18

Counting, and Accounting for, Online Audiences

Fernando Bermejo

The concept of the “audience” as it is commonly used today belongs to a set of ideas that refer to human collectivities.¹ In spite of this, the collective nature of the audience is rarely expressed directly. This is, of course, a historically bounded assertion, since, as Butsch (2000) has shown, the behavior of the audience as well as its capacity to express itself as such have starkly changed over the past three centuries. The audience of live performances, which dates back to antiquity, “was localized in place and time. An audience occupied an ‘auditorium,’ a space in which to hear and see what was going on and to respond directly” (McQuail, 1997, p. 3). It was thus able to perceive itself as a collective and to act as such. However, when technological devices were inserted into the communication process, physical co-presence ceased to be a necessary element of audiencehood, and immediate collective reactions became a much rarer occurrence.² In fact, the evolution of the mass media up to the late twentieth century can be described as a progressive separation from the lively audiences of performances (Butsch, 2000). If we consider the mass media that have been the dominant form of public communication throughout the past century, we can observe that the audience of those media hides in the privacy of media consumption,³ each member being disconnected from other members of the collective⁴ and only rarely asserting his or her own reality as part of an audience and acting upon it.⁵ In the era of mass communication media, the audience is for the most part the result of the defining work of others.⁶ And the audience measurement industry and the academic field of audience research are the most significant others generating definitions of the audience.

With the prominence achieved by online communication since the 1990s, both the audience measurement industry and the audience research field have faced the need to reconsider their tools and their concepts in order to keep providing those definitions. And this reconsideration has direct consequences on the institutional dynamics of the media, on the configuration of communication studies, and, ultimately, on the very reality of the media under study and of their audiences. And this is so because, for decades, the depictions of the audience provided by audience measurement have played a pivotal role in the economic and institutional dynamics of the media, especially in broadcasting (Ang, 1991; Napoli, 2003); because the field of audience studies has become in recent decades one of the most dynamic areas within communication research and conceptualizations of the audience have constituted a fundamental element of most communication theories (Allor, 1988); and because the way in which different social actors – from policymakers to media professionals and owners – conceive of and act upon the media depends to a large extent on the definitions of audiences that are available to them.⁷ In order to understand the changes triggered by the advent of the Internet as a relevant form of communication in these defining practices, the present chapter offers an analysis of the evolution of audience measurement in the online environment over the past 15 years – in others words, of the counting of online audiences. This analysis will in turn allow us to examine how commercial audience measurement may influence academic conceptions of and discourses on audiences – which can be considered modes of accounting for online audiences.

Counting Audiences

The practice of counting audiences, or audience measurement, has played an essential role in the economic and institutional functioning of the mass media for decades, and it has been closely linked to the use of advertising as a media business model.⁸ This practice can be conceived of as a subset within the field of market research, which, in turn, can be considered as one of the different forms of control exerted over collectivities that characterize modern societies and nation-states.⁹ However, the shape and relevance of these techniques of control have evolved over the decades, in parallel with the evolution of dominant technologies and modes of communication.¹⁰

As a form of control, audience measurement shows an almost paradoxical nature, since usually it is more necessary when its implementation seems more difficult. The origin of this paradoxical situation can be traced back to a large extent to the ways in which the degree of control over access to communication can orient different media technologies toward specific business models. In modes of communication in which there is a clear spatial component – such as live performances or theatrical film releases – it is usually simple (1) to demand some kind of payment in exchange for access; and (2) to calculate the size of the audience either by counting ticket sales or simply by counting the people in attendance. In fact, in these communicative modes, advertising often plays a secondary role – if any – and there is little need to implement a specific and complex mechanism for measuring audiences. In modes of communication where there is no access to a particular space that can be controlled but there is a physical container – such as a book or a disc – that holds information, it turns out to be relatively easy (1) to demand some kind of payment in exchange for the physical container; and (2) to come up with a rough estimate of the audience from the number or copies sold. Again, in these modes of communication reliance on advertising tends to be low, and estimates of the audience are easy to produce – even if they are not exact, since the precise relation between the number of copies sold and the actual size of the audience is often unknown.

Broadcasting, on the other hand, offers an obvious contrast with the modes of communication described above.¹¹ In broadcast radio and television there is no space to which access can be restricted, and there is no physical container that holds the information. Thus it turns out to be rather difficult to demand payment in exchange for content. As a result, advertising turns out to be an almost natural business model for broadcasting.¹² And, for advertising to flourish, it is necessary to know the size and profile of the audience. But, since there is no physical space in which to count audience members and no transaction record from which to infer audience size, there is a strong need to develop specialized and complex forms of audience measurement – hence the relevance of audience measurement in broadcasting.

The online environment is rather peculiar in this regard. First, pay-models have not been easy to implement in the online environment and advertising has become a very relevant source of revenue – which, in turn, opens the door to audience measurement efforts.¹³ Second, even if control over access to communication products is not always possible, online communication leaves a trace, and there is always a record of the connection between senders and receivers. However, this record is so complex, so difficult to clean and analyze, so open to errors and omissions, that it requires deploying elaborate measurement strategies to estimate the size of the audience.

