong emphasis of “the national interest” as a normative category.

One issue that has recently begun to be explored by scholars is the tendency of PSB companies to collaborate with commercial Internet platforms such as Facebook, a development that might lead to possible loss of institutional control (Moe 2013). Scheduling remains important in the multiplatform era as well (Ihlebæk et al. 2014).

Bennett (2013), Jackson (2013), and Larsen (2010) had empirically studied the ways in which digital progressivism is organizationally negotiated within PSBs. Bennett (2013, 7) argued that a key problem for PSBs in a multiplatform era is that their audiences are “at once too big and too small.” Although the size of national audiences necessitates automated modes of address and adaptation to lowest common denominators to maximize ratings (I label these “catch-all” strategies), these audiences are too small for making individual sub-groupings economically viable to uniquely identify and address. Furthermore, he argues, the question of metrics remains a significant challenge. How to measure and value online audiences is far from clear in the multiplatform era.

In the United States, the Future of Public Media research project surveyed recent innovation practices within PSM. The project identified new strategies in the following areas: inter-organizational collaboration, media literacy and access, engagement of youth, involvement of independent production agencies, journalism and reporting structures, multiplatform distribution, and community engagement via social media platforms (Clark and Aslama Horowitz 2013, 12). Jackson (2013) scrutinized the ways in which the BBC has negotiated user-generated content and audience interaction. From what she describes, it seems as if these activities have been treated largely as “social add-ons” in terms of quantitative audience impact—despite the corporation’s eager adoption of the concept of user interaction since 1998 at least. Tropes such as “second screens” and “social TV” are profuse in discourses on television and new media, both within academia and in the wider world. I will not dwell on these particular concepts, but suffice to say that my contribution here problematizes the tendency to see “socialness” as a functional auxiliary to existing broadcasting operations. All too often, socialness seems to be thought as a meta-function that serves to advertise the core function of broadcasting and create socially distributed “buzz” (particularly in a live event). Contrastingly, personalized audiovisual delivery constitutes a strategic shift at the very core of the broadcast medium, because it enacts more of an exhaustive adaptation to the user. Commercial broadcasters are anticipating this to better predict preferences to serve audiences more of what they are thought to desire and to allow for social sharing of playlists and hyperlinks. Yet, for PSB actors, merely serving audiences more of what they are thought to desire might not be as self-evident a goal.

I have previously (Andersson Schwarz and Palmås 2013) explored three examples of external innovation appropriated by Swedish PSBs, raising questions of structural hybridization and the feasibility of algorithmic approaches and integration with, for example, Facebook’s “social graph.” The Swedish Media Authority decreed in 2010 that overly enthusiastic PSB integration with Facebook and Twitter constituted an endorsement of these commercial services (Moe 2013, 118), something which led to a more “guarded social media use” (van Dijck and Poell 2014, 6) within these particular organizations.

Organizational Adaptation

To adapt is never solely a matter of exogenous (external) factors determining change; there are also endogenous (internal) influencing factors. This is doubly true when analyzing organizational adaptation (cf. Hernes 2008; Thrift 2005). Media institutions enjoy a particularly permeable relationship with their environment, primarily because they enact mediations with the surrounding world and, despite increasing commercialization and consolidation, continue to perform collective functions in society (Chadwick 2013; Hjarvard 2008). In the case of journalistic media, this duty of absorptivity is arguably even more pronounced, because their remit is to represent the surrounding world. Hjarvard (2008, 119) argues that media institutions have, since the early 1980s, become less “sender-steered, for example, steered by particular interests in the days of the party press or by the terms of public service broadcasting (PSB) concessions,” and more “receiver-steered,” catering more explicitly to their own users/audiences—partially as a concession to market demand and purchasing power and partially as a way to be less deferential to other social institutions. Ironically, although this has allowed media institutions to become more politically autonomous toward other established institutions, they arguably become less autonomous with regard to their “receivers.”

Organizational executives can be seen as “boundary workers” (Hernes 2008, 16–17) constantly negotiating inner and outer forces. Mansell (2014) has analyzed multinational discourses on developmental information and communication technology policy. She found that social actors who presuppose a media “environment” are co-constructing an evolutionary perspective by emphasizing change as exogenous (having an external cause or origin). Such a perspective privileges economic and technological factors. According to Mansell, endogenous perspectives are more ethnographic and/or practice based, and privilege factors associated with the learning and the local context of development processes. Ibrus and Merivee (2014) apply this evolutionary perspective in a recent analysis of Estonian PSB, noting that in strategy development, change is seen to be endemic, both externally and internally conditioned. An organization’s future is thought to be dependent on its ability “to transform itself from inside out on a daily basis—that is, to adapt incrementally” (Ibrus and Merivee 2014, 102).

