Beyond Global Streamers: Data-Driven Feedback at a Dutch Streaming Service

Introduction

Like other streamers, the Dutch subscription video-on-demand (SVOD) service Videoland collects extensive data on its audience’s viewing habits. However, unlike many of its global and well-researched counterparts, Videoland began sharing selected data with its content producers, aiming to use these metrics to inform script development for non-linear distribution. In 2018–19, at the outset of this initiative, we had the opportunity to conduct a small-scale ethnographic case study, observing how the metrics were discussed within various meetings. We also interviewed data managers, commissioning editors, producers, and screenwriters to explore the meaning and implications of these data and the ways in which they were used for their work (Keilbach and Surma 2022).

Five years later, we investigated whether Videoland had continued this data-sharing practice, which we initially observed during its experimental phase, and whether it delivered the intended outcome: better scripts tailored for the streaming environment. Interviews with two employees reveal that not only has sharing selected data become a standard procedure at Videoland, but the streaming service has also expanded what we have previously identified as forms of audience testing. Videoland thus continues to replicate traditional, established methods of audience research in the television industries, despite claims of being fundamentally different from linear television.

Our research is informed by Havens’ (2014) call for micro-level investigations to examine and analyze data-related practices within the media industries. It is situated within broader scholarship on the datafication and platformization of cultural production (e.g., Baym et al. 2021; Poell et al. 2022). This includes studies on the datafication of audiences and audience research (Athique 2018), the role of data analysis and machine learning in decision-making within film and television industries (Chow 2020; Doyle 2018), and the ways in which globally operating streaming services are impacting screen media production cultures (Afilipoaie et al. 2021; Castro and Cascajosa 2020; Haddad and Dhoest 2021; Idiz 2024; Rasmussen 2025). Of particular importance to our study are publications that examine how streaming data, and specifically unequal access to it—which Daphne Rena Idiz refers to as “performance data asymmetries” (2024, 2138)—affect creative work (Idiz and Poell 2025; Navar-Gill 2020; Rasmussen 2025).

With our small-scale project, we can contribute new insights into specific, data-related in-house practices at a Dutch SVOD-service, beyond existing research on global streamers. Focusing on Videoland and examining a procedure that differs from previously studied cases allows us to add a new perspective to ongoing debates about the role of streaming metrics in content production. In what follows, we first discuss the practice of data secrecy that is common among major streaming services. Next, we provide a brief overview of how Videoland gathers viewing data and how it shares this information with practitioners. Following this, we explore how professionals negotiate and make sense of these metrics. Finally, we outline how Videoland aims to gain even deeper insights into viewer preferences in the future.

Data Secrecy

Unlike the traditional (commercial) television industry, where success is measured through ratings, streaming services such as Netflix and Amazon Prime Video do not share detailed data about their shows (Lotz 2021; Wayne 2022; Wayne and Uribe Sandoval 2023). These streamers closely guard the performance metrics of their shows, restricting access even for their content producers. Such data secrecy creates a significant asymmetry of power, as “ownership of or access to data and to the techniques needed to make sense of data [. . .] are now key sources of competitive advantage,” as Doyle (2018, 60) argues in relation to the television industry.

Despite the data-driven approach to programming and content commissioning, metrics are not openly fed back into the creative production process. Instead, streaming services seem to privilege the idea that ignorance is a form of “creative bliss” (Navar-Gill 2020, 4), suggesting that producers do not need to know how well their shows performed, whether viewers binge-watched or at what point they stopped viewing. Creatives interpret this “culture of data secrecy” (Navar-Gill 2020, 5) as liberating, with the limited availability of data fostering a perception of streaming services as havens of creative freedom. However, all decision-making of these streamers is driven by data analysis, as Navar-Gill clarifies (2020, 9). Like other scholars (Doyle 2018; Idiz and Poell 2025), she emphasizes that the exclusion of practitioners actually increases asymmetry and concentrates power in the hands of the platforms.

