5. Surveillance Advertising Takes Shape

Protoplatformization

The internet advertising industry awakened to this challenge amid a convulsing dot-com stock market. Soaring company valuations driven by record levels of speculative investment were punctuated by sudden downturns and spates of postponed IPOs. This volatility contributed to a growing recognition among business leaders that the dot-com phenomenon was indeed a financial bubble. It was, as Netscape’s Jim Clark put it, “too hot not to cool down.”12 Although ad networks such as DoubleClick continued to grow at a rapid clip, a large portion of the demand for their services still came from dot-com start-ups eager to drive traffic to their online storefronts and portals. Despite the safety net provided by speculative capital, even the biggest ad networks understood that they needed to court national marketers in order to prosper in the long run. Failure to do so could spell disaster, but the potential upside was enormous. The combined ad-buying power of the FAST alliance alone was an estimated $50 billion a year, more than twenty-five times the total spent on internet advertising in 1998.13 As one ad industry representative reportedly told a group of marketers at the FAST Summit: “Just tell us what you want us to do and we’ll do it.”14

During this period, leading ad networks like DoubleClick and CMGI began to reconfigure their businesses around two overlapping objectives. The first, outlined in chapter 4, was to get big fast in order to squeeze out competition and secure a dominant market position in the web advertising sector. This involved using finance capital to buy competitors and invest in technology and labor. The second objective was to transition from a business model that primarily served web publishers (the supply side of the advertising market) to one that also provided services to marketers directly (the demand side). Up to this point, much of the industry had focused on helping publishers sell ads, leaving traditional ad agencies to help marketers buy them. Now, in their expansionary fervor, major ad networks attempted to facilitate both supply and demand, rolling out new products and services geared toward marketers and their ad agency proxies.

These efforts dovetailed to form a business strategy that prefigured what Nick Srnicek calls “platform monopoly.”15 In a platform monopoly, the goal is not just to successfully compete within a market but to facilitate the market itself as an essential intermediary. Successful platforms sit at the center of “multi-sided markets,” connecting “distinct groups of customers who value each other’s participation.”16 Using dot-com financial capital, ad networks began to create services not only for web publishers who got paid to host ads on their sites, but also for the ad agencies, marketers, and retailers who bought the ads and increasingly demanded proof of their results. Crucially, ad networks ramped up their promises to provide marketers with enhanced ROI. In a move that anticipated the platform monopolies of Google and Facebook, ad networks began a process of what might be called protoplatformization, constructing a sociotechnical infrastructure of business relationships and technical capacities to facilitate the capture and exchange of consumer information at an accelerating scale and pace.

DoubleClick was among the first to reorganize its business in this way. Although ad network executives did not use the term “platform monopoly,” they described their objectives clearly enough. DoubleClick’s CEO Kevin O’Connor framed the strategy in terms of “reintermediation.”17 O’Connor, who was something of an iconoclast, coined the term to invert the new economy rhetoric of “disintermediation,” which promised to connect buyers and sellers directly on the internet, eliminating inefficient go-betweens. By contrast, reintermediation signaled DoubleClick’s intention to leverage its market position and superior ad-targeting capacity to become web advertising’s indispensable middleman. The objective was not only to “deliver every ad to every consumer” but also to broker the broadest possible range of advertising transactions, taking a revenue slice of each one.18 DoubleClick was attempting to expand from providing a narrow band of outsourced ad services to web publishers to “building the infrastructure that makes marketing work in the digital world.”19

The gold standard for the platform monopoly approach was Microsoft’s unrivaled dominance in computer operating systems, which gave it inordinate power in the adjacent computing hardware and software sectors—and made it among the world’s most valuable companies. Many dot-coms including Netscape, Yahoo, and DoubleClick sought to mimic Microsoft’s playbook, even as the company underwent investigation by the U.S. Department of Justice for its anticompetitive practices. Making the patterning explicit, O’Conner described DoubleClick’s intention to become the “internal operating system for advertising on the net.”20

Protoplatformization meant that DoubleClick had to move beyond simply scaling up its existing ad network. Portals such as Yahoo and America Online had grown to the point where they too could put banner ads in front of large audiences. To differentiate itself, DoubleClick needed to create new markets to inhabit, generate sector-leading knowledge about consumer behavior, and develop better means to quantify ROI for marketers. This required not only attracting new marketer and publisher clients but also bringing those clients more squarely into the routine practice of consumer data collection and exchange. In other words, DoubleClick and its rivals sought to develop a shared capacity for internet surveillance among all entities of the marketing complex and to place themselves at the center of the burgeoning consumer data economy.