I would argue that for broadcasting corporations, an exogenous conception of social media integration—employing various forms of integration with social media platforms (or their functional auxiliaries, for example, the Facebook “social graph”) and/or back-end monitoring applications (e.g., Outbrain)—would imply forms of audience engagement that become para- or meta-functions to the overall editorial operation. By instead seeing PSBs as constituted in and through their audiences, a more endogenous interpretation of the public service remit is possible. Here, audiences are more directly involved (e.g., by offering personalized content suggestions). This would also constitute a more catholic approach that goes beyond the limitation of equating “social media” only with social media platforms (although, admittedly, current circulation of links and playlists almost entirely takes place through such platforms).

An even more expansive conception can be envisaged in which the public service remit is entirely re-conceptualized to include novel audiovisual formats, and/or interoperability with cultural institutions and public utilities that have historically been excluded from the notion of PSM, such as archives, libraries, museums, and publishing (cf. Andrejevic 2013; Moe 2014). In Sweden, this conception of PSM has been espoused by Rosengren (2014).

The Specificity of the Swedish Media Landscape

Nordic PSB actors were very early, proactive developers of Internet-mediated VOD/OTT delivery. They had the structural abilities with relatively well-equipped research and development (R&D) departments and no profit requirements. Contrastingly, the commercial incentives for many of their competing broadcasters employing VOD/OTT were lacking for a long time, due to the small market. The linear mass audience model was profitable enough, and the massive expansion of linear channels taking place in the 1990s was thought to be the answer to niche demand.

SVT pioneered audiovisual OTT delivery in Sweden by launching SVT Play in 2006. This service soon began altering fundamentally the way audiences received audiovisual broadcasting, prompting commercial actors to adopt similar services (Grafström 2010). Almost a decade later, Netflix is now estimated to parallel PSB streaming in its daily reach (Mediavision 2014). Among 16- to 25-year-olds, linear viewing has declined from 140 minutes per day in 2008 to 72 minutes in 2014 (MMS 2015). There is no doubt that while the consumption of linear television is clearly diminishing among certain groups (cf. Bardoel and d’Haenens 2008, 341), streaming is rapidly growing in some (partially overlapping) demographic groups. In early 2014, the Swedish television broadcasting landscape was shaken by a general downturn in linear viewing ratings across the board (Nordström 2014; Svedjetun 2014). Public service VOD/OTT viewing rates declined as well. Still, when looking back at the last decade, PSB audiences in Sweden as well as in Canada and Australia have not declined overall (Tambini 2014, 79). Furthermore, the ad-free model that is typical for the Nordic countries is interesting in itself; Tambini (2014, 83) observes that internationally, license fee and other public funds have risen compared with ad-based funding.

In the Nordic countries more generally, PSBs have gained legitimacy through their public mandates, which citizens, competitors, and regulators seem to associate primarily with high audience penetration (Nord and Grusell 2012). There has been strong opposition to what has been labeled a “monastery model” (Jakubowicz 2003), through which PSBs are marginalized, such as in the United States. Echoing a theme in my interviews, PSBs aspire to represent the whole country; they see themselves as key actors for the formation of opinion and identity. They have heeded fragmentation, be it via news avoidance (Strömbäck et al. 2013), via elite formations, or via the resurgence of partisan media. All these are seen, by my informants, as risks that act to deepen societal divides regarding participation, deliberation, and cohesion (Boczkowski and Mitchelstein 2013).

The mimicking of successful commercial tactics and formats (e.g., shows, quizzes, reality TV) in PSB during the 1990s has been portrayed as a break with the system’s paternalist heritage (Syvertsen 2003, 170; van Dijck and Poell 2014, 3). Although this mimicking to counter marginalization and work against the monastery model is a survival strategy for PSB, it routinely gets read by opponents of PSB as not only a naked ambition to maximize audiences, but also one that is often heavy handed in its execution.

For SVT, the overall amount of content produced has increased in recent years, due to a succession of new channels established in 1999, 2002, and 2004. These increased the potential for personalization. The increase in content breadth has been equally as important as the VOD/OTT playback function, because it determined the palette of programming to work from in the first place. Furthermore, SVT proactively supported terrestrial digital distribution early on for a range of reasons. SVT, in part, could not only retain control over distribution but also legitimately claim to reach the entire population. For the same reasons, SVT deliberately postponed investment in OTT until after the terrestrial switchover; the company would not have been able to legitimize a carrier technology that would have reached only a minority.