Based on interviews with screen workers, scholars such as Navar-Gill (2020), Rasmussen (2025) and Idiz (2024) examine how data secrecy and the asymmetry of power impact streaming production cultures and how practitioners make sense of it. Idiz addresses tensions arising from performance expectations in local markets (2024, 2,138), while Rasmussen highlights how practitioners “interact with streaming data despite widespread secrecy” (2025, 5,245) by sensing, generating and resisting it. Navar-Gill, on the other hand, focuses on how data secrecy has become discursively associated with the notion of creative freedom. She argues that, after a brief period in the early 2010s—when the entry of streaming services into content production and commissioning markets generated anxieties about data undermining creative skills and values (2020, 3)—the concept of “being data-driven” was reframed. Her study shows that the deliberate concealment of viewer information and the practice of “completely siloing data and creative functions apart from each other” (2020, 5) is a key element enabling the cultural imaginary of “creative freedom.”

Unlike the data secrecy typical of global streamers like Netflix and Amazon Prime Video, our case study highlights a contrasting practice: the sharing of selected viewing data with content producers. Based on six interviews and the observation of three meetings (of the data team, the content team, and with creatives of a fiction series) we saw how data are gathered and discussed and how practitioners make sense of them. Access to the meetings and our respondents was facilitated through a personal network contact and granted on the condition that no information would be disclosed concerning alternative uses of data, specific metrics, percentages, or the number of viewers and survey respondents associated with Videoland.

Gathering and Sharing Data at Videoland

Videoland was originally a video rental store chain that became a streaming service before being acquired by RTL Nederland. The streamer with a subscriber base of 1.4 million ¹ features three main types of content. First, it includes material previously broadcast on Dutch RTL television channels, essentially serving as RTL’s content library. ² Second, it licenses international series such as The Wire (2002–2008) and The Handmaid’s Tale (2017–2025). Finally, Videoland commissions original content (in Dutch), producing around 30 Videoland Originals annually. Of these, 6–8 are currently scripted productions, while 22–24 are unscripted, with two separate content teams managing these commissions.

Similar to other streaming services, Videoland systematically gathers quantitative data to analyze user behavior. This includes viewer and viewing information such as subscription numbers, viewing times, and content preferences, as well as details about the devices users employ to access the service and the first program that they watched after subscribing. ³ The collected data further encompasses metrics such as the reach of a particular show or series, the exact timing of viewership, and patterns of use regarding duration and frequency. In the case of serialized content, the data also provides insights into completion rates, interruptions between episodes and their duration, and the moment at which viewers discontinue watching. While in 2018–19, data gathering was conducted “manually” by data analysts and evaluated only for specific projects, by 2024, automated processes had been implemented, making streamlined data reports a standard feature. ⁴

When we conducted our research in 2018–19, Videoland had just begun sharing viewing and performance data with its content producers (Keilbach and Surma 2022). In hindsight, they were in the early stages of experimenting with how best to integrate and communicate this information. At the time, metrics were gathered three days, one week, one month, and two months after a show’s publication. Videoland’s data team, composed of nine female data analysts led by a male team leader, created a visualization in PowerPoint with the key data. This visualization was used by the content team to discuss the performance of the respective show in greater detail.

For the series we selected as our case study, the content team focused on metrics such as the number of viewers, the completion rate and the binge rate. As the data visualization highlighted a significant “dip” in the middle of one episode, this moment when viewership numbers suddenly dropped received particular attention during the meeting. After some speculation about the cliffhanger in the previous episode, a team member used the time code to locate the specific scene in question. The subsequent discussion of this scene informed the development of a topic list for a survey, which the content team used to gain deeper insights into audience viewing behaviors and preferences. Alongside general questions about the streaming service, the survey included an evaluation of the series’ content. Subscribers who had watched the show were asked about various aspects, such as the setting, characters, and storylines—areas that quantitative metrics cannot address.

