The Age of Surveillance Capitalism

by Shoshana Zuboff

  The critical factor in the divergence of capitalism from all other forms of “commercial society” was the development of certain social property relations that generated market imperatives and capitalist “laws of motion”… competitive production and profit-maximization, the compulsion to reinvest surpluses, and the relentless need to improve labour-productivity associated with capitalism.… Those laws of motion required vast social transformations and upheavals to set them in train. They required a transformation in the human metabolism with nature, in the provision of life’s basic necessities.11

  My argument here is that although surveillance capitalism does not abandon established capitalist “laws” such as competitive production, profit maximization, productivity, and growth, these earlier dynamics now operate in the context of a new logic of accumulation that also introduces its own distinctive laws of motion. Here and in following chapters, we will examine these foundational dynamics, including surveillance capitalism’s idiosyncratic economic imperatives defined by extraction and prediction, its unique approach to economies of scale and scope in raw-material supply, its necessary construction and elaboration of means of behavioral modification that incorporate its machine-intelligence–based “means of production” in a more complex system of action, and the ways in which the requirements of behavioral modification orient all operations toward totalities of information and control, creating the framework for an unprecedented instrumentarian power and its societal implications. For now, my aim is to reconstruct our appreciation of familiar ground through new lenses: Google’s early days of optimism, crisis, and invention.

  II. A Balance of Power

  Google was incorporated in 1998, founded by Stanford graduate students Larry Page and Sergey Brin just two years after the Mosaic browser threw open the doors of the world wide web to the computer-using public. From the start, the company embodied the promise of information capitalism as a liberating and democratic social force that galvanized and delighted second-modernity populations around the world.

  Thanks to this wide embrace, Google successfully imposed computer mediation on broad new domains of human behavior as people searched online and engaged with the web through a growing roster of Google services. As these new activities were informated for the first time, they produced wholly new data resources. For example, in addition to key words, each Google search query produces a wake of collateral data such as the number and pattern of search terms, how a query is phrased, spelling, punctuation, dwell times, click patterns, and location.

  Early on, these behavioral by-products were haphazardly stored and operationally ignored. Amit Patel, a young Stanford graduate student with a special interest in “data mining,” is frequently credited with the groundbreaking insight into the significance of Google’s accidental data caches. His work with these data logs persuaded him that detailed stories about each user—thoughts, feelings, interests—could be constructed from the wake of unstructured signals that trailed every online action. These data, he concluded, actually provided a “broad sensor of human behavior” and could be put to immediate use in realizing cofounder Larry Page’s dream of Search as a comprehensive artificial intelligence.12

  Google’s engineers soon grasped that the continuous flows of collateral behavioral data could turn the search engine into a recursive learning system that constantly improved search results and spurred product innovations such as spell check, translation, and voice recognition. As Kenneth Cukier observed at that time,

  Other search engines in the 1990s had the chance to do the same, but did not pursue it. Around 2000 Yahoo! saw the potential, but nothing came of the idea. It was Google that recognized the gold dust in the detritus of its interactions with its users and took the trouble to collect it up.… Google exploits information that is a by-product of user interactions, or data exhaust, which is automatically recycled to improve the service or create an entirely new product.13

  What had been regarded as waste material—“data exhaust” spewed into Google’s servers during the combustive action of Search—was quickly reimagined as a critical element in the transformation of Google’s search engine into a reflexive process of continuous learning and improvement.

  At that early stage of Google’s development, the feedback loops involved in improving its Search functions produced a balance of power: Search needed people to learn from, and people needed Search to learn from. This symbiosis enabled Google’s algorithms to learn and produce ever-more relevant and comprehensive search results. More queries meant more learning; more learning produced more relevance. More relevance meant more searches and more users.14 By the time the young company held its first press conference in 1999, to announce a $25 million equity investment from two of the most revered Silicon Valley venture capital firms, Sequoia Capital and Kleiner Perkins, Google Search was already fielding seven million requests each day.15 A few years later, Hal Varian, who joined Google as its chief economist in 2002, would note, “Every action a user performs is considered a signal to be analyzed and fed back into the system.”16 The Page Rank algorithm, named after its founder, had already given Google a significant advantage in identifying the most popular results for queries. Over the course of the next few years it would be the capture, storage, analysis, and learning from the by-products of those search queries that would turn Google into the gold standard of web search.

  The key point for us rests on a critical distinction. During this early period, behavioral data were put to work entirely on the user’s behalf. User data provided value at no cost, and that value was reinvested in the user experience in the form of improved services: enhancements that were also offered at no cost to users. Users provided the raw material in the form of behavioral data, and those data were harvested to improve speed, accuracy, and relevance and to help build ancillary products such as translation. I call this the behavioral value reinvestment cycle, in which all behavioral data are reinvested in the improvement of the product or service (see Figure 1).

