YouTube’s Operational Logic: “The View” as Pervasive Category

Abstract

Launched in 2005 as a platform for user-generated content (UGC), YouTube is one of the most popular websites in the world. In this article, I focus on the site’s use of “the view,” which I argue serves as a pervasive category enacted through the platform, in its information regimes and beyond. The view supports a myth of viewer intentionality and satisfaction and serves as the operational logic of the platform as a whole. It is a category in Durkheim’s sense, ordering practices and naturalizing hierarchies and inequalities. These hierarchies concern, and impact on, participation, financial compensation, visibility, and popularity. In making my claims, I demonstrate how the celebratory discourse around YouTube as an empowering tool that levels the media playing field was positively misguided. I make a plea for a critical reading of the view, which can enhance our understanding of the platform and its culture.

Launched in 2005 as a platform for user-generated content (UGC), YouTube is one of the most popular websites in the world. Since 2006, the platform operates as an Alphabet (Google) subsidiary. The majority of the videos uploaded to the site are free to view and supported by advertising. The standard for charging for advertising on YouTube has become cost-per-view (CPV) bidding—the number of views an ad receives establishes this cost. But YouTube views, although important to the platform’s business model, represent just one dimension of “the view,” a category that has become deeply embedded in the thoughts and actions of YouTube stakeholders. What I aim to reveal in this article is that far from being merely a transactional metric, the view functions as the operational logic of the platform as a whole. It plays a critical role in creating a kind of platform culture that is now being scrutinized both in the popular press and by academics. Many have expressed their hopes for the platform, focusing on its participatory potential—but it is becoming increasingly evident that inequalities and hierarchies are inherent to YouTube as it currently exists. Not all content and its creators have equal standing. Its search and recommendation algorithms make some channels and content more visible than others, to give just one example. The power of the view in ordering interaction thus warrants further consideration.

In this article, I argue that “the view” has established itself as the central structuring agent of YouTube, and as a focus of attention, it enhances our understanding of the hierarchies and practices that have emerged on the platform. Academic research on YouTube have been conducted from the perspective of participatory culture (Burgess and Green 2009; Strangelove 2010), political economy (Andrejevic 2009; Fuchs 2017; Kim 2012; Vonderau 2016; Wasko and Erikson 2009), and within an algorithm studies context (Bishop 2018; Rieder et al. 2018). However, the concerns expressed in these scholarly works provide only partial insights into YouTube and its culture. A concern for the view opens up the assessment of the view’s value beyond that of a transactional metric and illuminates how it works as a structuring agent within the platform. The view, in fact, naturalizes power relations on the platform, as is also evident from everyday thinking, speaking, and action. To explore this idea, I use the concept of the “multi-sided market” (Rieder and Sire 2014) to draw attention to the multiple stakeholders around the view and the value it holds for them. In doing so, I take up Gerlitz’s (2016) proposal of exploring from this perspective how the data points of platforms hold together these stakeholders, which include not only YouTube and its viewers but also advertisers, content creators, and multichannel networks (MCNs). The view is embedded in the networks of actors who make and use the data points associated with it.

To develop my argument, I begin by sketching the idea of market and user information regimes—systems in which the view is firmly embedded. I reflect on ideas as to how measures, including those operating as part of those regimes, are reactive (Espeland and Sauder 2007), and argue that the view works as a categorical distinction that structures the platform. Next, I consider the market information regime of YouTube, exploring the monetization of views on the platform. I then discuss the meaning and construction of the view by looking at how views are counted. I move on to the user information regime, exploring how YouTube constructs the visibility of content through algorithms optimized for the view. I also consider how creators think and act in relation to their ideas about what the platform is and how it operates. In conclusion, I reflect on the implications of the view as, what I call, a pervasive category on the platform.

Information Regimes and Metrics

The competition among media over scarce human attention has created an “attention economy” (Goldhaber 1997; Webster 2014; Wu 2016). Media institutions and producers need to know their audiences to influence consumption and guarantee viewers for their content. As a result, audiences have become the object of scientific inquiry. To measure audiences and render them calculable, media institutions have to make them visible through quantification (i.e., numbers and statistics) and classification (Ang 1991). In commercial television, audience measurement has been the central instrument used to construct “the audience” for quite a while (Ang 1991) and, as such, it is also the dominant means to generate information about the television audience.