In fact the birth of the field of online audience measurement can be dated to 1995: it coincides with the advertisers' increasing interest in new communication technologies such as interactive television and the Internet. During the first years of the 1990s the Internet underwent a profound transformation and, while preserving most of its structural and defining traits, it become a very different network from the ARPANET, from which it grew.¹⁴ In those years, three related processes took place that help explain the birth of online audience measurement at that particular juncture. First, the Internet underwent a process of privatization and commercialization. All the different areas of the Internet's management and infrastructure were transferred into private hands, and commercial interests became a staple at all levels of online communication – from the actual telecommunication backbones to the content and services provided online; from the software necessary to connect to the net to the actual connectivity through Internet service providers. Second, the Internet became a mass medium in terms of the number of its users. For the first time, online users could be counted in the dozens of millions, and for a few years in the mid-1990s the number of Internet users worldwide more than doubled yearly. Third, asymmetry became a feature of online communication due to the introduction of a communication tool – the World Wide Web – that was basically unidirectional.¹⁵ The popularization of the Internet in the mid-'90s was due, to a great extent, to the advent of the web, a tool that made it extremely easy to consume information published by others but did so at the expense of becoming, at least for a few years, a medium in which content was published by a few and consumed by many (Roscoe, 1999). In this environment of commercialization, popularization, and asymmetrical communication, it is no surprise that the advertising industry became interested in online communication, since these were the same defining traits of the environment where it had thrived for decades: the traditional mass media.¹⁶

The first advertising banner appeared in 1994. At that moment, online publishers and the advertising industry understood that online advertising could become a significant source of revenue for all, and that this would only be the case if an online ratings system was set in place (Turow, 1997). As a result, starting in the spring of 1995 and in the space of only a few months, several companies and organizations decided to implement different procedures to measure online audiences. The online audience measurement industry had been born.¹⁷

Measuring Online Audiences

The online audience measurement industry created in the mid-1990s was clearly shaped by three main forces that conditioned its form and evolution. First, the general institutional framework from which it grew called for measurement efforts able to produce data that could serve as a coin of exchange in the trade of audiences constituting the core of the mass advertising business. This general framework required, as had been the case in the measurement of the audience of previous mass media, data on audience size and composition of the audience of specific communication vehicles. Second, the new industry did not grow in a vacuum, but in the context of an established audience measurement industry that had for decades played a very relevant role in the functioning of the mass media. This ratings industry had structured itself around the model of independent third-party organizations that were able to reconcile the conflicting interests of the different stakeholders in the audience market (Miller, 1994). These organizations established a form of measurement that revolved around the idea of exposure to specific communication vehicles – what was actually measured was the number of people who had been exposed to a particular communicative product; any other objects of measurement were ignored for the most part. On top of this, the audience measurement industry had accumulated a great deal of methodological know-how that could be put to use in the measurement of online audiences (Webster, Phalen, & Lichty, 2000). In particular, the industry had focused on developing combinations of sampling methods and sophisticated data collection tools that could satisfy the demands of rigor among sellers and buyers of audiences. Third, while the general institutional framework and the legacy of the existing ratings industry pushed online audience measurement toward a line of continuity with previous forms of measurement, the new industry was confronted with a medium that presented rather novel technological features and generated distinctive patterns of use. If these evolving methods of measurement could not be properly adjusted to the new context, the whole established commercial institutional arrangements surrounding the counting of audiences could be put into question.

The persistence of the institutional framework demanding a coin of exchange for the audience trade was guaranteed by the determination with which the advertising industry promoted the development of online audience measurement in the second half of the 1990s. Led by Procter & Gamble and other large advertisers, and with the support of the main trade associations, the advertising industry facilitated the creation of different coalitions and organizations aimed at promoting and shaping online audience measurement. Through the Coalition for Advertising Supported Information and Entertainment (CASIE), the Future of Advertising Stakeholders (FAST), and the most lasting and influential Interactive Advertising Bureau (IAB), advertisers and, to a lesser degree, publishers attempted to fuel the online advertising industry through the development of standards and best practices that could be adopted by all participants in the audience trade. In turn, the legacy of the preexisting audience measurement industry has manifested itself through the role played by many of the established measurement organizations in the development of the new sector and through their attempts at becoming the independent third-party of reference by improving on existing methods and/or acquiring a position of market dominance.

But, in spite of general continuity in the measurement effort in terms of aim, logic, and procedures, online audience measurement had to adjust to some specific traits of the new environment in which measurement operations were being implemented. The Internet as a communication tool presents obvious differences compared to previous media, and some of those differences have direct consequences on the particular forms measurement should take if it has to perform the role that is required of it. In particular, it should be noted that the Internet is an intrinsically global medium; it is a medium that is not bounded to the home, and it is regularly used from multiple access locations; it is neither a flow – like radio or television – nor a closed product, like print publications; it leaves a trace of every transaction of information taking place over it; it presents an endless number of communicative vehicles; and it can be accessed from a multiplicity of platforms (Bermejo, 2007).