  The cycle emulates the logic of the iPod; it worked beautifully at Google but with one critical difference: the absence of a sustainable market transaction. In the case of the iPod, the cycle was triggered by the purchase of a high-margin physical product. Subsequent reciprocities improved the iPod product and led to increased sales. Customers were the subjects of the commercial process, which promised alignment with their “what I want, when I want, where I want” demands. At Google, the cycle was similarly oriented toward the individual as its subject, but without a physical product to sell, it floated outside the marketplace, an interaction with “users” rather than a market transaction with customers.

  This helps to explain why it is inaccurate to think of Google’s users as its customers: there is no economic exchange, no price, and no profit. Nor do users function in the role of workers. When a capitalist hires workers and provides them with wages and means of production, the products that they produce belong to the capitalist to sell at a profit. Not so here. Users are not paid for their labor, nor do they operate the means of production, as we’ll discuss in more depth later in this chapter. Finally, people often say that the user is the “product.” This is also misleading, and it is a point that we will revisit more than once. For now let’s say that users are not products, but rather we are the sources of raw-material supply. As we shall see, surveillance capitalism’s unusual products manage to be derived from our behavior while remaining indifferent to our behavior. Its products are about predicting us, without actually caring what we do or what is done to us.

  To summarize, at this early stage of Google’s development, whatever Search users inadvertently gave up that was of value to the company they also used up in the form of improved services. In this reinvestment cycle, serving users with amazing Search results “consumed” all the value that users created when they provided extra behavioral data. The fact that users needed Search about as much as Search needed users created a balance of power between Google and its popula
tions. People were treated as ends in themselves, the subjects of a nonmarket, self-contained cycle that was perfectly aligned with Google’s stated mission “to organize the world’s information, making it universally accessible and useful.”

  Figure 1: The Behavioral Value Reinvestment Cycle

  III. Search for Capitalism: Impatient Money and the State of Exception

  By 1999, despite the splendor of Google’s new world of searchable web pages, its growing computer science capabilities, and its glamorous venture backers, there was no reliable way to turn investors’ money into revenue. The behavioral value reinvestment cycle produced a very cool search function, but it was not yet capitalism. The balance of power made it financially risky and possibly counterproductive to charge users a fee for search services. Selling search results would also have set a dangerous precedent for the firm, assigning a price to indexed information that Google’s web crawler had already taken from others without payment. Without a device like Apple’s iPod or its digital songs, there were no margins, no surplus, nothing left over to sell and turn into revenue.

  Google had relegated advertising to steerage class: its AdWords team consisted of seven people, most of whom shared the founders’ general antipathy toward ads. The tone had been set in Sergey Brin and Larry Page’s milestone paper that unveiled their search engine conception, “The Anatomy of a Large-Scale Hypertextual Web Search Engine,” presented at the 1998 World Wide Web Conference: “We expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers. This type of bias is very difficult to detect but could still have a significant effect on the market… we believe the issue of advertising causes enough mixed incentives that it is crucial to have a competitive search engine that is transparent and in the academic realm.”17

  Google’s first revenues depended instead on exclusive licensing deals to provide web services to portals such as Yahoo! and Japan’s BIGLOBE.18 It also generated modest revenue from sponsored ads linked to search query keywords.19 There were other models for consideration. Rival search engines such as Overture, used exclusively by the then-giant portal AOL, or Inktomi, the search engine adopted by Microsoft, collected revenues from the sites whose pages they indexed. Overture was also successful in attracting online ads with its policy of allowing advertisers to pay for high-ranking search listings, the very format that Brin and Page scorned.20

  Prominent analysts publicly doubted whether Google could compete with its more-established rivals. As the New York Times asked, “Can Google create a business model even remotely as good as its technology?”21 A well-known Forrester Research analyst proclaimed that there were only a few ways for Google to make money with Search: “build a portal [like Yahoo!]… partner with a portal… license the technology… wait for a big company to purchase them.”22

  Despite these general misgivings about Google’s viability, the firm’s prestigious venture backing gave the founders confidence in their ability to raise money. This changed abruptly in April 2000, when the legendary dot-com economy began its steep plunge into recession, and Silicon Valley’s Garden of Eden unexpectedly became the epicenter of a financial earthquake.