The ratings industry produces the audience. More specifically, ratings services construct what Napoli (2010), building on the work of Ettema and Whitney (1994), refers to as “institutionally effective audiences.” These, he explains, are audiences that can be integrated into the economics of media industries. Their representation functions as currency in the marketplace. These constructs are not fixed; rather, over time, they are renegotiated between stakeholders under the influence of technological and institutional forces (Napoli 2010). The construction resulting from this process serves particular needs and interests and should therefore be critiqued (Bourdon and Méadel 2014; Meehan 1984; Napoli 2010). Studies that take up this task examine the political economy of ratings production. In doing so, they are greatly indebted to the influential work of Meehan (1984, 217), who asked the important question, “How do ratings and the ratings industry fit into the production of the commodity message?” In her work, she criticized the dominant perspectives of the early 1980s, which were grounded in assumptions about the commodity produced by mass communication industries (e.g., Shanks 1977; Skornia 1965; Smythe 1977). Each of those approaches regarded the audience as raw material to be captured and overlooked the shaping of ratings by technologies and corporate interests. Analyzing how ratings construct the audience is of great importance because such constructions affect how content is produced and distributed.

Ratings services are a good example of what sociologists call “market information regimes” (Webster 2014). These regimes “are the medium through which producers observe each other and market participants make sense of their world” (Anand and Peterson 2000, 272). Within the mass media industry, independent measurement firms supply the measure that is used to facilitate transaction. These firms are needed because broadcasters and advertisers do not trust one another to measure the audience (Meehan 1984, 222). In the United States, Nielsen ratings have become the dominant market information regime around which the broadcast television industry operates. The Nielsen Company measures and reports on the size and composition of the audience for specific programs. Media companies and brands plan and program on the basis of the ratings Nielsen supplies.

Like the broadcast television industry, YouTube’s business model centers primarily, though not exclusively, on selling eyeballs to advertisers. As mentioned, the platform transacts sales of video ads on CPV bidding where advertisers pay for video views (or certain interactions related to them). In the process of calculating those views, YouTube reduces the complexity of watching to a single, observable human behavior (cf. Ang 1991): the act of clicking on a video so as to play it. The fact that online audiences are produced is clear, since social media platforms calculate views in significantly different ways. A view of a video on Facebook and Instagram is counted after three seconds of play (and on these platforms, videos autoplay when scrolled over). Twitter, in contrast, records a view by a click on a video embedded in a tweet. On YouTube, a view is counted after roughly thirty seconds—although, as I will show, the determination is more complex and unstable.

YouTube’s market information regime is distinct from that of traditional television in at least two ways. First, the company relies on its own metrics and does so based on its complete records of all viewers rather than on the sort of sampling techniques used in TV audience measurement (Braun 2014, 137). As it happens, the dominant self-reporting of metrics by social media companies, as well as the methodological discrepancies among platforms for establishing metrics, have frustrated advertisers (Hermann and Isaac 2016). Only gradually have third-party verifications been allowed after some pressure from advertisers.

Second, traditional ratings offered producers indirect control over the audience—lacking, however, the ability to exert direct, material influence on them (Ang 1991). Traditional television scheduling and programming employed strategies intended to capture audiences. One such strategy is hammocking: the scheduling of an unpopular television program between two popular programs in the hopes that people will watch it rather than change the channel or turn the set off. The digital environment brought with it a new information regime that helped users navigate the abundance of online content, as can be seen in tools such as search and recommendation systems. Together, these contribute to the “user information regimes” that now record user behavior to create rankings, recommendations, and statistical summaries (Webster 2014). Such means alter the way viewers relate to programming. Metadata and adaptive agents, rather than the viewer or the television programmer, now inform the flow of programming (Uricchio 2004). Their algorithms hierarchize ideas, people, and objects (Striphas 2015). User information regimes focus public attention in particular ways and shape media consumption. Importantly, just like market information regimes, they serve particular interests (Webster 2014).