In preparation for the feedback meeting with practitioners, the commissioning editors selected both quantitative metrics and qualitative survey responses. In our case study, visualizations of these data were presented during a meeting with the creative producer, director, and two screenwriters. The presentation combined viewing metrics expressed as percentages and selected quotes from the survey, aiming to inform the creators about audience responses. Notably, this feedback took place only after the decision to produce another season had already been made, indicating that viewing data, at the time of our research, was primarily used as a tool for reflection. This sharply contrasts with other streamers’ practices of relying on performance metrics primarily to cancel shows—or greenlight a second season—after a set evaluation period (Renfro 2020). The timing might have contributed to the relaxed atmosphere during the meeting, further enhanced by the commissioning editors’ repeated expressions of satisfaction with the program’s performance. The noticeable “dip” in viewership that was clearly visible in the data presentation, sparked a lively discussion about potential reasons why viewers might have stopped watching. The selected responses to the survey questions were also discussed in detail, as some of them came as a surprise to the producers.

The time planning indicates that data was not primarily analyzed to decide whether to commission a second season. This aligns with the content team’s understanding of data-driven feedback as serving an educational purpose, aimed at assisting creatives in improving the quality of their work. They consider this particularly relevant given that many of the scripts Videoland receives are categorized as “linear material,” which the content team regards unsuitable for streaming.

Five years later, after the automation of data gathering and the standardization of in-house data reports, the discussion of viewing data within various teams and with creative practitioners remains a core practice at Videoland. However, only selected metrics, which have been authorized by the management, are included in these reports, while other metrics are deliberately withheld—such as data that could reveal insights into the actual number of viewers. Furthermore, in some instances of Original commissions, data-driven briefings have become integral to the early stage of content development. In these cases, commissioning editors utilize streaming data that Videoland has systematically gathered over time to inform the creative decisions of a new project, once producers and screenwriters are involved.

While qualitative surveys continue to provide valuable insights into viewer preferences, Videoland has recently also introduced test screenings for some Original content projects, follow-up seasons of existing content, and trailers for new formats to gain an even deeper understanding of their audience. During these screenings, carefully selected subscribers watch, for example, (parts of) specific episodes or trailers—either at home or at screenings in the in-house cinema theater or in simulated “living-room” settings, followed by interviews. By adopting this approach, Videoland is drawing on traditional audience research methods from the television (and film) industry (see Ang 1991 or Redvall 2017 on audience research in the television industry; see also Zafirau 2009 or Napoli 2003 on audience testing and measurement in the screen media industries).

Making Sense of Data

While practitioners in the US sometimes value the siloing of viewing data, perceiving this concealment as a safeguard for creative freedom (Navar-Gill 2020), Videoland’s contrasting approach prompted us to explore how Dutch content producers make sense of the detailed information shared with them. If data secrecy is seen as liberating, could data-driven feedback, conversely, be perceived as a constraint on creativity?

In the feedback meeting we observed in 2019, all attending practitioners (the creative producer, the director, and two screenwriters) were open and engaged. Given the novelty of the process, they initially asked numerous questions about the terminology and the significance of specific percentages, pointing out that they do not know how to read the data. In response, the commissioning editors took the time to explain key metrics, such as the “completion rate” and “binge rate,” and reassured the creative team that the numbers were “really good.”

The remainder of the meeting was characterized by lively discussions about the series’ content. The presentation of quantitative data on viewer behavior, for example, sparked debates about the narrative structure, character development, and storylines. The previously mentioned “dip” of viewership in the middle of an episode led the writers and producer to critically examine potential narrative and content-related weaknesses. Additionally, the qualitative results, presented as systematically organized quotes from the questionnaires, occasionally provided the writers with surprising insights—for instance, the unexpectedly positive reaction to a side story that they had regarded merely as a spontaneous idea or byproduct. The analysis of both quantitative and qualitative data led to animated discussions among all participants in the meeting and fostered a reflective exchange.