This platform monopoly approach was controversial to some in the broader advertising sector, where double dipping from both the supply and demand sides of the market was seen as a conflict of interest. But the growing market power of dot-coms like DoubleClick, coupled with the sense that old rules no longer applied in the new economy, enabled ad networks to chart this new terrain relatively unfettered by the norms of Madison Avenue. Engaging in a bit of revisionist history, the CEO of CMGI’s Engage now exclaimed: “We’ve always thought the marketer is the one with the power, so we are innovating for them.”21

The remainder of this chapter chronicles ad networks’ efforts to become platform monopolies and outlines the broader implications of the late dot-com period (1998–2000) for surveillance advertising’s development. The trend of protoplatformization expanded surveillance capacities and created strong market pressures for all commercial entities to participate in consumer data collection. It also opened the doors to increasingly invasive, manipulative, and discriminatory practices of behavioral profiling and ad targeting. In all of these ways, these early attempts to build platform monopolies prefigure important elements of today’s surveillance advertising economy.

Open the Network, Close the Loop

One of the first prerequisites for platformization was breaking open the business model that helped launch ad networks to prominence in the preceding years. DoubleClick had made its bones by building the web’s premier ad network, but now the exclusivity of the network model threatened to constrain the company’s growth. Ad networks had relied on making ad targeting a premium service, available only to web publishers who signed on to exclusive partnerships. If you wanted to use DoubleClick’s state-of-the art technology, you had to join its network and refrain from using others. This afforded DoubleClick leverage to charge steep prices (publishers often relinquished half of all ad revenue), but it restricted DoubleClick’s ability to serve ads and collect data from nonaffiliated sites. Though DoubleClick garnered significant audience traffic across its stable of publishers, the company’s activities were nonetheless confined by the network format. Wedded to exclusive partnerships, ad networks were ceding significant chunks of the market without so much as a fight.

DoubleClick was among the first to crack open the ad network model by relaxing exclusive partnership requirements and offering its DART ad-serving technology to web publishers as a stand-alone service. Selling its technology à la carte risked undermining its network business. O’Connor said the decision “almost ripped the company apart.”22 But to win the business of national marketers, the company needed to develop new products and move into new markets. The move proved successful, attracting hundreds of web publishers who wanted better ad-targeting capabilities but had balked at joining an ad network. DoubleClick launched in 1996 with approximately thirty participating web publishers.23 By the end of 1998, DoubleClick’s client roster had grown to 570 publishers representing thousands of websites, including premium sites like the Wall Street Journal, NBC, and CBS.24

Competing ad networks began to relax their terms as well, experimenting with nonexclusive contracts and systems for bartering unsold ad inventory. CMGI broadened the purview of its Engage advertising division to include clients outside of the holding company’s portfolio.25 Engage’s Real-Time Visitor Intelligence service gave any web publisher the ability to tap into its proprietary profile database in order to deliver targeted ads and personalized offers.26 No exclusive network contract was required. As promoted on its product page: “As soon as a visitor clicks on your website, you can use information about their previous internet activities to provide them with customized web pages, targeted ads, promotions, products and services—even if it’s their first visit to your site.”27 So-called discount networks pitched low costs and ease of use to bring in as many publishers as possible. One such company, LinkExchange, compiled a massive network of more than 400,000 small-traffic websites.28

DoubleClick also created a system for swapping unsold, or remnant, ad inventory among publishers on a one-to-one exchange basis.29 In the event that DoubleClick could not deliver a targeted ad—for example, if no appropriate consumer profile could be located—it would simply serve a default banner representing itself or one of its affiliates. In O’Connor’s words, the goal of reintermediation was to ensure that there was “never an unused ad space.”30 Likewise, the CMGI property FlyCast developed a system to auction off remnant inventory at reduced prices across a network of a thousand sites that included premier publishers such as Yahoo.31 Other companies in the remnant auction market included Adauction and Adbot.32 As explained by Adauction’s CEO:

The online auction works for publishers because it complements existing sales programs without creating channel conflict. Unlike other outlets that essentially replace the publisher’s direct sales of advertising, the auction is similar to the inventory yield and management system that airlines use to sell off empty seats to brokers, travel agents and other companies. Ad availability and media are as perishable as an airplane reservation. The auction format helps efficiently sell advertising late in the sales cycle or, to complete the metaphor, just before the plane takes off.33

Expanding their range of publisher partners was only the beginning of ad networks’ turn toward platformization. To attract marketer clients, ad networks needed to demonstrate that internet advertising could provide strong ROI. Targeting a potential customer with an ad was all well and good, but demonstrating a resulting action—say, a product purchase or newsletter signup—was far better. To this end, DoubleClick created a marketer-focused version of its DART ad-targeting technology, running a variety of programs under the heading “Closed Loop Marketing Solutions,” which purported to show the “true relationship between advertising and sales.”34 The idea was to augment targeted advertising with so-called post-click analysis to better understand what users did (and did not do) after they clicked through banners to reach marketers’ home pages and digital storefronts.

One of the most successful applications of post-click analysis was what is now known as remarketing or retargeting, a practice that should be familiar to anyone who has experienced an ad seemingly following them around the web. Remarketing enables advertisers to reconnect with users who previously clicked their ads or visited their sites by sending a follow up message, either through a customized ad or email. In 1999 DoubleClick introduced Boomerang, one of the web’s first remarketing applications. The service was pitched as a way for marketers to reach “precisely the individuals who have displayed an explicit interest” in their products by delivering follow-up ads across any of DoubleClick’s thousands of partner sites.35 Ads could be targeted using a range of variables, giving marketers the capability to reach prospects who had previously visited specific pages on their sites, including those who visited but had not completed a desirable transaction like a purchase or email registration.

With remarketing, advertisers could deploy tailored ads with the capacity to follow specific users as they traversed the web. It was the online equivalent of a mattress salesperson slipping into the backseat of your car as you left the showroom, quietly riding along with you to your next destination, and then tapping your shoulder to remind you of the upcoming President’s Day sale extravaganza. Failing that, the salesperson might also approach others who happen to share your behavioral or demographic traits, under the assumption that they too might be in the market for a queen-size pillow-top mattress.

In order to work, remarketing and similar services required that new channels of communication be established among web publishers (where ads appeared), marketers (where consumer actions occurred), and ad networks (who coordinated and conducted the transactions). For example, DoubleClick placed invisible tracking codes on its marketer clients’ sites in order to enable post-click functionality. More advanced services required marketers to give DoubleClick access to the guts of their online shopping applications and customer databases, effectively building a patchwork infrastructure for consumer data transmission among advertising partners with DoubleClick as the hub.

Closing the loop between ads and consumer actions was hardly a revolutionary advertising strategy in itself. This had long been the premise of direct response marketing like mail-in coupons and catalogs. Similarly, ad networks had been developing centralized surveillance systems since 1996. As described in chapter 3, the first generation of ad networks combined ad delivery and data collection into a unified process, enabling a kind of localized surveillance within the boundaries of a given ad network. But subsequent efforts to close the loop marked a new stage of internet advertising’s development because they extended surveillance capacities beyond the websites where ads appeared to marketers’ own home pages and storefronts, and eventually beyond the web itself.

Racing to achieve platform monopoly status, ad networks also sought to build bridges with the well-established world of off-line data brokering in order to enhance their consumer databases. This relied on extensive partnerships among all manner of data brokers, both new and old. By 1998, CMGI’s AdKnowledge media buying service integrated an array of commercial data, including “audience demographics from [online ratings services] MediaMetrix and NetRatings; psychographic data from SRI Consulting; web site ratings and descriptions from NetGuide; and web traffic audit data from BPA Interactive.”36 Ad networks had used external data in limited ways for several years. DoubleClick was among the first companies to augment ad targeting with information obtained from the U.S. census and other public sources. Nevertheless, advertising in cyberspace remained largely separate from the marketing and consumption practices of the physical world. Consumer profiles were generally limited to information that was collected passively as people surfed the web. Profiles were linked to unique IDs but only contained personally identifying information when it was supplied actively—say, as part of a website registration or purchase.