These peculiarities explain, in turn, some of the problems and limitations of the different methods that have arisen to assess the size and composition of online audiences. In this sense, when the industry started to develop online audience measurement initiatives, two main routes seemed to be at its disposal. The first route entailed the adaptation of the measurement methods already in use in the audience ratings industry, in particular surveys and panels. The second route involved the development of new tools, which were based on the use of the records left behind by online transactions in order to estimate the size and profile of the audience – these took the form of log-file analyses and tag systems. Both routes have been followed by different measurement operations since the mid-1990s, and surveys, electronic measurement panels, log-file analyses, and tags have all been part of the measurement panoply used by the industry. However, all these methods have shown obvious shortcomings in their attempts at measuring online audiences and none has been able to become the standard of reference.¹⁸

Surveys – defined as the use of standardized questionnaires to collect information from a representative sample of a population – have traditionally been an essential pillar of the processes of audience measurement, particularly in the measurement of those media that do not easily lend themselves to the use of automatic devices attached to receivers in order to collect information on usage – mostly print publications and, for most of its history, radio.¹⁹ However, when they have been used to measure the audience of online media, their limitations have only been accentuated. This has been the case both because of sample size demands and because surveys essentially depend on interviewees' memory and sincerity –neither of which can be easily relied upon when measuring online audiences. The larger the number of vehicles – radio stations, television channels, websites, and so on – that need to be measured, the larger the sample required to offer stable data from all the different vehicles. If the audience is scattered across many different options, those options that attract a small portion of the total audience will face larger margins of error and high variation across subsequent measures of their audience. Due to the huge number of possible options available online to capture audiences' attention, this issue turns out to be particularly relevant, and surveys need to have sample sizes that are not always possible or viable in practical terms. In addition, the expectation that online users will remember, and be able to recount, all the different websites they have visited and to specify how much time they have spent in each and what they have done while there is simply unrealistic. Likewise, expecting that survey interviewees will be willing to share with interviewers the details of all their online activity does not seem sustainable. As a result, surveys have played a marginal role in attempts at measuring online audiences. They have been used, for the most part, to estimate the global size of the online population and to understand macro-level issues such as frequency of use, speed connectivity, location of use, and so on. But they haven't been able to provide relevant data on the size and composition of the audience of specific online outlets.

The essential difference between panels and surveys lies in the use of samples. While in surveys the population sample from which information is gathered is normally used once, in panels the same sample is used over time to collect information on subsequent occasions, thus providing a longitudinal view of the evolution of the variables being measured. In their implementation for audience measurement purposes, panels are usually associated with the use of metering devices; while there are panels that use diaries instead of meters to collect information from the members of the sample, measurement operations that use meters are almost by default panels, since it does not make practical sense to install meters in a sample in order to gather information from it just once. As such, panels have been the dominant technique for the measurement of the television audience through the use of “audimeters” (that is, meters for audience measuring). And, since television panels have been, for decades, socially and economically the most relevant form of measurement, it seemed only natural to apply the same model to the online environment when the need to start measuring the Internet audience became apparent (Buzzard, 2003). The only significant difference in the implementation of panels has derived from the different nature of metering devices. While in broadcasting the meters have always been a piece of hardware attached to radio and television receiving devices – and, most recently, carried around by members of the sample being measured – in the online space these meters take the form of software tools that can be installed in computers to monitor online activities.

However, when used to measure the online audience, panels have faced three main hurdles that were not present – at least not to the same degree – in television measurement. The first one, as is the case with surveys, has to do with the size of the sample and the need to draw very large ones in order to cover a significant portion of the number of available online vehicles. Traditional online panels have offered statistically significant data for only a few hundred websites – often less than 300 – which represent a very small portion of all available websites.²⁰ The second hurdle faced by online panels has to do with the geography of online use. Audience measurement panels have usually drawn their samples from national populations. This means that the audience they measure is circumscribed to specific national borders. And, while that might not have been a relevant issue when measuring broadcasting audiences – in spite of the international reach of some of its vehicles, broadcasting has traditionally been bounded to national borders – it actually turns out to be rather significant in dealing with such a transnational medium as the Internet. As a result, online measurement panels have been unable to measure the portion of the audience of online vehicles that are located outside the country where the panels are deployed. The third main hurdle faced by online panels has to do with the different patterns of media consumption. While television, the medium in which panels have achieved an indisputable success, is clearly tethered to the home – television consumption has traditionally taken place in domestic spaces – that is not the case with the Internet. If an online panel wants to measure the online audience, its sample cannot be limited to homes but has to take into account work spaces and “third places” – cybercafés, libraries, community centers, and the like. This has proven to be – in spite of significant efforts – an insurmountable hurdle for online panels. It is almost impossible to draw a representative sample of computers connected to the Internet in places other than homes and, even if such a sample could be drawn, it would be very difficult to deploy due to general resistance from companies and organizations to installing metering devices in their networks' computers. In spite of these obvious limitations, online panels have been widely used over the last 15 years or so, as a source of data on online audiences.

But, besides other methods already used in the measurement of previous media audiences, the Internet offers an alternative route that is not open to other media. Online activity generates endless amounts of information that is constantly recorded, stored, and analyzed. This information can help in measuring online audiences, and two main techniques have been used to take advantage of it. The first technique is usually referred to as log-file analysis. Log-files are text files that are automatically generated by Internet servers; these files record the activity generated by requests to them. They store information such as Internet protocol (IP) addresses from which requests are made, cookies in the users' browser, browser type, operating system used by the client, time stamp, and so on. But, while the richness of this information is unquestionable, measuring online audiences through the analysis of log-files turns out to be a rather complex process, fraught with technical problems, imprecision, and limitations. However, besides the fastidiousness of the cleaning and analyzing process, perhaps the main problem with the use of log-files for audience measurement purposes has been the inability to link the data recorded on online activity with individual users in a consistent and precise way, and the impossibility of deriving a clear demographic profile of those users.²¹ The use of tags, tiny invisible markers and code inserted in web pages that allow for the collection of information similar to that stored by log-files, has facilitated the automatization of the analysis, but it has not been able to overcome limitations in connecting activity with users.²² This means that the analysis of the records of online activity has serious shortcomings when it comes to offering an exact quantification of the size of the audience and a detailed profile of its composition.