This stands in stark contrast to experiences with ratings in linear television. During our interview with the producer, she described the anticipation of ratings and the communication that followed the day after a show aired as nerve-wracking. By contrast, she found Videoland’s approach to sharing metrics to be a form of productive feedback, which she characterized as pleasant, interesting, and inspiring. This perspective was echoed by other members of the creative team, confirming that in our case study, data was not perceived as a limitation to creative freedom.

The screenwriters explicitly aligned the metrics with their “gut instincts,” making sense of them as a confirmation of their creative intuitions. Reflecting on the “dip” in viewership highlighted in the data, they mentioned that they could have anticipated issues with the storyline and acknowledged that they should have trusted their gut instinct. This means the data validates doubts they already had during the writing process but were unable to act on due to the constraints of a tight production schedule. Interestingly, their surprise at the unexpected popularity of a minor storyline was not seen as a contradiction to their gut feeling.

This recalls Zafirau’s (2009) observation that Hollywood producers and other professionals continually refine and validate their “gut instincts” as part of their daily routine to better understand audience preferences. This process involves, for example, putting themselves in the shoes of viewers, such as by regularly attending movie screenings. In the context of our case study, data-based feedback, as provided by Videoland, can be understood as serving as an additional tool, helping screenwriters and producers validate and refine their instincts that guide their creative decisions. Whether aligning with or contradicting their “gut feeling”—data-driven insights are a resource to shape creatives’ instincts.

Videoland’s commissioning editors, by contrast, embraced the use of metrics for their educational potential. In our interviews, they highlighted the value of data for improving scripts to better suit the demands of the streaming context. This perspective directly addresses the persistent issue of being offered an abundance of “linear material” which poses a challenge for Videoland as the streaming service seeks to distinguish itself from “television” and its established form of storytelling.

Despite the commissioning editors’ hope of educating practitioners through data-driven feedback, the screenwriters and producer maintained that data would not influence their creative decisions. In our 2019 interviews, both the creative producer and the screenwriters explicitly stated that writing and producing for a streaming service was not different from working for traditional television—regardless of metrics. In that respect, our findings align with the statements of screen media workers interviewed by Daphne Rena Idiz who, overall, “felt that working with Netflix was not significantly different from working with European broadcasters or other streamers” (2024, 2,136).

By 2024, as data-driven feedback has become an established practice at Videoland, responses to metrics appear to have diversified. In an interview, an employee identifies two dominant reactions among creative practitioners: some screenwriters and producers embrace such feedback as a valuable resource, echoing our observations made in 2019, while others reject its influence in order to safeguard their creative instincts. The commissioning editors, meanwhile, place even greater emphasis on the educational function of data-driven practices. They point to the slow pace of change in the Dutch media industry as a key reason why data remain relevant.

However, belief in the potential of performance data to inform creative decisions and improve content—understood as making it more suitable for streaming—does not imply an uncritical enthusiasm for data. Instead, this belief is accompanied by increased skepticism toward quantitative metrics. Videoland staff remain keenly aware of the importance of having data-savvy personnel and prioritize human interpretation of data over relying solely on automated conclusions. Additionally, qualitative data, such as insights derived from surveys and more recently from test screenings, has become increasingly central to their decision-making process.

Audience Testing

Discussing their data-driven feedback practices, the mid-level executives we interviewed emphasized Videoland’s approach as a distinguishing feature from linear television. However, while the streaming industry tends to highlight transformation, we, as television studies scholars, see these practices more as a continuation of traditional audience research long used in broadcast television. Gathering data through surveys and test screenings are well-established methods for studying audiences (Ang 1991). Furthermore, the ways in which practitioners are making sense of metrics and the utilization of data to inform creative production resonates with specific forms of audience testing commonly used in the film and television industries.