These boundaries began to break down when CMGI’s AdForce secured an exclusive partnership with the global information services conglomerate Experian to provide data for ad-targeting purposes.37 The off-line data broker industry itself was undergoing an expansionary trend as companies like Experian began to transition from primarily providing financial information services to a much broader range of activity, including credit reporting and consumer profiling.38 Ad networks were particularly keen to obtain consumers’ off-line purchasing records, thought by many marketers to be a valuable predictor of future buying behavior. This is certainly what DoubleClick was after in its 1999 acquisition of Abacus Direct, a data broker that tracked the buying habits of some 88 million U.S. households.39 “Advertisers rely on demographic information to target ads when they don’t have anything else,” said DoubleClick’s Kevin Ryan, talking up the significance of the acquisition. “Transaction information is much better.”40 DoubleClick’s move to integrate its profile databases with Abacus’ records, which included personally identifying information such as names, addresses, and telephone numbers, caused a public relations debacle and enflamed a larger public policy confrontation. I explore this incident in more detail in chapter 6. For now, it is sufficient to note that although DoubleClick’s intention to integrate Abacus’ data with its own was the most prominent example of the increasing intrusiveness of surveillance advertising, it was by no means an outlier.

The Social Relations of Surveillance Advertising

One of the most troubling consequences of surveillance advertising’s expansion in the late 1990s was the increased use of behavioral profiling to categorize people according to determinations of their social worth. In 1993, Oscar Gandy wrote a prescient book called The Panoptic Sort, which articulated a theory of information technology–based surveillance through which individuals and groups are “sorted according to their presumed economic or political value.”41 Looking at a range of examples, including credit reports and data from the U.S. Census Bureau, Gandy warned of an emerging system of classification that enabled “organized interests,” whether selling shoes, insurance, or political ideology, “to identify, isolate, and communicate differentially with individuals in order to increase their influence over how consumers make selections among these options.”42

One of Gandy’s overarching concerns was that information technology–based sorting accelerates forms of discrimination in accordance with institutionalized biases of “race, gender, age, class, culture, and consciousness.”43 Surveillance advertising’s proliferation in the dot-com period only moved such practices of “social sorting” further toward the front of the marketing complex’s business agenda.44 Joseph Turow has documented the progression of this trend whereby marketers increasingly use profile databases “to determine whether to consider particular Americans to be ‘targets’ or ‘waste.’”45

Although the practice of tailoring ad messages to specific audiences is not new, online behavioral profiling is of a qualitatively different character than the probability-based methods used to analyze and target mass media audiences. For instance, consumer classification is greatly enhanced by surveillance practices that combine past purchasing records with online behavioral data and demographic information. The shift is significant. As Turow explains: “For decades, marketing and media firms learned as much as they could about social groups and then tried to target people they thought were members of these groups. The emerging process is almost the opposite: They learn enormous amounts about individuals, consign them to various groups, and then determine whether and how they want to deal with them.”46

Aggregating data across the publisher/marketer divide put companies like DoubleClick in a unique position to experiment with intensified forms of behavioral profiling. In some instances, behavioral profiling meant collecting novel types of data, such as attempts to monitor the ads that users hovered over with their computer mouse but did not click. How long did they linger over the ad? What was clicked immediately after hovering? But more significant was the buildup of capacities to aggregate data from an expanding array of sources in order to develop more detailed browsing histories, analyze post-click behaviors, classify consumers in various ways, and otherwise pursue what Phillip Napoli calls “the rationalization of audience understanding.”47

Consumer classification became an increasingly important component of the targeting algorithms that ad networks touted as the key to improve marketer ROI. CMGI’s Engage reportedly created consumer interest profiles using a matrix of some 800 potential attributes.48 With the help of the Internet Archive’s Wayback Machine, it is possible to see a snapshot of Engage’s 1998 product description for a service called Next-Generation Profiling:

As a visitor browses through an Engage-enabled website, the Engage Suite software builds individual profiles based on the type of content viewed, the time spent viewing, and other factors including their frequency and recency of visits to a particular interest category. This information is processed by a patent-pending algorithm into profiles that include a user identification number, an interest category code, and an interest score to indicate a level of interest in a particular category. A single visitor can have several entries in the database—one for each observed interest category. As the number of recorded visits of a single visitor grows, the accuracy and depth of the profile is improved. Unlike static registration information, Engage profiles are constantly changing to more accurately reflect the current interests of an individual.49