Casting the Role of Measurement

Besides – and partly because of – the difficulty in adjusting the measurement process to the peculiar characteristics of the online environment, audience measurement has also been confronted with a transformation in its institutional role within the field of online commercial media. In broadcasting, ratings constituted an essential pillar of institutional arrangements: they guided most programming decisions, shaped the internal functioning of the media, and determined their financial success.²³ The extensive power held by ratings derived from the role they played as the fundamental piece of information that shaped advertisers' relations with the media. Ratings provided advertisers with (1) the guarantee that they were receiving what they were paying for;²⁴ and (2) the information necessary to target their ads – in terms of audience composition.²⁵ However, the online environment offers a much more complex picture.

In this scenario, measuring the size and profiles of an audience remains an important activity, but it does not occupy the central role that it held in broadcasting. The reasons for this are multiple. Of course, the difficulty in measuring the online audience analyzed above makes it harder for the results of the measurement to be accepted as a common currency in online economic exchanges involving advertising. A decade and half after its birth, and in spite of a process of consolidation, online audience measurement does not show signs of achieving the stability that the ratings industry presents in other media – there are competing measurement operations, competing methodologies, and competing general approaches to measurement. Besides this, the very diverse shapes taken by online communication and the many different business models that sustain them contrast with the relative homogeneity of previous media and business models, making it more difficult to find a common point of reference from which to explain the economic logic of the medium and to refer to an anchoring institutional arrangement. In this context, audience measurement seems to have lost its centrality in the institutional arrangements of the media in their online transition, because new forms of managing and targeting advertising have emerged – such as search-related and behavioral advertising – that seem to comply better with the characteristics of the online environment while being able to function without any audience measurement as such; of course, other types of data collection practices are necessary. Flaws, disputes, and competition in the measurement process are not in themselves enough to dethrone established institutional arrangements. As Napoli (2011) has pointed out, a viable alternative to these arrangements needs to be available if significant changes in the existing order are to take place.

In order to understand the viability of these possible alternatives to the system of trading audiences that is based on data provided by the measurement industry, some general dynamics of the online environment and of the online advertising management process need to be examined. There are, in particular, three specific characteristics of the online environment that seem to shape the evolution of online advertising and to influence the relevance of measurement:

1. The Internet Leaves a Trace: As pointed out before, a record of online activity is routinely generated and automatically collected as part of the activity itself. And if this record can be used as the source of information for targeting ads – and not necessarily in the form of audience measurement – there is no need to devise and implement complex and costly data gathering procedures for the advertising market to flourish.
2. The Internet Affords More Interactivity Than Any Other Communication Medium: While previous communication media were for the most part unidirectional and their usage flexibility was rather limited, the Internet allows for endless forms of user activity: reading, viewing, clicking, writing, playing, purchasing, and so on. This means that the restriction to exposure as the key piece of information for managing online advertising does not account for the complex relationship between media and users. Thus, and in combination with the above, huge amounts of information about audiences, derived from the many different forms in which they interact with the medium, can be easily and automatically collected and then used in the process of targeting ads.
3. Over the Internet, Ads and Content Are Managed Separately and per Request: There is no necessarily predetermined link between specific ads and specific communicative products, nor between specific ads and specific users. That is, every online request can return a different ad, breaking the traditional relationship between media outlets and advertising – in which content, ads, and audiences were bundled in a comprehensive whole.

If we look at these three issues in reverse order, we can say that, while the traditional form of managing the advertising process was based on the implementation of complex measurement procedures that sampled portions of the population to assess the exposure of select groups of individuals to specific communication vehicles, in the online context (1) there is no need to measure specific vehicles – since the content has been unbundled and advertising can be managed by individual requests; (2) the idea of exposure constitutes but a tiny fraction of users' engagement with the medium – it does not cover most of the interactive nature of the relationship; and (3) information can be collected on all users – not only on a sample of them – by simply capturing and analyzing the trace left by online interactions. Taking this into account, it seems easier to understand how traditional forms of audience measurement may be losing their central role in the context of online communication.

If we examine the evolution of online advertising over the past decade, it is possible to see new forms of advertising – such as search engine and behavioral advertising – that have gained increasing relevance and have adjusted to the three characteristics described above without demanding audience measurement – at least not in the way in which it was developed in broadcasting, or even in the first years of online measurement. Search engine advertising is based on the keyword searches of users seeking answers to their queries. Both because of the particular relation of users to the search process – they are looking for something – and because their interests, wants, and desires are expressed in relatively concrete and concise terms, using those keywords as a way to link users and ads has turned out to be a rather effective and lucrative way of managing online advertising (Bermejo, 2009). Thus search engine advertising utilizes information provided by users – not captured through any additional data collection process – that goes beyond exposure – users are not exposed to content, they are searching for it – and these specific quests for information are used to target individual users – there is no longer a need to target a collective or a mass/undifferentiated audience.²⁶

Behavioral advertising is based on the collection of information concerning all possible online activities of users, a collection that is made in order to create an individual profile of each of those users; and this profile serves as the key piece of information to determine which ads are going to be displayed. Of course, the system is not perfect, and it faces problems such as the ability of users to delete or reject tracking devices, or the issue of multiple users per machine and multiple machines per user. However, it is becoming a regular form of managing online advertising and, again, it does not require any form of audience measurement because it relies on routinely collected information that individual users generate automatically. Behavioral advertising targets specific users at specific moments in time.

