Zucked

As impressive as all those numbers are, the IPO itself was something of a train wreck. Trading glitches occurred during the first day, preventing some trades from going through, and the stock struggled to stay above the IPO price. The deal set a record for trading volume on the first day after an IPO: 460 million shares.

The months leading up to the IPO saw weakness in Facebook’s advertising sales that triggered reductions in the company’s revenue forecast. When a company is preparing for an IPO, forecast reductions can be disastrous, as public investors have no incentive to buy into uncertainty. In Facebook’s case, investors’ extreme enthusiasm for the company—based primarily on user growth and Facebook’s increasing impact on society—meant the IPO could survive the reduction in forecast, but Zuck’s dream of a record-setting offering might be at risk. As described by former Facebook advertising targeting manager Antonio García Martínez in his book Chaos Monkeys, “The narratives the company had woven about the new magic of social-media marketing were in deep reruns with advertisers, many of whom were beginning to openly question the fortunes they had spent on Facebook thus far, often with little to show for it.” For all its success with users, Facebook had not yet created an advertising product that provided the targeting necessary to provide appropriate results for advertisers. Martínez went on to say, “A colossal yearlong bet the company had made on a product called Open Graph, and its accompanying monetization spin-off, Sponsored Stories, had been an absolute failure in the market.” Advertisers had paid a lot of money to Facebook, believing the company’s promises about ad results, but did not get the value they felt they deserved. For Facebook, this was a moment of truth. By pushing the IPO valuation to record levels, Facebook set itself up for a rocky start as a public company.

The newly public stock sold off almost immediately and went into free fall after Yahoo Finance reported that the investment banks that had underwritten the IPO had reduced their earnings forecasts just before the offering. In the heat of the deal, had those forecast changes been effectively communicated to buyers of the stock? The situation was sufficiently disturbing that regulatory authorities initiated a review. Lawsuits followed, alleging a range of violations with respect to the trading glitches and the actions of one underwriter. A subsequent set of lawsuits named the underwriters, Zuck and Facebook’s board, and Nasdaq. The Wall Street Journal characterized the IPO as a “fiasco.”

For Facebook’s business, though, the IPO was an undisputed blessing. The company received a staggering amount of free publicity before the deal, essentially all of it good. That turbocharged user growth, news of which enabled Facebook to survive the IPO issues with relatively little damage. Investors trusted that a company with such impressive user growth would eventually figure out monetization. Once again, Facebook pushed the envelope, stumbled, and got away with it. Then they did something really aggressive.

The data from inside Facebook alone did not deliver enough value for advertisers. Thanks to Connect and the ubiquitous Like and Share buttons, Facebook had gathered staggering amounts of data about user behavior from around the web. The company had chosen not to use the off-site data for commercial purposes, a self-imposed rule that it decided to discard when the business slowed down. No one knew yet how valuable the external data would be, but they decided to find out. As Martínez describes it, Zuck and Sheryl began cautiously, fearful of alienating users.

Thanks to the IPO, Facebook enjoyed a tsunami of user growth. Within a few months, user growth restored investor confidence. It also overwhelmed the complaints from advertisers, who had to go where their customers were, even if the ad vehicles on Facebook were disappointing. The pressure to integrate user data from activities away from Facebook into the ad products lessened a bit, but the fundamental issues with targeting and the value of ads remained. As a result, the decision to integrate user data from outside Facebook would not be reversed.

In early October 2012, the company announced it had surpassed one billion monthly users, with 600 million mobile users, 219 billion photo uploads, and 140 billion friend connections. Despite the mess of the IPO—and not being privy to the issues with ads—I took great pride in Facebook’s success. The stock turned out to be a game changer for Elevation. Even though my partners had turned down our first opportunity to invest, Elevation subsequently made a large investment at a relatively low price, ensuring on its own that the fund would be a winner.

Only eight and a half years from Zuck’s dorm room, Facebook had become a powerful economic engine. Thanks to the philosophy of “move fast and break things,” no one at Facebook was satisfied with a record-setting IPO. They began hacking away at the problem of monetizing users. There were several challenges. As Martínez wrote in Chaos Monkeys, the advertising team around the time of the IPO was, for the most part, young people who had no previous work experience in advertising or even media. They learned everything by trial and error. For every innovation, there were many mistakes, some of which would have been obvious to a more experienced team. The team may have been young, but they were smart, highly motivated, and persistent. Their leadership, with Sheryl Sandberg at the top, created a successful sales culture. They took a long view and learned from every mistake. They focused on metrics.