It is one thing when extensive profiles, compiled from disparate off-line and online sources over which individuals have little knowledge or control, become the basis for the commercial messages they encounter on the internet. It is perhaps another when these processes dictate the availability of broader social opportunities and material necessities. As Lori Andrews asks: “When young people in poor neighborhoods are bombarded with advertisements for trade schools, will they be more likely than others their age to forgo college? And when women are shown articles about celebrities rather than stock market trends, will they be less likely to develop financial savvy?”50 Assessing these outcomes, Joseph Turow laments the dissolution of a media landscape that, while flawed, offered a pluralistic balance of society-making mass media and segment-making niche media.51 But even these critiques do not capture the full social costs of unrestrained consumer surveillance.

As early as 1999, the data broker Acxiom offered a service called InfoBase Ethnicity System that provided a “precise breakdown of ethnic, religious, and minority classifications” that could be matched with name, income, housing information, and other demographic data.52 According to promotional materials, the service enabled marketers to target, or perhaps exclude, categories of consumers as specific as “full-figured African American women.” By 2000, critics had coined the term “weblining” to describe the practice of denying people opportunities on the internet based on their marketing profile.53 The concept derives from “redlining,” a descriptor of earlier discriminatory practices whereby the boundaries of poor, often minoritized, neighborhoods were mapped in red to indicate that services such as banking or telecommunications need not be offered. Redlining gets to the root of the issue, which is a hardening of social inequalities, whereby surveillance advertising practices reproduce social discrimination and economic exploitation on the internet and beyond. In this scenario, the marketing complex is doing more than just sorting profiles according to different consumer typologies. It is also creating a system that, when functioning optimally, catalogs individuals as either valuable or worthless via a range of processes ultimately beholden to profit maximization.

Far from creating objective representations of the world as it is, consumer surveillance reproduces and often reinforces existing structural inequalities.54 In 1993, Gandy illustrated the prejudicial capacity of the panoptic sort by recounting the systematic denial of credit to qualified applicants of color based on a discriminatory index of race embedded in preweb automated loan processing systems. As Tamara Shepherd argues, data-driven surveillance marketing “does not just reflect user desires, it produces them in ways that are differential according to already existing structures of privilege.”55 In other words, discrimination is most likely to be exercised against the interests of working-class people, people of color, women, and other groups who might be variously ascertained as “waste” by an inhumane surveillance advertising apparatus.56

Mass marketing was in many cases about selling as much as possible to as many as possible. Although this approach certainly continues, surveillance advertising was developed in no small part to mitigate the dissolution of the mass audience and the inefficiencies of mass appeals. Surveillance advertising is about boosting profits by selling more to the right people, the flip side of which is eliminating efforts to market to consumers deemed undesirable, and even pruning existing customers who fail to meet profitability standards. In a purely economic sense, these practices are justifiable if they produce efficiency gains that compensate for the loss of low-value customers. Wherever exclusion can be implemented more profitably than inclusion, we should expect increasingly stratified and discriminatory business practices within an increasingly stratified and discriminatory capitalism. In the United States today, there are few areas of social life that remain insulated from this logic. Health care, employment, housing, education—all are seemingly fair game.

Mark Andrejevic’s concept of “digital enclosure,” described as the ongoing construction of interactive spaces “where every action generates information about itself,” provides another useful lens to make sense of the social relationships produced by surveillance advertising and the industry’s move toward platformization.57 In digital enclosure, the persistent monitoring of interactive spaces enables the production of information commodities (e.g., behavioral profiles), which become the property of those conducting the surveillance and over which the subjects of monitoring have little control. As digital enclosure has proliferated, the ironic outcome for those being watched is an “asymmetrical loss of privacy.” Individuals are more scrutinized than ever, while the institutions conducting surveillance remain “stubbornly opaque.”58

The trajectory of protoplatformization in the late 1990s was a key moment of digital enclosure on the internet. It began to normalize a default mode of media engagement whereby participation in interactive spaces required submission to surveillance, a reversal of interactivity’s much-lauded emancipatory potential. The theory of enclosure is powerful because it highlights the deep connections between capital accumulation and social division. Its reference point is a Marxist understanding of the land enclosure movements associated with the historical transition from feudalism to capitalism. In addition to creating new propertied areas for the production of commodities, land enclosures created social divisions among those who owned and controlled land and those who did not. Whereas land enclosures formalized social cleavages around property ownership and control over the means of production, digital enclosures create class-like divisions around the ownership and control of interactive spaces and the information commodities produced therein.