Accounting for Audiences

The adaptation of audience measurement to the online environment appears to face significant challenges and seems to be on its way to losing the central institutional role it has played for decades. We could surmise that this situation does not significantly affect the process of accounting for audiences, which lies at the heart of audience studies. Because the evolution of audience measurement presented here has a direct impact on a specific subsection of communication media – online media – and seems to be relevant largely for commercial audience research, we might be led to conclude that audience studies in the academy can remain unaltered and linger at the margins of these developments. On the contrary, I would like to argue that the situation described here actually has a direct influence on the field of audience studies, and I will do so by addressing six related issues.

First, even if we were to argue that the dynamics examined above are circumscribed to the online environment, the increasing relevance of online communication should make us wonder about the actual reach of these phenomena. The dominant forms of public communication at particular times in history have often exerted a strong influence over the general understanding of audiences (Neuendorf, 2001). In this sense, the overwhelming dominance of television as the medium of reference throughout the second half of the twentieth century has clearly shaped our understanding of audiences, our ways of accounting for them. It would be naïve to think that television has lost that central place in society in general and in audience studies in particular. It would be equally naïve to think that this is going to be the case for very long. The rapidly growing relevance of online communication should lead us to consider the possibility that the specific dynamics of a subsection of the media in the current situation today might well transform tomorrow into the general overarching trends of media communication.

Second, even though the crisis underway in audience measurement seems to have its epicenter in the context of online communication, the truth is that its consequences have been felt across the board in the audience measurement industry. Although the history of commercial audience measurement includes frequent struggles over methodological procedures, market dominance, and the validity of measurement results, the general approach to measurement and its pivotal role in the economic functioning of commercial media has remained unquestioned. Currently, however, the problems that online audience measurement faces seem to be spilling over into all forms of measurement, and we are now witnessing a general questioning of audience measurement across all media (Turow, 2006). Considering the reverberations in other parts of the media world, it seems difficult to maintain that the issues discussed above only affect a small section of communication media. Rather, they are shaking the whole audience measurement industry.

Third, even if the two previous points were well taken, and we accepted that the situation examined here affects directly those forms of media communication that are on their way to becoming dominant and affects indirectly all other forms of commercial media communication, it still could be argued that this is not an audience studies' problem. After all, academic approaches to studying audiences have followed a different logic from commercial ones, have often used different methods, and have produced very distinctive views of the audience. However, we should consider Mosco and Kaye's (2000) assertion that “it was not until broadcasters needed to determine the actual size of the unknown audience that the analytical objectification of the audience began [. . .] eventually bequeathing to the field of mass communications an analytical category loaded with meaning” (p. 35). It is obvious that, once the concept entered the field of communication, it took on a new life, independent of the commercial interests from which it was derived. But, if the field of mass communication inherited the concept of “audience” from commercial research, and commercial research seems to be heading in the direction of discarding it as a useful way of understanding new media, we should at least wonder whether we want to keep using a concept when the field from which it originated does not seem to find a use for it any more.

Fourth, even though commercial and academic audience research have followed very different – often opposite – paths in terms of goals, methods, and results, ratings have a long tradition of being a source of data for academic audience research. Researchers of mass communication have explicitly relied on industry audience measurement sources to study media audiences, from the inception of radio (Beville, 1940) and television (Stavitsky, 2000) to the web (Webster & Lin, 2002). Given this fact, the influence of audience measurement data on academic research is not merely implicit and diffuse. To the extent that the nature of our knowledge of audiences is changing, the types of questions that can be answered using old conceptions of audiencehood will also change.

Fifth, if we examine the evolution of audience studies over the past few decades, we can see that academic research has moved in the direction of seeking a more nuanced view of the constitution of audiences and of the meanings of audience activity. Scholars have argued that the mere counting of audiences acts as a form of control that reduces the complexity and richness of those audiences to mere data on exposure. To the extent that commercial research moves beyond the idea of exposure and in the direction of a more detailed depiction of audiences' practices, perhaps there may be increasing possibilities for collaboration and mutual understanding. This is what Napoli (2011) refers to when he says that the evolution in the field of audience measurement may represent changes in the relevance of different forms of inquiry in audience studies “and, perhaps most interesting of all, an opportunity for bridging gulfs that have developed between academic traditions and between academic and industry researchers.” We need to acknowledge, though, that, in spite of this increasing possibility of convergence, the difference in goals and approaches will no doubt keep commercial and academic research in different camps, as contrasting and independent modes of studying and understanding audiences.