Nordic PSB Personalization and SVT Flow

In February 2014, the Finnish PSB YLE launched their news app Uutisvahti, which presents a personalized newsfeed based on the user’s own choice of emphasis and reading history, much like the American public radio broadcaster NPR. Like the commercial Swedish news aggregator app Omni, which the Schibsted media conglomerate launched in November 2013, the user does not actively have to seek news. The app notifies the user with a push notification when something deemed critically newsworthy has happened. In June 2014, the Norwegian PSB NRK revealed that its developers were experimenting with personalization (Breivik 2014). Their announcement included an online questionnaire, as user dialogue was anticipated. NRK noted that although none of the content would be locked away, various types of logins could enable a higher degree of relevancy to users. Tellingly, one of the developers later blogged about the risks of personalization, emphasizing linear TV’s ability for “accidental learning” (Hofseth 2014). Much of this reasoning was looked on favorably also by Swedish PSB executives in my fieldwork. The NRK login would be implemented primarily on web and on the iOS and Android mobile apps (Aaeng 2015). As of May, 2015, the Danish PSB DR had begun implementing personal login for their streaming service, DR TV.

In April 2014, SVT launched SVT Flow, a new functional and aesthetic layer on top of the underlying SVT Play engine (figure 1). This begged the question whether it constituted an entirely new service, or merely a cosmetic “add-on.” We will never know the whole answer to this, because SVT Flow was shelved within less than a year of operation. ² It presented a hybrid of on-demand viewing and what SVT calls “flow viewing” (scheduled broadcasting), adding push functionality to the typical pull delivery inherent to VOD. It offered playlists that were manually curated by editors and not personalized through any form of algorithmic user adaptation; this allowed the users to freely follow these or switch back to SVT Play. According to the advertising agency Forsman & Bodenfors, which co-produced SVT Flow together with SVT, the service was intended as a catch-up offering for precisely those abovementioned groups who shun linear TV and who are believed to value convenience and more casual modes of viewing. Its interface resembled that of Netflix, while its programming resembled conventional TV scheduling.

OPEN IN VIEWER

Figure 1.

Screen interface, SVT Flow, June 16, 2014.

The main reason for shelving SVT Flow was that its functionality was not sufficiently clear to the users. According to Glaser, viewers had considerable trouble finding it. Like many of my respondents, she too emphasizes the current difficulties of measuring online viewing with the same accuracy as linear viewing (Silvergren Blåder 2015). The anticipation that online streaming enables better measurement is contradicted by the simple recognition that unless unique user identifiers are implemented, what is measured are simply clicks and IP addresses, enabling little or no knowledge of who is actually in front of the screen.

PSB Perspectives on Audience Data

PSB managers want to know their audiences and what they think of the services, but for entirely different reasons than commercial broadcasters. The manager of audience analysis and corporate intelligence at SR (Interviewee 02) contrasted the PSB’s need for audience intelligence with that of the commercial broadcasters that she had previously worked for. She maintained that SR genuinely wants to be able to provide its audiences better value. She used a metaphor of putting ethnic minorities “on the front cover” to differentiate SR from commercial broadcasters. Although these would present an outward image of ostensible diversity, the PSB approached diversity intrinsically, from the inside out by having certain types of ethnic employees to access, understand, and describe different worlds. This is not to say that the PSB would primarily adhere to an internal mode of adaptability, while the commercial broadcasters would primarily appreciate external appearance: the inside-out hermeneutics she seemed to describe was complex and highly reflexive; it sought to find in the outside world what the internal organization needed with the employees in her example clearly serving as “boundary workers” (Hernes 2008, 16–17). The hermeneutics of seeing broadcasting “as primarily a social rather than an economic process, as something with moral, cultural, intellectual, and creative purpose” (Tracey 1998, 19) opposes digital reductionism.

In another interview, I asked a manager of content and output analysis at SVT (Interviewee 09) about the risk of adapting a “click-maximization mindset” in which content is designed to engender maximum shareability or likability. He answered that such a tendency, if it were to exist, would signify that “something is broken”:

That kind of fear or desperation I don’t see in the company. In contrast, we know that our mission is focused on the entire population but that we don’t have enough strength in our offering to reach the entire population. That is partly because the groups are of different sizes and have very different preferences. We struggle to be equally strong or valuable for everyone. We have, if I may use the term, a competitive situation where some groups are subject to more intense commercial targeting than other groups.

As I understood him, the respondent posited that the inadequate relationship SVT has with certain age groups hampers its ability to do outreach; engagement is hard to generate if there is not even awareness to begin with. The broadcaster has to use a tactic in which more attractive pieces of content would lead audiences to, or at least make them aware of, democratically important content. One example would be to syndicate popular U.S. drama shows and schedule these to stimulate mass appeal. He explained the need to create relationships that extend beyond piecemeal programming and to make PSB more meaningful as a place to discover unexpected things. He maintained how, in terms of sheer numbers, linear viewing was dominant. Audiences rarely used TV for information, he noted, quoting Zillmann’s (1988) emphasis on emotional patterning and TV as “mood management.”