Both information regimes rely on measures. It is measurement that allows for the making of judgments and the creation of hierarchies of winners and losers (Beer 2016, 1). In other words, measurement is the mechanism that enables the performance of competition. Metrics are, put simply, “those data that are used to provide some measure of the world” (Beer 2016, 1). Views on YouTube, used to evaluate the success of a video or channel, are a case in point.

As Beer (2016) explains, metrics circulate through various paths and enjoy varying degrees of visibility. And although numerous metrics are produced, not all are consequential. Only some become integrated into practices and decisions (Beer 2016). In other words, they are “reactive”—meaning that people adjust their actions in response to them (Espeland and Sauder 2007, 1). Views, I argue, enjoy a particular prominence on YouTube. Not only are they reported in YouTube Analytics as part of the company’s market information regime but they are also central to search and recommendation (albeit as watch time) and are publicly visible to all users in the interface. Moreover, as I will explore, views inform practices and decisions.

In the case of social media metrics, “the same data-points can operate in and be relevant for different valuation regimes as they can be interpreted differently” (Gerlitz 2016, 23). With this insight, Gerlitz has updated the influential work of Boltanski and Thevenot (1991), who claimed that distinct orders of worth function as sources of conflict. To clarify Gerlitz’s argument, consider again the example of the view count on YouTube. It serves variously as an indicator of social appreciation, cultural relevance, or promotional success. Different stakeholders interpret and find different meanings in measures, which, in this way, facilitate the multisided market.

In what follows, I seek to push forward the insights outlined above by claiming that what can be understood as “the view” is crucial in grasping the operation of YouTube. I aim to show how the view is a category involved in ordering interaction on the platform and demonstrate that it has material force in the world (Bowker and Star 1999). My argument is that the view works as a category in Durkheim’s sense, in that it is “put to use in various forms of structured action that naturalize wider power relationships” (Couldry 2004, 354). Enacted throughout the platform, and not exclusively in the information regimes, it becomes what I would call a pervasive category. This argument connects to the ideas discussed above about measures and their reactive capacity.

Monetizing Views

To understand the view’s significance and its importance to the production of value, we need first to understand in greater detail how YouTube works as a transactional interface. YouTube brings together different types of end-users and generates revenue in the process. Although it has experimented with subscription services (e.g., pay-to-watch channels, YouTube Red and YouTube TV), most YouTube videos are free to watch and are ad-supported. Advertisers seek to promote their brands and products to large audiences. Views have been regarded as a transactional metric through which ad-space can be bought and sold. To sell views, YouTube needs to attract viewers and keep them watching on the platform. To do so, they employ search and recommendation algorithms that help locate content with a high probability of appealing to the viewer. Here, views play a crucial role. Content creators get a percentage of the ad-revenue generated by their content and its popularity (through a high view count) and can leverage this visibility to acquire sponsorship deals with brands. Content creators have developed multiple tactics to generate views, for example, adapting their videos and their metadata according to how they think the search and recommendation algorithm operates (Bucher 2017). They can also sign up with MCNs to help mediate their relationship with YouTube. As for YouTube viewers, they can use the platform without paying and search for content that interests them. They use view counts to assess the popularity and the cultural significance of the videos they may want to watch and their content creators.

Advertisers have multiple bid-strategy options for advertising and campaigning on YouTube. To start a video campaign, they possess two options: bumper ads and TrueView ads. Bumper ads are those that employ a nonskippable six-second video format and use cost-per-thousand impressions (CPM) bidding. With CPM, advertisers pay each time an ad is shown a thousand times; this is also an option for text ads, image ads and mobile ads. ¹ Then there is the pay-per-view ad format called the TrueView video ad, where CPV bidding is the default for TrueView video ads and in-stream ads. Buyers offer the highest amount they are willing to pay for a view. Advertisers pay only if a viewer watches their ad for thirty seconds or until the end of the video, whichever comes first. However, for a variety of reasons, not all YouTube content is monetized. Perhaps the ad space has not been sold, the ad is blocked by software installed on the user’s computer, or a viewer has been on the site for a longtime.