And sixth, over the last couple of decades, multiple traditions of research within the field of audience studies have been questioning the reality of the audience, and even the concept of the audience itself. Perhaps it was Allor (1988) who first explicitly stated this line of reasoning when he said that the “audience exists nowhere; it inhabits no real space, only positions within analytic discourses” (p. 228).²⁷ Thus, throughout the 1990s, a succession of authors began the discursive process of eroding the audience as an ahistorical abstraction or a naturalized concept. But before we decide on the demise of the audience as an analytic category, we should consider that the audience has had a long history of adaptation to new phenomena, an adaptation process that points to its resilience and flexibility. The concept of “audience” has changed from being applied to communicative situations in which there is a physical co-presence to other situations, in which technological mediation allows for communication at a distance; it has gone from referring to the act of hearing – whence its etymology (derivative of audire, “to listen” in Latin) – to encompassing the act of watching, or even reading (Radway, 1988). Thus, before we can herald the end of the audience as a useful academic or commercial heuristic, perhaps this discursive erosion and questioning needs to be paired with significant changes in the communication environment. The increasing relevance of digital and interactive forms of communication may actually offer the kind of change that encourages a productive engagement with the utility of the audience as a concept. While this history of conceptual adaptability might indicate that we are just witnessing a new expansion in the uses and meanings of the word “audience,” it seems necessary to remark that across all of its past incarnations the only element that has remained stable in the situations to which the word refers is the fact that there is a clear separation and asymmetry between senders and receivers – from popes and kings who confer the status of audience to lay people and subjects, to media organizations that produce content for the consumption of the mass audience. As Angus (1994) stated right before the Internet started to achieve its current relevance, in the most significant contemporary communication systems “audiences tend to remain simply audiences; that is, communication systems tend to sever audiences from reciprocal production of social knowledge and engagement in decision making” (p. 233). However, online forms of communication allow for the dissolution of that asymmetry and thus may break with the only trait of continuity in the different incarnations of our models of the audience. Online audiences often do not remain simply audiences. Still, the perfect symmetry and interchangeability between senders and receivers remains a counter-factual ideal in most communicative contexts, and there remains, for example, a clear difference between those who create and manage an online social network and those who take part it in as users. In the online context, asymmetry and power relations take a new, different form. However, the erosion of asymmetry in online forms of communication should at least make us wonder whether other concepts can better account for these new communicative realities. And it might well be that the general crisis that audience measurement is undergoing will serve as the final impulse needed for jettisoning our attachment to the concept of the audience. If that is the case, perhaps the same practices and the same modes of communication that the audience measurement industry seems unable to tame may serve as the source of inspiration for devising new ways of accounting for the relationship between people and their media practices of reading, watching, browsing, using, producing, and consuming.

ACKNOWLEDGMENT

I would like to thank the “paper working group” at the Berkman Center (in particular Hal Roberts, Wendy Seltzer, David Ardia, and Donnie Hao Dong) for their constructive comments on previous versions of this chapter. My residence at the Berkman Center was partly made possible by a José Castillejo grant awarded by the Spanish Ministry of Education.

NOTES

1 Perhaps it could even be included within what Herbert Blumer (1951) called “forms of collective behavior.” See, for instance, Long (1994) for a use of the concept of “collective action” to analyze the practice of reading.

2 Writing can be considered to be the first technological device inserted in the communication process. The printing press was the means through which the technology of writing achieved an unprecedented social relevance, and it is often taken as the starting point for the history of modern communication technologies.

3 As Livingstone (2003) put it: “With the shift away from visible and audible participation by live audiences, the activities of the new, mass mediated audiences became highly interiorized and hence inaccessible to the researcher” (p. 17).

4 This lack of connection among members of the audience is partially eliminated in the home celebration of media events, as Dayan and Katz (1992) have shown.

5 For a study of some of these rare instances of “audience self-expression” in the age of mass media, see Newman's (2004) work on radio audiences.

6 This is what Dahlgren (1998) meant when he said that “‘audiences’ are always at least in part discursive constructs, shaped by specific institutional needs and discursive domains” (p. 307). In the same vein, see Bratich (2005).

7 See Webster and Phalen (1994) for a discussion of the different images of the audience implied in communication policy, and Hagen (1999) for a view on the impact of audience measurement on journalists and media producers.

8 For a detailed history of audience measurement in broadcasting, see Beville (1988) and Buzzard (1990).

9 Space limitation does not allow for a proper historical and conceptual contextualization of audience measurement within the field of technologies of control. See Beniger (1986), in particular Chapter 8, for that contextualization.

10 It would be questionable, at the least, simply to say that it is the evolution of technology that determines the relevance and form of audience measurement. There are obvious and complex institutional forces at play in these processes – and their analysis widely exceeds the reach of this essay. It is possible, however, and even fruitful, to look at the evolution of audience measurement from the point of view of the change in the communication technologies whose use constitutes the audiences being measured. See, along these lines, Napoli (2001, 2003, and 2011).

11 Print periodicals can be considered, in this context, a hybrid mode. On the one hand, there is a physical container that can be sold in exchange for payment. On the other hand, print periodicals have relied on advertising as a significant source of revenue for most of their history. The hybrid business model of the press obviously points to the insufficiency of the technological argument to explain the degree of reliance on advertising, and thus the importance of audience measurement – no doubt economic, legal, social, and historical factors need to be taken into account as well. Non-broadcast television also offers an interesting point of comparison, since control of access to communicative products does not rely on restricting spaces or on selling containers, but rather on technological barriers raised between producers and audiences – scramblers, etc. – in a fashion similar to the ways in which pay models have been implemented online.

12 Again, a narrative more nuanced than the one being offered here would be needed to explain how advertising became the predominant business model for broadcasting. See, for instance, Smulyan's (1994) history of the commercialization of radio in the United States.