Whether the object of enclosure is common land or behavioral information, the purpose is to wring profits from yet untapped social domains, to expand the sphere of commodification.59 Beyond the business model of any given internet advertising platform, digital enclosure is an extension of the marketing complex’s long-standing prime objective to generate high levels of consumption to match the output of an incredibly productive consumer capitalism. Ultimately, surveillance advertising as digital enclosure makes media engagement more productive for capital by creating new opportunities for information commodification, centralizing control over interactive spaces, and making it increasingly difficult for people to abstain from consumer surveillance.

Here Come the Marketers

Marketers began to adopt the web as an advertising channel in earnest during the later years of the dot-com bubble. Data compiled by Advertising Age showed that in 1998, just nineteen of the largest hundred national marketers advertised on the web. By 1999, that number had jumped to eighty-seven.60 Between 1999 and 2000, the total number of marketers purchasing online advertising quadrupled.61 Not only did more marketers come online, they also began to allocate proportionally more of their ad budgets to internet ads. A member survey by the Association of National Advertisers found that respondents had tripled their web ad spending from 1998 to 1999.62 The total number of ad impressions purchased by the top ten traditional marketers increased by a factor of five from 1999 to 2000, while aggregate online ad spending more than doubled from 1998 to 1999 and nearly did so again from 1999 to 2000.63

As ad networks focused their services around providing marketer ROI, they outflanked and then won business from some of their biggest detractors: traditional advertising agencies. Although some agencies had been skeptical of upstarts like DoubleClick, they increasingly turned to “outsiders possessing technological muscle” to conduct web campaigns.64 Of course, agencies partnered with ad networks largely in response to marketers’ requests for “proof that their online campaigns [were] paying off.”65 Those that did not have the ability to meet these expectations were forced to contract outside help. A significant chunk of this business went to the protoplatforms, further contributing to the sector’s consolidation. DoubleClick attracted business from big-name marketers including AT&T, Charles Schwab, IBM, Intel, and Microsoft.66 By 2000, DoubleClick alone served some 4,400 marketer clients representing over half of Fortune 100 companies.67

Though DoubleClick served hundreds of billions of targeted banners, it bears repeating that banner ad impressions are just one measure of surveillance advertising’s development. The platform monopoly strategy was an attempt to dominate online advertising by reintermediating every possible advertising transaction, regardless of any specific ad format or technology. Responding to an interviewer’s suggestion that DoubleClick was simply “peddling banners,” DoubleClick’s Kevin O’Connor articulated the agnosticism of the platform strategy as such: “You’re fixating on a single piece of real estate that happens to be shaped like a banner. Our business is, we take bytes. . . . whether they’re shaped like a banner or a pop-up or a sponsorship. . . . It could be a video, it doesn’t matter. We take those bytes and we target them at consumers, and we watch how consumers react.”68

O’Connor was right. DoubleClick and the other platform hopefuls were not just peddling banners. They were selling surveillance advertising as a plug-and-play service, facilitating their clients’ participation in a widening dragnet of consumer data collection. As ad networks embraced platformization, options for linking up with centralized commercial monitoring systems proliferated. Publishers could join any number of ad networks, license ad-server technology, or use cloud-based applications to collect and combine data on web users. Marketers could use dedicated ad campaign management platforms and customer relations management systems to conduct new kinds of targeting such as remarketing. As O’Connor put it: “You can think of us as the Burger King of internet advertising. We give it any way you want.”69 Although perhaps not as easy as pulling up to the drive-through, the barriers for marketers and publishers to participate in surveillance advertising were being lowered dramatically.