Brands have been fearful of advertising alongside UGC (Kim 2012), because YouTube relies on “programmatic advertising,” meaning that algorithms, rather than humans, dictate the placement of ads. This process has resulted in numerous public incidents such as the Adpocalypse, an advertiser boycott of YouTube that led to a new policy of automated demonetization of content not deemed advertiser friendly. The boycott had been triggered by reports that Hezbollah was monetizing videos on YouTube to promote terrorism and concerns over the exploitation of children. Content creators have been angry about the removal of content and their exclusion from advertising revenue through moderation on the basis of unclear and inconsistent reasons (Solon 2017). As YouTube star PewDiePie (2018) explains, “Whenever my video gets demonetized I have to play this guessing game of what caused it. What was it in this ten-minute video that wasn’t allowed?” YouTube CEO (chief executive officer) Susan Wojcicki then promised to improve the appeals system after demonetization, including the development of greater transparency and increased human review of content (Wojcicki 2018).

To have access to revenue sharing and promotional opportunities, content creators must participate in the YouTube Partner Program (YPP), which was initiated in 2007. Initially intended only for the creators of the most-viewed content (YouTube 2007), the program was made available to others over the years. During its debut and expansion, a channel’s view-count served as the main factor determining whether a creator qualified for monetization. Between 2012 and 2017, the program was opened to everyone. This expansion not only produced growth but also brought with it the problem of original content being re-uploaded by others for profit; thus, YouTube decided to place ads only on channels with more than ten thousand lifetime views (Bardin 2017). This participation threshold was updated in 2018, in response to backlash over the pairing of ads with controversial content. Henceforth, channels were required to have at least one thousand subscribers and more than four thousand hours of watch time in the past twelve months to be considered for the program. ² This benchmark makes explicit YouTube’s desire that content creators assess their channels’ performance in terms of watch time. The shift in focus is important because YouTube performs better as a platform for video than other social media when watch time rather than views is used as the metric used for comparison.

Although corporations such as Disney, Sony, and Universal own the majority of the most-viewed videos on YouTube (Fuchs 2017), there are a handful of independent stars like Daniel Middleton (DanTDM), PewDiePie, and Jake Paul making millions of dollars on YouTube. The standard advertising split is 45 percent for YouTube and 55 percent for content providers. These days, the proprietors of YouTube channels often sign with MCNs to increase their audience and advertising income. MCNs sign up popular channels to their network, then sell advertising, cross-promote affiliated channels, and develop brands around celebrities. In exchange, they receive a percentage of ad revenue. ³ MCNs have contributed to asymmetrical relations between content producers and YouTube by standardizing production, reorganizing CPM ad sales and, through aggregation, shifting the value away from videos to channels (Vonderau 2016).

The successful YouTube gamer Kwebbelkop (Jordi van den Bussche) has revealed that 30 percent of his income derives from views, with the rest coming from brand deals. He mentions that views were important for making money when he had just started, but later in his career brand sponsorships and deals brought in more money. He recognizes, however, that these successes were a direct result of his high view count. As he states, “Everything is actually based on those views. Views still are what’s most important” (Kwebbelkop in VPRO Tegenlicht 2017, author’s translation). Interestingly, the YouTube Creators Academy, where YouTube offers free instruction and training for creators, tries to persuade creators that views matter less than watch time and subscriber count. ⁴ The Academy explains that views are simply video loads and are therefore not important; watch time, by contrast, is what truly registers engagement, since here one can see that people are explicitly choosing to watch something. That YouTube feels that it has to address whether views are important underscores their perceived social significance.

Counting Views

When asked to define a view, YouTube analytics product manager Ted Hamilton responded,

A view should be a video playback that was requested by an actual user who got what they were intending to get and had a good user experience. We think of views as a currency and therefore we have to make a significant effort to eliminate counterfeit views, if you will. (Von Baldegg 2012)

The statement squarely positions views as a form of currency. It furthermore reveals what YouTube thinks the view acts as a proxy for: intentionality and subsequent satisfaction. In what follows, I address how views are counted, touching on two issues: the gaming of the system, which relates also to the moderation of views, and the malleability of the view’s definition.