13 See Bermejo (2009) for an evolution of online advertising revenues.

14 ARPANET was a wide-area computer network created in the late 1960s with the support of the Advanced Research Project Agency of the United States Department of Defense. It is commonly considered the precursor of the current Internet.

15 Obviously, asymmetry is not a necessary trait of the World Wide Web. In fact, the original design of the web by Tim Berners-Lee allowed users to modify web pages easily. However, as Berners-Lee (2000) has explained, the spread of the web in its inception was accomplished at the expense of eliminating the original ability to edit the content displayed. The subsequent evolution of the web into what has been named Web 2.0 is in a sense a return – though incomplete and often highly controlled – to the original symmetry of the web's design.

16 For a detailed analysis of these three elements and of the birth of online audience measurement, see Bermejo (2007).

17 In fact the “online audience measurement industry” could as well be called the “web audience measurement industry,” since, among all the different online communication tools, it was the web that attracted the most advertising interest and on which measurement efforts would concentrate.

18 In other media, even though different methods may be used to measure their audiences, there is usually a method of reference, and also, usually, a measurement operation of reference per medium per country. This has not been the case for online audience measurement.

19 Actually, the first metering device ever used to measure the media audience was a radio audimeter (Beville, 1988). However, the increasing proliferation, miniaturization, and portability of radio receivers made it impossible to use metering devices for most of radio's history. In recent years, the advent of portable people meters has allowed for the reintroduction of audimeters as a way of measuring radio audiences.

20 Those that provide data for more than a few hundred websites use non-traditional sampling techniques in order to generate very large samples that need to be statistically manipulated to correct for bias.

21 Indirect inferences and complementary sources of data can aid in this endeavor. However, most records of online activity do not provide any direct demographic information that could be used to produce a precise profile of the audience being measured.

22 Tags are one of the tracking devices in use across the Internet. The first of these devices to appear were cookies, small textual files that are stored by websites in users' browsers in order to identify repeated visits by the same user/browser. Tags, in their different denominations and forms – web beacons, clear Graphic Interchange Formats (GIFs), JavaScript tags, etc. – serve the same function of identifying subsequent requests by the same browser, but they allow greater flexibility in terms of the range of information collected and they facilitate the automatization of the analysis of this information. Flash-cookies also belong in this family of tracking devices currently used in online communication.

23 In this sense, the current state of online audience measurement and the new possibilities of obtaining information about the audience are also affecting the ways in which journalists see their role – see, for instance, McGregor (2007) – and the ways in which news organizations manage their business – see Graves and Kelly (2010).

24 It is obvious that the final goal of advertising is the generation of sales. In that sense, advertisers cannot be satisfied with simply knowing that their ads have been seen, heard, or read by an audience with a specific size and composition. However, in the world of the mass media, this is precisely what advertisers pay for. The online environment seems to have raised the accountability bar, and advertisers seem to be increasingly dissatisfied with advertising campaigns that only guarantee a certain audience. They increasingly demand proof that the audience is moving toward purchases, and this explains the interest in clicks and the growing relevance of performance-based pricing models. See Hoffman and Novak (2003) for an examination of the different pricing models used in online advertising.

25 This centrality of ratings has lead Meehan (1984) to provide a third answer to the question “What is the commodity produced by advertising supported communication media?” To the “naïve” answer – programs – and Smythe's (1977) answer – audiences – she added a third one: ratings. For her, the commodity that is traded in broadcasting's advertising market is not audiences – since they do not go through any real manufacturing and sale process – but rather the information produced about audiences through a complex set of institutional, economic, and methodological arrangements: the ratings.

26 On top of this, and because of the effectiveness of the system and of the ability to keep track of users' reactions to the ads in the form of clicks, search engines have been able to offer advertisers performance pricing models that seem to be more attractive to them, since they provide proof that users are closer to becoming consumers.

27 This led Dahlgren (1998) to contrast the past with the present of audience studies: “In some slightly mythic recent past – of no more than a few decades ago – ‘audiences’ used to be a rather unproblematic notion. They may have been seen at times as difficult (and expensive) to research, but the concept and the empirical reality it referred to had a fairly common-sense quality. [. . .] The status – ontological and epistemological – of ‘audiences’ was generally not in itself a topic which engaged researchers. Today we find ourselves in a very different situation” (p. 298).

REFERENCES

Allor, M. (1988). Relocating the site of the audience. Critical Studies in Media Communication, 5(3), 217–233.

Ang, I. (1991). Desperately seeking the audience. London, UK: Routledge.

Angus, I. (1994). Democracy and the constitution of the audience: A comparative media theory perspective. In J. Cruz & J. Lewis (Eds.), Viewing, reading, listening: Audiences and cultural reception (pp. 233–252). Boulder, CO: Westview.

Beniger, J. (1986). The control revolution: Technological and economic origins of the information society. Cambridge, MA: Harvard University Press.

Bermejo, F. (2007). The Internet audience: Constitution and measurement. New York, NY: Peter Lang.

Bermejo, F. (2009). Audience manufacture in historical perspective: From broadcasting to Google. New Media and Society, 11(1/2), 133–154.

Berners-Lee, T. (2000). Weaving the web. New York, NY: Harper Collins.

Beville, H. M. (1940). The ABCs of radio audience. Public Opinion Quarterly, 4(2), 195–206.

Beville, H. M. (1988). Audience ratings: Radio, television, and cable. Hillsdale, NJ: Lawrence Erlbaum Associates.