In her historical analysis of audience research at the Danish Broadcasting Corporation (DR), Eva Novrup Redvall presents a form of audience testing that differs from the way it is used in the US television industry. Despite ongoing discussions about their value, test screenings in the US “still hold great power” (Redvall 2017, 3), often informing decisions to greenlight or cancel shows based on audience feedback. In contrast, the alternative model at DR, as described by Redvall, uses audience testing as a “dialogue-based tool” (Redvall 2017, 2) that fosters creative learning. By focusing on viewer perceptions of storylines, characters, and aesthetics, DR repurposed audience testing as a “development tool” (Redvall 2017, 5) with a diagnostic function. Rather than being seen as a threat to creative freedom, viewer evaluations became a source of inspiration, prompting constructive conversations, self-reflection among creators, and ultimately enhancing the quality of the series.

Positioning viewer data, as gathered by streaming services such as Videoland, within the broader tradition of audience testing, emphasizes its function as a productive tool for creative development. Our observations show that viewer evaluations, whether gathered through structured feedback mechanisms or aggregated viewing metrics, can drive the creative process, sparking discussions about content and encouraging constructive self-evaluations among practitioners. This perspective challenges the notion of data-driven practices in streaming as fundamentally innovative or transformative, situating them instead within an ongoing evolution of audience research practices.

However, this optimistic framing must also grapple with the realities of unequal access to data and the problematic power dynamics inherent in its use. Commissioning decisions based on viewing data still occur behind closed doors, with limited transparency about how these data are interpreted or applied. Videoland selectively discloses data to creative practitioners, with the intention of “improving” content and thus mainly after a series has been renewed for another season. Shows that are not renewed typically do not receive such insights, leaving many creators in the dark about how their work was perceived. Moreover, it can be safely assumed that metrics that could weaken the streamer’s negotiating position in the future are withheld, and that only the numbers that align with their strategic interests are shared.

This selective and asymmetric sharing of viewing data underscores the complex power dynamics inherent in these practices (Doyle 2018; Idiz and Poell 2025), even within an environment where the sharing of viewing data has been structurally implemented. While data-driven feedback has the potential to serve as a development tool, its use remains dependent on the streaming service’s discretion, restricting its broader applicability and reinforcing streamers’ dominance in shaping creative processes. Although viewing data can and do function as a productive instrument for creative development, their use in the Dutch streaming industry still reflects entrenched hierarchies and access disparities.

Conclusion

By sharing selected metrics and qualitative audience feedback with practitioners, Videoland challenges the prevailing narrative of data secrecy often associated with SVOD services. Their approach underscores the distinctiveness of local streaming services compared to global streamers and serves as a reminder that multiple paradigms exist within the streaming industry. Accordingly, Videoland employees position the service as distinct not only from “television” and its established forms of storytelling but also from global players like Netflix or Amazon Prime Video, emphasizing a unique production culture based on long-standing relationships with local screen media workers. With our study, we emphasize the importance of examining these local SVODs and their practices, as they complicate the understanding of streaming services and their use of data. Rather than siloing data from the creative process, some streamers, like Videoland, utilize metrics as a development tool that fosters creative learning.

Based on our observations, we have discussed two (of the many) ways data are made sense of: commissioning editors view data as a tool to help screenwriters and producers improve scripts to align with the demands of a streaming service, while creatives use the same data to validate or refine their instincts, thereby continuing long-standing practices in the media industries. This suggests that, for practitioners, creating content for streaming services is not radically different from working within the traditional audiovisual industry.

Despite the streaming industry’s rhetoric of transformation, our observations indicate that production practices largely mirror established methods and processes, borrowing extensively from traditional approaches used in the television industry. This continuity raises critical questions about the actual impact of data-driven practices. While data-driven feedback may offer opportunities for iterative development and optimization, whether it genuinely leads to the creation of “better” scripts remains uncertain. What is clear, however, is that rather than revolutionizing creative production, the use of data in the Dutch streaming industry reflects an ongoing evolution—one firmly rooted in the practices of its broadcast predecessors.