To offer creators insight into the performance of their content, YouTube Insight was introduced in 2008. This tool supplied analytical information about videos and included data about the frequency of traffic and views and their geographic origin; it also offered insights into the life cycle of videos. In announcing the tool, then product manager Tracy Chan (2008) claimed it would help creators “to increase views and become more popular.” Her phrase suggests that views are indeed sought after and relates them to the idea of popularity. Popular discourse clearly shows that for users of the platform, views are an indication of success and mark out those who should be heard. The aforementioned tool has since been rebranded as YouTube Analytics, and its features are continually being tweaked. The selection of metrics included in the dashboard is significant in that it communicates what constitutes productive YouTube usage and shapes how the platform is employed in practice (Beer 2016, 93). Watch time and the watch time-report replaced the view report in YouTube Analytics in 2015. I would argue that watch time, rather than competing with views, is in fact nothing more than a prolonged view. It is a prolonged view in the sense that generating watch time also requires generating a view. Watch time does, however, shift emphasis to the duration of views. It is actually a means to hierarchize views. Watch time fuels the myth of viewer intentionality and gratification, on which the understanding of views has been perfected, and therein contributes to the category “the view.”

In YouTube Analytics, aside from what appears on the dashboard, which allows creators to check on the performance of their videos, some other metrics are also made publicly visible. YouTube displays the total number of likes, dislikes, and comments that a video has received on the content page. Yet here, too, views function as a privileged metric, apparent from the count tally’s prominent position—directly beneath the play button—and comparatively large font size. The design of such interfaces plays an important role in shaping how metrics are understood and used (Beer 2016, 101).

Counting views is clearly a way of evaluating success on YouTube. It allows the performance of videos and channels to be compared. Views are a focus of aspiration. As Espeland and Stevens (2008, 416) explain, “The ‘outliers,’ ‘under-achievers,’ and ‘under-performers’ produced by performance measures become targets of manipulation, disapproval and anxious self-scrutiny.” Indeed, content creators on YouTube try to game the system and manipulate metrics like views. That they do so is telling, indicating the view’s enduring value. Manipulation happens through automation or trickery. Automation involves the artificial increasing of metrics by automatic systems. There are also many online companies that help with this sort of activity, selling YouTube views, likes, and comments. Trickery concerns the use of misleading titles and thumbnails to drive clicks on a particular video. Here, a line is drawn between misleading viewers and optimizing videos for search and recommendation algorithms. Because of this, YouTube has repeatedly warned content creators about artificially inflating views out of fear that these practices affect accuracy and hence might diminish trust in the metric. They also have threatened to take disciplinary action (e.g., removal of videos, suspension of accounts) against those creators caught paying third parties for views.

In 2012, in a widely publicized effort, YouTube stripped more than three billion “fake” views from channels showing music videos. It claimed it was enforcing its view-count policy but provided no means to allow independent checks on its measurement or verification. It has since introduced a new moderation system for views that claims to count views from “real people” as they are recorded. Prior to the introduction of this new verification method, they would freeze the counter at 301 while the system subjected a video—stored on multiple servers around the world—to a statistical verification process designed to verify traffic. They removed views that represented the activity of bots and those that had been amassed with the help of misleading thumbnails. This process took several hours and videos with fewer than 301 views were not verified. Nowadays, views deemed valid by the system (the actual criteria used are unknown) are shown in the first hours after the video is published. The view count is updated later as more views are validated.

YouTube is not transparent about how its spam and security-breach system works. Moreover, no one outside the company can evaluate either the verification or validation of the methods they use to identify attempts to inflate view counts (or other metrics) or the specifics as to why content has been removed. This is perhaps not even possible within YouTube, because in deep learning the software engineers do not understand the relations among inputs made by the algorithms. Tweets directed at the @YouTubeCreators account on Twitter reveal that creators experience fluctuations in metrics and want to know why their content has been demonetized. These incidents result either from the work of YouTube verification methods, “one-time error” (YouTube 2013) or changes to their methods. Although YouTube responds to these issues, they are always vague about what has happened and how it has been resolved.