  By mid-April, Silicon Valley’s fast-money culture of privilege was under siege with the implosion of what came to be known as the “dot-com bubble.” It is easy to forget exactly how terrifying things were for the valley’s ambitious young people and their slightly older investors. Startups with outsized valuations just months earlier were suddenly forced to shutter. Prominent articles such as “Doom Stalks the Dotcoms” noted that the stock prices of Wall Street’s most-revered internet “high flyers” were “down for the count,” with many of them trading below their initial offering price: “With many dotcoms declining, neither venture capitalists nor Wall Street is eager to give them a dime.…”23 The news brimmed with descriptions of shell-shocked investors. The week of April 10 saw the worst decline in the history of the NASDAQ, where many internet companies had gone public, and there was a growing consensus that the “game” had irreversibly changed.24

  As the business environment in Silicon Valley unraveled, investors’ prospects for cashing out by selling Google to a big company seemed far less likely, and they were not immune to the rising tide of panic. Many Google investors began to express doubts about the company’s prospects, and some threatened to withdraw support. Pressure for profit mounted sharply, despite the fact that Google Search was widely considered the best of all the search engines, traffic to its website was surging, and a thousand résumés flooded the firm’s Mountain View office each day. Page and Brin were seen to be moving too slowly, and their top venture capitalists, John Doerr from Kleiner Perkins and Michael Moritz from Sequoia, were frustrated.25 According to Google chronicler Steven Levy,

  “The VCs were screaming bloody murder. Tech’s salad days were over, and it wasn’t certain that Google would avoid becoming another crushed radish.”26

  The specific character of Silicon Valley’s venture funding, especially during the years leading up to dangerous levels of startup inflation, also contributed to a growing sense of emergency at Google. As Stanford sociologist Mark Granovetter and his colleague Michel Ferrary found in their study of valley venture firms, “A connection with a high-status VC firm signals the high status of the startup and encourages other agents to link to it.”27 These themes may seem obvious now, but it is useful to mark the anxiety of those months of sudden crisis. Prestigious risk investment functioned as a form of vetting—much like acceptance to a top university sorts and legitimates students, elevating a few against the backdrop of the many—especially in the “uncertain” environment characteristic of high-tech investing. Loss of that high-status signaling power assigned a young company to a long list of also-rans in Silicon Valley’s fast-moving saga.

  Other research findings point to the consequences of the impatient money that flooded the valley as inflationary hype drew speculators and ratcheted up the volatility of venture funding.28 Studies of pre-bubble investment patterns showed a “big-score” mentality in which bad results tended to stimulate increased investing as funders chased the belief that some young company would suddenly discover the elusive business model destined to turn all their bets into rivers of gold.29 Startup mortality rates in Silicon Valley outstripped those for other venture capital centers such as Boston and Washington, DC, with impatient money producing a few big wins and many losses.30 Impatient money is also reflected in the size of Silicon Valley startups, which during this period were significantly smaller than in other regions, employing an average of 68 employees as compared to an average of 112 in the rest of the country.31 This reflects an interest in quick returns without spending much time on growing a business or deepening its talent base, let alone developing the institutional capabilities that Joseph Schumpeter would have advised. These propensities were exacerbated by the larger Silicon Valley culture, where net worth was celebrated as the sole measure of success for valley parents and their children.32

  For all their genius and principled insights, Brin and Page could not ignore the mounting sense of emergency. By December 2000, the Wall Street Journal reported on the new “mantra” emerging from Silicon Valley’s investment community: “Simply displaying the ability to make money will not be enough to remain a major player in the years ahead. What will be required will be an ability to show sustained and exponential profits.”33

  IV. The Discovery of Behavioral Surplus

  The declaration of a state of exception functions in politics as cover for the suspension of the rule of law and the introduction of new executive powers justified by crisis.34 At Google in late 2000, it became a rationale for annulling the reciprocal relationship that existed between Google and its users, steeling the founders to abandon their passionate and public opposition to advertising. As a specific response to investors’ anxiety, the founders tasked the tiny AdWords team with the objective of looking for ways to make more m
oney.35 Page demanded that the whole process be simplified for advertisers. In this new approach, he insisted that advertisers “shouldn’t even get involved with choosing keywords—Google would choose them.”36

  Operationally, this meant that Google would turn its own growing cache of behavioral data and its computational power and expertise toward the single task of matching ads with queries. New rhetoric took hold to legitimate this unusual move. If there was to be advertising, then it had to be “relevant” to users. Ads would no longer be linked to keywords in a search query, but rather a particular ad would be “targeted” to a particular individual. Securing this holy grail of advertising would ensure relevance to users and value to advertisers.

  Absent from the new rhetoric was the fact that in pursuit of this new aim, Google would cross into virgin territory by exploiting sensitivities that only its exclusive and detailed collateral behavioral data about millions and later billions of users could reveal. To meet the new objective, the behavioral value reinvestment cycle was rapidly and secretly subordinated to a larger and more complex undertaking. The raw materials that had been solely used to improve the quality of search results would now also be put to use in the service of targeting advertising to individual users. Some data would continue to be applied to service improvement, but the growing stores of collateral signals would be repurposed to improve the profitability of ads for both Google and its advertisers. These behavioral data available for uses beyond service improvement constituted a surplus, and it was on the strength of this behavioral surplus that the young company would find its way to the “sustained and exponential profits” that would be necessary for survival. Thanks to a perceived emergency, a new mutation began to gather form and quietly slip its moorings in the implicit advocacy-oriented social contract of the firm’s original relationship with users.