Blumer, H. (1951). Collective behavior. In A. M. Lee (Ed.), Principles of sociology (pp. 67–121). New York, NY: Barnes & Noble.

Bratich, J. Z. (2005). Amassing the multitude: Revisiting early audience studies. Communication Theory, 15(3), 242–265.

Butsch, R. (2000). The making of American audiences: From stage to television, 1750–1990. Cambridge, UK: Cambridge University Press.

Buzzard, K. (1990). Chains of gold: Marketing the ratings and rating the markets. Metuchen, NJ: Scarecrow.

Buzzard, K. (2003). Net ratings: Defining a new medium by the old, measuring Internet audiences. In A. Everett and J. T. Caldwell (Eds.), New media: Theories and practices of digitextuality (pp. 197–208). New York, NY: Routledge.

Dahlgren, P. (1998). Critique: Elusive audiences. In R. Dickinson, R. Harindranath, & O. Linné (Eds.), Approaches to audiences: A reader (pp. 298–310). London, UK: Arnold.

Dayan, D., & Katz, E. (1992). Media events: The live broadcasting of history. Cambridge, MA: Harvard University Press.

Graves, L., & Kelly, J. (2010). Confusion online: Faulty metrics and the future of digital journalism. New York, NY: Tow Center, Columbia University. Retrieved October 1, 2011, from http://www.journalism.columbia.edu/page/633-confusion-online-faulty-metrics-the-future-of-digital-journalism/437

Hagen, I. (1999). Slaves of the ratings tyranny? Media images of the audience. In P. Alasuutari (Ed.), Rethinking the media audience (pp. 130–150). London, UK: Sage.

Hoffman, D. L., & Novak, T. P. (2003). Advertising pricing models for the world wide web. In B. Kahin & H. R. Varian (Eds.), Internet publishing and beyond: The economics of digital information and intellectual property (pp. 45–61). Cambridge, MA: MIT Press.

Livingstone, S. (2003). The changing nature of audiences: From the mass audience to the interactive media user. In A. Valdivia (Ed.), A companion to media studies (pp. 337–359). Malden, MA: Blackwell.

Long, E. (1994). Textual interpretation as collective action. In J. Cruz & J. Lewis (Eds.), Viewing, reading, listening: Audiences and cultural reception (pp. 180–211). Boulder, CO: Westview.

McGregor, P. (2007). Tracking the online audience. Journalism Studies, 8(2), 280–298.

McQuail, D. (1997). Audience analysis. Thousand Oaks, CA: Sage.

Meehan, E. R. (1984). Ratings and the institutional approach: A third answer to the commodity question. Critical Studies in Mass Communication, 1(2), 216–225.

Miller, P. (1994). Made-to-order and standardized audiences: Forms of reality inaudience measurement. In J. Ettema & C. Whitney (Eds.), Audiencemaking: How the media create the audience (pp. 57–74). Thousand Oaks, CA: Sage.

Mosco, V., & Kaye, L. (2000). Questioning the concept of the audience. In I. Hagen & J. Wasko (Eds.), Consuming audiences? Production and reception in media research (pp. 31–46). Cresskill, NJ: Hampton.

Napoli, P. (2001). The audience product and the new media environment: Implications for the economics of media industries. International Journal on Media Management, 3(2), 66–73.

Napoli, P. (2003). Audience economics. Media institutions and the audience marketplace. New York, NY: Columbia University Press.

Napoli, P. (2011). Audience evolution: New technologies and the transformation of media audiences. New York, NY: Columbia University Press.

Neuendorf, K. (2001). Viewing alone? Recent considerations of media audience studies. Journal of Broadcasting and Electronic Media, 45(2), 345–354.

Newman, K. M. (2004). Radio active: Advertising and consumer activism 1935–1947. Berkeley, CA: University of California Press.

Smulyan, S. (1994). Selling radio: The commercialization of American broadcasting 1920–1934. Washington, DC: Smithsonian Institution.

Smythe, D. W. (1977). Communications: Blindspot of Western Marxism. Canadian Journal of Political and Social Theory, 1(3), 1–27.

Stavitsky, A. G. (2000). By the numbers: The use of ratings data in academic research. Journal of Broadcasting and Electronic Media, 44(3), 535–539.

Radway, J. (1988). Reception study: Ethnography and the problems of dispersed audiences and nomadic subjects. Cultural Studies, 2(3), 359–376.

Roscoe, T. (1999). The construction of the world wide web audience. Media, Culture and Society, 21(5), 673–684.

Turow, J. (1997). Breaking-up America: Advertisers and the new media world. Chicago, IL: University of Chicago Press.

Turow, J. (2006). Niche envy: Marketing discrimination in the digital age. Cambridge, MA: MIT Press.

Webster, J. G., & Lin, S. (2002). The Internet audience: Web use as mass behavior. Journal of Broadcasting and Electronic Media, 46(1), 1–12.

Webster, J. G., & Phalen, P. (1994). Victim, consumer, or commodity: Audience models in communication policy. In J. Ettema & C. Whitney (Eds.), Audiencemaking: How the media create the audience (pp. 19–37). Thousand Oaks, CA: Sage.

Webster, J. G., Phalen, P. F., & Lichty, L. W. (2000). Ratings analysis: The theory and practice of audience research. Mahwah, NJ: Lawrence Erlbaum Associates.