Although content creators game the system in various ways, YouTube always has the authority to change the way views are counted. TechCrunch has reported on how TrueView ads allow for a legitimate way to inflate video counts by changing how views are counted. With TrueView ads, views are counted toward the video views for a given channel. Movie trailers, for instance, often seem to generate millions of views. However, such numbers derive largely from views of the teaser ad rather than the full trailer (Schonfeld 2011). Asked to comment on this, a YouTube spokesperson explained,

When it comes to paid advertising, view count of a video increases only when it’s clear that a viewer has made a choice to watch a video. For example, a viewer might make that intention known by clicking to watch a Promoted Video, opting-in to watch a TrueView in-stream ad, or playing a YouTube video embedded in a homepage ad. . . . To be clear, with standard in-stream ads where a viewer has no choice in its selection, we do not increase the video’s view count. (Schonfeld 2011)

Again we see how the view is connected to viewers’ intentions as determined by their actions. TrueView ads play automatically, meaning that the user has to deliberately choose not to watch. With autoplay set as the default, users are nudged toward the consumption of these ads. By design, the nonskippable ad format of bumper ads helps boost view count. Content creators have expressed dissatisfaction with this practice, stating that it has resulted in “fake views” and has been misleading about their level of popularity.

Similarly, in 2015, YouTube introduced cards for TrueView in-stream and reclassified certain kinds of engagement (i.e., clicks) as views. More specifically, advertisers started paying for TrueView in-stream ads when the user watches thirty seconds or the entire ad, whichever comes first, or clicks on any of the following: call-to-action overlay, a front or end card, a companion banner, or any link in the video viewer. This change marks a significant shift in the understanding of views, as a user clicking on a particular item is also counted as a view, even though the video may not necessarily be watched for thirty seconds or more. The expansion was justified with the claim: “we believe this [certain clicks] is a strong indication that they are interested and engaged” (Fein et al. 2015). YouTube’s ability to alter how views are counted highlights how the view is a construction and reveals an imbalance in its power relations with advertisers and content creators. The way that YouTube justifies expanding its definition of views underscores that the view represents a myth of viewer intentionality and satisfaction.

Algorithms, “the View,” and Visibility

As mentioned, “the view” plays a crucial role in the user information regimes of YouTube—that is, the recommendation systems that rank content and channels. What is desired and aspired to by multiple YouTube stakeholders is content that generates a high number of views. MCNs and creators adjust their practices in making and publishing their content according to their understanding of how the algorithms of these systems work.

YouTube offers multiple features for content discovery. In addition to channel subscriptions, there is its search and recommendation system, which shapes what its billions of users watch. They serve to keep users on the site and to influence what is popular. YouTube’s Chief Product Officer has stated that more than 70 percent of viewing hours on the site come from recommendation-driven viewing (Solsman 2018). YouTube’s ranking algorithms use metrics produced for videos and channels. However, in these systems, not all metrics are valued the same.

With regard to the YouTube recommendation algorithm, view counts were initially an important factor in filtering and ranking. YouTube now claims that it no longer ranks on the basis of views because such ranking rewards clickbait content and prevents users from seeing new videos (Meyerson 2012). The company claims to have developed “better” signals for determining how engaging content is. Subsequently, in 2012, YouTube updated its algorithm to optimize for watch time:

If viewers are watching more YouTube, it signals to us that they’re happier with the content they’ve found. It means that creators are attracting more engaged audiences. It also opens up more opportunities to generate revenue for our partners. (Meyerson 2012)

The underlying assumption here is that the length of time spent on YouTube can serve as a proxy for a viewer’s happiness and level of engagement. This conception is how YouTube extended the notion of the view and the intentionality of viewers that it captures. Watch time is more than a simple multiplication of views with average view duration. It concerns the overall time people spend watching YouTube (an entire session) and how that relates to a specific video (Meyerson 2012). The link between watch time and YouTube’s business model is clear in the above citation: generating revenue for partners means generating revenue for YouTube as well. This particular construction of the audience impacts content production and distribution.