Majoritarianism and Paternalism

Several respondents confirmed that during the first decade of the millennium, the ambition to target one audience at a time shifted toward a more pronounced remit to include everyone. The current mode for driving public appeal in Swedish PSB paradoxically resolves the problem of catering to minorities through firmly majoritarian modes of address. This means that each product, while maintaining to represent and/or include minorities, should be popular with the masses. “What happened in the old strategy [of niche address] was that cost became immense in relation to the numbers of viewers actually reached” (Interviewee 09). Even worse, the low awareness among certain members of the population, such as pensioners, led to the failure of PSB targeting. Messages intended for youths reached pensioners instead.

PSBs still have a long way to go in giving up on their broad, all-inclusive modes of address. In radio, the conventional linear channels that reach the bulk of the audience, reach far fewer people in the online, streamed version. Individual shows might have enjoyed a reach of only 0.3 to 0.4 percent of the total audience. VOD has had a similar tendency. Operating within a small language area has taught important lessons, not least of all that it has dis-incentivized piecemeal approaches. Because most audience numbers for individual programs/shows represent such small fractions of the audience, they become statistically inadequate.

Prior to its gradual deregulation in the 1980s, scholars most often associated PSBs with paternalism. On the surface, algorithmic personalization might seem like the polar opposite of paternalism, by giving audiences the power to collectively decide what to consume. What such a reading of digital populism glosses over, nevertheless, is the fact that social media platforms are riddled with bias (van Dijck and Poell 2013). The highly commercial infrastructures for social sharing are rigged to favor certain types of interaction and certain forms of content. Furthermore, the market for personalized Internet-based services is dominated by a handful of mainly U.S.-based, highly corporate actors.

Key Findings

In this article, I have explored some of the various ways in which Swedish PSB executives and strategists apprehend exogenous trends of personalization, social shareability, and predictive audience management. Although algorithmically based audience prediction is idealized in the current media economy, ad-free PSB pinpoints some limitations of such suggestive visions of “datafication” and tailored delivery.

There are two modes of adaptation: the cybernetic adaptation to users’ desires and the organizational adaptation to normative techno-cultural trends in society. By studying the ways in which PSB executives experienced the organizational adaptations, we see how both these modes are not only viewpoint dependent but also contingent on particular factors, primarily the original public service remit for the broadcasting operation.

For algorithmic/correlational approaches to become relevant at all, two requirements have to be met. First, the data accessed and retained have to be detailed enough to cover relevant behavior without being intrusive. Second, the data mining and correlational analysis have to operate at an economy of scale, partly because the data analysis requires a significant degree of manual labor, partly to render statistical populations that are adequately large. Even within large populations, if the subsets probed are too small, the noise and/or probabilistic certainty would distort the findings. Hence, despite the benefits of the algorithmic approach, with their convoluted processes of identifying relevant patterns and then tailoring communication to those fractions of the population, the well-worn, catch-all broadcasting heuristic still has a function in society. Because algorithmic prediction always entails probability calculations, minority programming is riskier; statistical predictions are more likely to be wrong the smaller the target populations are. High granularity in broadcasting entails an information problem. In a cost-benefit calculation, a catch-all strategy of dissemination thus might be preferable instead of niche address, especially in the case of small language programming. Indiscriminate broadcasting has non-reciprocal and inclusive properties (Scannell 2004), making it a deliberate counterpoint to personalized, contextualized, granular media experiences.

SVT is not likely to embrace full-scale personalization. Rather, innovations that divulge such a move should primarily be thought of as a form of “image management.” SVT Flow in many ways constituted one such venture. The service outwardly appeared to indicate more inductive programming, while, in fact, it was premised on a rather conventional editorial selection. As brands go, many PSBs want to appear attuned to a contemporary modernity whose signifiers are convenience, personalization, and particular user-interface aesthetics, such as speed and lack of friction. One might think that the appearance of organizational progressiveness and modernity satisfies the search for societal legitimacy and public appeal. If all other media corporations are engaging in personalization techniques, PSBs should be expected to follow by this logic. According to several of my respondents, Nordic PSBs have struggled to be attractive to discerning, technology-oriented media users, particularly in the younger age groups (15–40 years of age). This helps explain the conspicuous dimension behind the institutional implementation of technology (e.g., up-to-date apps, players, etc.) as a way of adding brand value.

Huge corporations such as SR and SVT are characterized by certain degrees of internal heterogeneity. Different departments and divisions have somewhat different aims and goals, pulling in somewhat different directions, and resulting in a degree of institutional divergence. At SVT, internal organizational heterogeneity seemed to be less of a problem than within SR. Quite likely, the SVT strategy to maintain a catch-all, rather than a niche, programming approach contributed to institutional consensus. The organization’s web-exclusive content and new services, such as SVT Flow, appeared to have been anchored in this consensus. Within SR, however, I noted some dissensus, particularly when some of new online services lacked wider organizational backing. According to some of my respondents, these services never were marketed widely enough, or allowed enough shelf life, for widespread user adoption.