In the early days, Facebook did its best to create effective advertising products and tools from profile data, users’ friend relationships, and user actions on the site. My band, Moonalice, was an early advertiser, with a budget of less than ten thousand dollars a year. Our first ads, a few years before the IPO, were tiny rectangles on the side of the page, with a few words of text and maybe a link. The goal was to introduce Moonalice to new fans. We promoted a song called “It’s 4:20 Somewhere” this way. We ran an ad for several years—typically spending ten or twenty dollars a day—and people downloaded the song 4.6 million times, a number cited by the Rock & Roll Hall of Fame as a record from any band’s own website. A little Facebook ad running every day for three years made it possible. But when the only option was a tiny rectangle, Facebook ads were ineffective for many advertisers and products. The same format that worked so well for downloading a song was worthless for promoting a conference. I have no idea why it worked for one thing and not the other. It did not matter much, because in those days, Facebook allowed plenty of free distribution, which they called “organic reach.” Back then, for a fan page likes ours, Facebook would let a really compelling post reach about 15 percent of our fans for free. The value of organic reach on Facebook compelled us and millions of others to shift the focus of our communications from a website to Facebook. We trained our fans to interact with us on Facebook and to use our website as a content archive. Many others did the same, helping to cement Facebook’s position as the social hub of the web. Embracing Facebook worked really well for Moonalice, and our page eventually gathered more than 420,000 followers.

Not surprisingly, there was a catch. Every year or so, Facebook would adjust the algorithm to reduce organic reach. The company made its money from advertising, and having convinced millions of organizations to set up shop on the platform, Facebook held all the cards. The biggest beneficiaries of organic reach had no choice but to buy ads to maintain their overall reach. They had invested too much time and had established too much brand equity on Facebook to abandon the platform. Organic reach declined in fits and starts until it finally bottomed at about 1 percent or less. Fortunately, Facebook would periodically introduce a new product—the Facebook Live video service, for example—and give those new products greater organic reach to persuade people like us to use them. We signed up for Facebook Live on the first day it was available and streamed a concert that same day. The reach was fantastic. Facebook Live and Moonalice were made for each other. I streamed one set of the very first concert by Dead & Company, a spin-off from the Grateful Dead, and so many people watched it that the band saw a surge in ticket sales for other dates on the tour. As a result, one of the band’s managers invited me to stream their next show from the stage.

At the time of the IPO, targeting options for Facebook ads were limited to demographic information from activity on the site, things like age, sex, and location, as well as interests and relationships. The introduction of Open Graph and News Feed ads in 2012 set the stage for much better targeting, which improved rapidly when Facebook integrated “off-site” data into the suite available to advertisers. If Moonalice wanted to promote a concert, we would target demographic data—say, people over twenty-one in the city where our gig was—and then filter that audience with interests, like “concerts,” “Beatles,” and “hippie.” We would spend perhaps one hundred dollars to promote a show on Facebook. We would get a few thousand “impressions,” which in theory meant users who saw the ad. The nature of News Feed is such that users race past a lot of posts. To capture attention, we switched from promoting Events—which is what Facebook called our concerts—to creating posts that included a rich graphic element, such as a poster. In doing so, we ran afoul of the 20 percent rule. As it was explained to me by a senior Facebook executive, Zuck had decided that too much text made ads boring, so he set an arbitrary limit of 20 percent text. Our posters are works of art, but many of them violated the 20 percent rule because rock posters sometimes integrate lots of text into the art. Facebook would reject those ads, so I learned to superimpose a little bit of text onto attention-grabbing photos to create compelling images that would comply with Facebook’s rules.

Moonalice is not a sophisticated advertiser, but we were not alone in that regard. Facebook enabled millions of organizations with small budgets to reach audiences for a fraction of the cost of print, radio, or TV advertisements. But Facebook recognized that the really big money would come from attracting the advertisers who had historically spent giant budgets on traditional media. Such advertisers had completely different expectations. They expected to reach large, targeted audiences at a reasonable cost with complete transparency, which is to say they wanted proof that their messages reached their intended audiences. At the time of the IPO, Facebook could not meet those expectations consistently. In 2013, Facebook began experimenting with data from user activity outside Facebook. They created tools for advertisers to exploit that data. The tools enabled Facebook advertisers the ability to target audiences whose emotions were being triggered in predictable ways.

Facebook’s culture matched its advertising challenge perfectly. A company that prided itself on its software hacking roots perfected a new model to monetize its success. Growth hacking applies the intensely focused, iterative model of software hacking to the problem of increasing user count, time on site, and revenue. It works only when a company has a successful product and a form of monetization that can benefit from tinkering, but for the right kind of company, growth hacking can be transformational. Obsessive focus on metrics is a central feature of growth hacking, so it really matters that you pick the correct metrics.