Google researchers, Covington et al. (2016), published a paper describing YouTube’s video recommendations system. They explain how the system uses two neural networks for filtering; one for candidate generation and the other for ranking. Filtering for candidate generation is achieved on the basis of a user’s YouTube activity history and collaborative filtering (i.e., what similar viewers have watched). Subsequently, the ranking filter

assign[s] a score to each video according to a desired objective function using a rich set of features describing the video and user. The highest scoring videos are presented to the user, ranked by their score. (Covington et al. 2016)

The “desired objective function” is the prediction of expected watch time. Many factors are used to determine which videos would seem most relevant, but the actual formula behind this assessment has been kept secret. The paper does reveal that “freshness” (meaning recently uploaded content) and frequency of uploads are important. It also suggests that videos are demoted if users choose not to click on them after they have been recommended.

As for search recommendations, YouTube stated in a 2014 handbook that the factors involved in ranking relevance included watch time, recency, duplicates, and query relevance. ⁵ More recently, Rieder et al. (2018) explored search ranking on YouTube. Although they were unable to link ranking to metrics such as likes, views, or content, they did find that “newsy” moments drive ranking and found a strong preference for YouTube–native content over mainstream material. It has been speculated that the search algorithm is, however, optimized for relevance rather than watch time (Chaslot 2017).

The Algorithmic Imaginary

As mentioned, creators, in pursuit of view counts, adapt their practices to their understanding of the YouTube algorithms. Their understanding of what algorithms do and how they work has been referred to as the “algorithmic imaginary” (Bucher 2017). This imaginary is shaped by the experiences of creators and a glut of (online) resources explaining the primary principles of YouTube’s algorithm. It is an imaginary in that it concerns how these algorithms are envisioned to work rather than how they actually operate (which is often unknowable). It is, however, necessary to expand the focus a bit and also to reflect more broadly on how the use of the platform is affected by the imagination of what the platform is and how it operates. Users’ interpretation of the valuation of metrics here plays an important role.

Bardo Ellens and his team, responsible for producing comedy sketches for the YouTube channel Banjomovies, as well as the popular YouTube gamer Kwebbelkop state that they attempt to boost viewership of their channel by making their content more appealing to the algorithm (VPRO Tegenlicht 2017). Specifically, they mention making content that is longer than ten minutes, as well as their frequent uploading of content to the platform. The idea that creating longer videos improves watch time is also found on websites that provide YouTube creators with tips to succeed. The YouTube Creator Playbook for Brands provides advice about “best practices” to maximize discoverability. These suggestions include the making of videos around trending events, the creation of “evergreen” videos (videos about topics that people always search for, such as how-to videos), and the optimization of metadata. ⁶ This has resulted in an array of creator practices: creating know-how content (Bishop 2018), issue hacking, and tagging videos with terms related to current events to gain views (Rieder et al. 2018). MCNs play an important role here as they drive their clients’ content to the top of a results list by making them appeal to YouTube’s algorithms as they understand them (Lobato 2016).

The sort of content that is largely determined by YouTube algorithms in the quest for visibility has been criticized. YouTube promotes videos consistent with its commercial aims (Bishop 2018), which means prioritizing watch time. These videos are more easily found through search and recommendation. PewDiePie (2018) discussed his frustration with this phenomenon:

YouTube right now really pushes for watch time. Unless you upload regularly and long videos you are going to suffer. . . . This is what you should be pissed off about. That putting a lot of effort and really good channels don’t get as much exposure as they might deserve because they don’t fit in this quota.

These remarks reveal how the algorithmic imaginary incentivizes content creators to make longer videos and to post frequently. The goal of extending the amount of time people spend on videos is important to YouTube’s bottom line; the quality and diversity of content that people see is at stake but is irrelevant. The algorithms, as a result of their focus on the view, are optimized to entertain rather than enlighten viewers and broaden their horizons. This contributes to the culture of the platform, as these algorithms cause particular types of content to rise to the surface.