As it became easier to participate in consumer surveillance, it also became more difficult to abstain. This was particularly true for web publishers, who became increasingly desperate to attract advertising support as alternative funding models like paid-subscriptions faltered and dot-com investment dried up. “If the money is right, many online publishers are willing to strike whatever sort of partnerships an advertiser might want,” reported the Wall Street Journal.70 “I don’t see how any site cannot [collect user data]. Advertisers will demand it,” said one marketing executive.71 By 2000, the Wall Street Journal surmised that it was “too late to tinker” with tracking technologies such as cookies: “They power practically every online shopping cart and every paid advertisement. In less than six years, they’ve become part of the fabric of the web.”72 Corroborating the Journal’s claims, a series of studies by the Federal Trade Commission found the consumer surveillance was “nearly ubiquitous” online.73 The FTC predicted “a 99% chance that, during a one-month period, a consumer surfing the busiest sites on the web will visit a site that collects personal identifying information.”74

Throughout the late dot-com period, web advertising remained a work in progress. Marketers continued to press for strategic advantage and made hay of instances of targeting gone wrong—flogging home mortgages on sites aimed at teenagers, for example—that undermined the rhetoric of precision put forth by the surveillance advertising industry.75 Well aware of the importance of positive publicity and the appearance of forward momentum for attracting clients and investors, online ad executives routinely oversold their capabilities. So there was a heavy dose of irony in statements like those by Paul Schaut, CEO of Engage, who told Adweek in 1999: “Now, we don’t have to wave our arms to talk about the promise of profiling. We have arrived.”76 Ad networks had indeed greatly expanded their surveillance capacities, but the industry had not “arrived” in the sense of fulfilling any “promise of profiling.” Claiming that the waving of arms was no longer necessary was in fact a vigorous display of arm waving.

It might be expected that the gap between the rhetoric and reality of surveillance advertising’s execution would have curtailed marketers’ adoption. National marketers in particular had been complaining about the web’s shortcomings from the very beginning, but ignoring the internet was never actually an option. As Dan Schiller argues, confronted by the internet’s exploding popularity and the threat of its noncommercial development, marketers “had little choice but to spring into action.”77 The example of Procter & Gamble is again illustrative. In the early 1990s, P&G was among the first to articulate the marketing complex’s collective need to ensure that the new interactive media would serve the interests of advertisers. The company threw its early support behind the cable industry’s “interactive television” projects, all of which withered as the web rose to prominence. Thereafter P&G became one of web advertising’s most reliable complainers, convening the FAST Summit and trumpeting the medium’s shortcomings regarding ROI. In 1998, P&G spent about one tenth of 1 percent of its U.S. ad budget on the internet.78 “The current state of web advertising just isn’t effective enough to warrant any truly meaningful investment from us,” said Denis Beausejour, P&G’s vice president of advertising.79

But even the world’s biggest advertiser could not afford to sit out the internet. The coercive power of capitalist competition is inescapable. Between 1998 and 2000, P&G ramped up investment in its own corporate sites and quadrupled its online advertising budget. Lauding the company’s “tremendous influence in leading the industry, along with its own real commitment to the medium in ad spending and creation of innovative, successful campaigns,” Advertising Age named P&G its 1999 “interactive marketer of the year.”80 Beausejour, the same executive who had lamented the web’s inefficiency a year before, explained the reversal in these terms: “The interactive world is where consumers are going, where it’s happening, and we need to be there.”81

Beausejour was speaking for more than just P&G. Ultimately, all elements of the marketing complex—including the upstart ad networks—shared an overriding interest in maximizing the internet’s utility for selling. As Advertising Age observed: “While industry leaders may disagree about the mechanics of developing internet ads . . . the most important aspect to come out of FAST is that parties who used to sit at opposite sides of the table are now working together to figure out how to make the internet work as an ad medium.”82 The push for increasingly rationalized forms of surveillance advertising was part of a broader effort by national marketers to mold the internet’s development to suit their interests, beginning with the trade association lobbying that I outlined in the first chapters of this book.

The flashy dot-coms with billion-dollar valuations and the marketing titans with billion-dollar ad budgets often butted heads, but they quickly converged on the notion that consumer data was to be the lifeblood of the commercial internet. Although the particulars were implemented on a competitive and ad hoc basis, a surveillance advertising consensus had taken hold. Closed-loop advertising services like remarketing laid bare the structural composition of this arrangement, demonstrating that ROI required consumer monitoring conducted collaboratively by web publishers, marketers, and platform intermediaries. Simply collecting data on one’s own customers and keeping that data in house would not be sufficient. Surveillance advertising would be, in key respects, a team effort.