From late 2012 to 2017, Facebook perfected growth hacking. The company experimented constantly with algorithms, new data types, and small changes in design, measuring everything. Every action a user took gave Facebook a better understanding of that user—and of that user’s friends—enabling the company to make tiny improvements in the “user experience” every day, which is to say they got better at manipulating the attention of users. The goal of growth hacking is to generate more revenue and profits, and at Facebook those metrics blocked out all other considerations. In the world of growth hacking, users are a metric, not people. It is unlikely that civic responsibility ever came up in Facebook’s internal conversations about growth hacking. Once the company started applying user data from outside the platform, there was no turning back. The data from outside Facebook transformed targeting inside Facebook. Additional data improved it more, giving Facebook an incentive to gather data anywhere it could be gathered. The algorithms looked for and found unexpected correlations in the data that could be monetized effectively. Before long, Facebook’s surveillance capabilities rivaled those of an intelligence agency.

To deliver better targeting, Facebook introduced new tools for advertisers. Relative to the 2016 election, the two most important may have been Custom Audiences and Lookalike Audiences. As described in Facebook’s Advertiser Help Center, a Custom Audience is “a type of audience you can create made up of your existing customers. You can target ads to the audience you’ve created on Facebook, Instagram, and Audience Network,” the last of which is “a network of publisher-owned apps and sites where you can show your ads” outside the Facebook platform. Introduced in 2013, Custom Audiences had two very important forms of value to advertisers: first, advertisers could build an ad campaign around known customers, and second, a Custom Audience could be used to create a Lookalike Audience, which finds other Facebook users who share characteristics with the Custom Audience. Lookalike scales without limit, so advertisers have the option to find every user on Facebook with a given set of characteristics. You could start with as few as one hundred people, but the larger the Custom Audience, the better the Lookalike will be. Facebook recommends using a Custom Audience between one thousand and fifty thousand.

Thanks to growth hacking, Facebook made continuous improvements in its advertising tools, as well as growing its audience, increasing time on site, and gathering astonishing amounts of data. Progress against these metrics translated into explosive revenue growth. From one billion users at year-end 2012, Facebook grew to 1.2 billion in 2013, 1.4 billion in 2014, 1.6 billion in 2015, nearly 1.9 billion in 2016, 2.1 billion in 2017, 2.3 billion in 2018, and 2.4 billion in late 2019. From just more than $5 billion in sales in the IPO year of 2012, Facebook grew to $7.8 billion in 2013, $12.5 billion in 2014, $17.9 billion in 2015, $27.6 billion in 2016, $40.7 billion in 2017, and $55.8 billion in 2018. There were issues along the way. Advertisers complained about the lack of transparency relative to advertising, and Facebook has been sued for inflating some metrics, including ad views and video views. But Facebook had become a juggernaut. The customers that advertisers needed to reach were on Facebook. This gave Facebook enormous leverage. When advertisers complained, Facebook could get away with apologies and marginal fixes. With its customary laser focus on a handful of metrics, Facebook did not devote any energy to questioning its decisions. If there was any soul searching about the morality of intense surveillance and the manipulation of user attention, or about protecting users against unintended consequences, I have been able to find no evidence of it. If Zuck and the Facebook team noticed that usage of Facebook differed materially from their ideal, they showed no concern. If anyone noticed the increasingly extreme behavior inside some Facebook Groups, no one took action. The Russians exploited this to sow dissention among Americans and Western Europeans, beginning in 2014. When The Guardian newspaper in the UK broke the story in December 2015 that Cambridge Analytica had misappropriated profiles from at least fifty million Facebook users, it precipitated an intense but brief scandal. Facebook apologized and made Cambridge Analytica sign a piece of paper, certifying that it had destroyed the data set, but then quickly returned to business as usual. While always careful to protect itself from legal liability, Facebook seemed oblivious to signs of trouble. The benefits of growth—in terms of revenue, profits, and influence—were obvious, the problems easy to ignore. At Facebook, everyone remained focused on their metrics.

As 2016 began, Facebook was on a huge roll. Aside from a few PR headaches, the company had not skipped a beat since its IPO. Almost everything that mattered had changed since my days as a mentor to Zuck, and I knew only what had been publicly disclosed or what I had seen with my own eyes. Like its users, Facebook showed the public only the good news, which is why I was so surprised by what I saw during 2016. Bad actors using Facebook’s tools to harm innocent people did not compute for me, but I saw the evidence and could not let it go. It was only when I reengaged with Zuck and Sheryl just before the election that I began to appreciate that my view of Facebook was inaccurate. It took me longer than I would have liked to understand the problem. More than four years of relentless success at Facebook had bred overconfidence. The company was in a filter bubble of its own. Every day, there were more users, spending more time on the site, generating more revenue and earnings, which pushed the stock to new highs. The Midas Effect may have set in, causing Zuck and his team to believe that everything they did was right, always for the best, and uncontestably good for humanity. Humility went out the window. Facebook subordinated everything to growth. Eventually that would create problems it could not resolve with an apology and a promise to do better.