There are also strategies used by creators that do not focus on the algorithm and its emphasis on watch time. For instance, Kwebbelkop reveals that in deciding which games to play, he searches for those games that are already popular on YouTube. He identifies popular games through the number of views generated by his competitors. Here, views are being valued not for their transactional potential, at least not directly, but as a measure of what is popular.

(Re)producing Hierarchies and Inequalities

Although content on YouTube is measured in various ways (through clicks and other forms of engagement), it is monetized through views. The view interconnects YouTube, advertisers, creators, and MCNs, but it is YouTube that decides how to count views and which content to monetize. Advertisers, creators, and MCNs have rather limited means to hold YouTube accountable for its procedures. In this article, I hope to have shown that the view is a category that helps to legitimize hierarchies on the YouTube platform. As explored, these hierarchies concern, and impact, participation, financial compensation, visibility, and popularity. This suggests that the celebratory discourse around YouTube as an empowering tool that levels the media playing field (Jenkins 2006) was positively misguided.

Part of the hierarchies and inequalities mentioned are produced through the information regimes of YouTube. These help to construct the media environment (Webster 2014) and therefore require scrutiny. With regard to the market information regime of YouTube, problems emerge as to the transparency and accountability of reported metrics. There is no independent measurement firm supplying YouTube with metrics, and so YouTube has the leeway to define the audience in a way that best supports its own interests. Moreover, the programmatic placement of ads on YouTube has made advertisers anxious that the platform is unknowingly helping to broadcast extremist messages and hate speech. For this reason, they demand more information about the placement of ads, threatening to withdraw their advertising dollars if action is not taken to address these problems. Brands, paying per view of video ads, additionally face the challenge that views are constructed differently across the online media landscape.

In addition, YouTube’s user information regimes make channels and content with higher watch time, understood as belonging to the category of the view, more visible in the recommendation system. They allow for direct behavioral control of audiences by nudging them to certain kinds of consumption. The reality of traditional television, where the observation and regulation of bodies were disentangled through audience measurement (Ang 1991), is no longer with us. Not surprisingly, content creators, in pursuit of views, attempt to make their content appealing to the site’s search and recommendation algorithms. Their ideas about which types of content are made more visible by these algorithms influences choices about what to produce, when to circulate one’s content and how to input metadata.

The view also produces hierarchies and inequalities outside of these information regimes. For instance, it does so not only in the assessment of popularity of videos and channels but also in the conditions for participating in the YPP. With the entrance of MCNs into the marketplace, the value of views has changed as well. MCNs cluster channels and sell market segments to advertisers, creating multiple marketplaces within YouTube. Their resources and knowledge enable them to deploy strategies that make associated content (creators) more visible and that package views to sell to advertisers. YouTube holds out the egalitarian promise that anyone with an Internet connection and a video camera can become a media tycoon. In reality, however, creators are beholden to the power of YouTube and, more specifically, to the power of the view.

Bowker and Star (1999) were concerned with the work of classification in ordering human interaction and how the categories involved in systems of classification are made and kept invisible. They argued that classifications do political and ethical work. My goal has been not only to make visible the construction of “the view” as a category but also to expose its deep entanglement in YouTube’s ontology. It is a pervasive category in that it is enacted throughout the platform at multiple sites. In these places, it orders practices and naturalizes hierarchies and inequalities. It is particularly influential in the central role it has acquired in the YouTube information regimes. Its involvement in coordination makes it a category in Durkheim’s sense and connects it to the workings of power. As a metric, it has gained extreme visibility and has led to grave consequences. As a myth, it supports the idea that the view says something meaningful about audience consumption and preferences. Its centrality has contributed to what has been described as the toxic culture of YouTube. With so much at stake with views, the need is apparent for more transparency and accountability over how the YouTube audience is constructed. Rather than to place the blame on the YouTube algorithms alone, which has been the tendency, I hope to have clarified that it is the myth of the view that needs contesting.