DoubleClick executives came to argue that advertising’s future hinged on new forms of cooperative business practices around the collection, analysis, and deployment of marketing data:

It has become clear that the old business silos of direct marketing versus brand advertising, or offline versus online media development, or the marketing area versus the IT department, no longer make sense. The common trait possessed by the most effective marketers and advertisers is that they have broken down these walls and allowed the groups to learn from each other’s experiences. Successful companies must aggregate customer data from various channels in order to have a holistic view of their customers. They need the ability to measure the effectiveness of their campaigns across these various media. They also need technology to integrate their sales, customer service, IT, marketing, and advertising efforts.83

Consumer surveillance became a focal point in the perpetual business struggle to maximize returns, which demanded breaking down the walls that separated marketers and their intermediaries, from ad agencies to publishers to online advertising companies. The protoplatformization of ad networks epitomized these developments. DoubleClick’s NetGravity subsidiary worked not only with publishers but also with marketers and retailers, and its services became increasingly integrated with its clients’ own growing customer databases. “Is the business consulting or hardware?” asked Advertising Age. “Once the server is installed, there’s a lot of time, resources and intellectual property that gets shared among clients, agencies and third-party ad servers.”84 Institutional and operational distinctions were blurred to facilitate surveillance advertising and assert the marketing function online. “All this means vendors such as NetGravity will be working even more closely with advertisers, agencies and publishers in the future,” said a company executive. “It’s not a piecemeal approach.”85

As more marketers began to participate in surveillance advertising, ad networks benefited from an enlarged capacity to aggregate data and the incremental entrenchment of their position as the most important strategic intermediaries—protoplatform monopolies—in the online advertising sector. This made the fortunes of market leaders like DoubleClick even as it structured the entire industry in important ways, tilting internet development toward ubiquitous commercial surveillance and increasingly discriminatory practices of behavioral profiling. Protoplatformization pointed to a future where a few big winners would dominate an internet advertising industry organized around the widespread collection and exchange of consumer data.

What is significant about this period is not DoubleClick’s particular business history per se, but its leadership in enlarging the social domain of consumer surveillance and creating a prototype for surveillance advertising’s subsequent propulsion by market forces. As I discuss in the Conclusion of this book, DoubleClick was purchased by Google in the mid-2000s and became an integral part of the search giant’s embrace of surveillance advertising. But already in the dot-com era, competitive pressure pushed all marketers and publishers toward indiscriminate data collection. It was a classic race to the bottom in which participating in surveillance advertising became more or less mandatory. In order to get the most value from web advertising, publishers and marketers alike had to become more than buyers and sellers of banner ads; they had to become terminals of surveillance, continuously collecting consumer data and exchanging it with the platform hubs.

As Adweek reported, “No one really knows just what data matters. . . . Companies are tracking everything they can now, with the idea of figuring out what’s important later.”86 “People are erring on the side of collecting too much data rather than too little,” admitted one dot-com executive.87 Ad networks were particularly voracious. DoubleClick boasted a database of 120 million user profiles in 2000, twelve times what it had at the time of its IPO two years before, and roughly on par with the total number of internet users in the United States at the time.88 Competitors amassed immense profile caches as well, with CMGI’s Engage at 70 million, MatchLogic at 65 million, and 24/7 Media at 60 million.89 The trend was so pervasive that data systems provider Oracle reported a 30 percent increase in orders for enterprise-class data warehouses and explicitly linked the surge in demand to dot-coms looking to store hundreds of terabytes of data collected from the web.90 At that time, only the largest brick-and-mortar businesses had databases requiring storage in the terabyte range, and even then only after years of data collection.

The web of the late 1990s became a surveillance free-for-all as monitoring exploded without regard for consumer awareness, let alone consent. The prevailing business logic accepted such practices as beneficial to consumers and necessary “for mainstream advertisers to do more than kick the tires of online advertising.”91 Significant political challenges arose only after DoubleClick and others announced plans to integrate web profiles with consumers’ off-line personal information. The resulting political battles are the subject of the next chapter.