by Gillian Tett
Not everyone in the academic community agreed with Dunbar. After he published his groundbreaking research, other anthropologists, neurologists, and biologists tried their own tests and some concluded that the optimal size of a social group could be twice this size. But Zuckerberg and the other founders of Facebook were fascinated by Dunbar’s number, and they eventually asked Dunbar to provide some consulting advice. Initially, their interest was commercially focused. As they designed their site the Facebook engineers wanted to know how many friends a Facebook user was likely to have so they could build their systems accordingly. However, as the Facebook engineers talked to Dunbar, they realized that his findings not only had implications for how Facebook should build the website for external users. The findings could also affect how employees interacted with each other inside the company. Back in the early days when Zuckerberg had first created his company, the employees had worked as a single group. Some of them lived together, they worked in close quarters, knew each other well, and had joint rituals, such as ordering food from a local Chinese takeout. However, as the company swelled in size, it was harder to maintain this sense of group identity.
Facebook was not the only company to face that challenge. It afflicted every other successful technology start-up. And the history of Silicon Valley suggested that for many companies the problems posed by a rapid expansion in size were deadly. After all, the technology world is littered with tales of companies that have started life as small, freewheeling entities, which then enjoyed great success but later turned into gigantic bureaucracies, riddled with infighting and silos. Sony was one case. Xerox another. The example that particularly worried some Facebook engineers, however, was Microsoft. Although the Seattle-based group had started off as a dynamic and creative entity, by the turn of the century it was plagued with silos. This fragmentation was not as extreme as at Sony, but nevertheless undermined Microsoft’s ability to compete.
So was there any way to avoid that fate? The Facebook engineers were determined to try. “We want to be the anti-Sony, the anti-Microsoft—we look at companies like that and see what we don’t want to become,” one senior manager later recalled. They started tossing ideas around about how to combat the problem. In the summer of 2008, one of the early Facebook founders, Andrew Bosworth, a burly, baldheaded, tattooed, engineer, floated a novel idea.17 Bosworth was known as “Boz” to his colleagues. (The Facebook engineers loved to give each other nicknames; it was part of the process of social grooming.) In previous months, Boz had been trying to create a training program for new recruits. His goal was to ensure that when computing engineers joined Facebook they knew the same set of computer codes as all the other employees, and were assigned to a team that used their skills most effectively. So he created an introductory course that showed them the company and taught them crucial coding knowledge. But Boz then realized that the course could do more than impart technical knowledge. It could also be a tool of social engineering. After all, if you put all the new recruits through a common training experience in small groups, you could create a mechanism for some social grooming and bonding. And while these groups of trainees would not stay together as a unit, since they were destined to be scattered across the company, the joint experience could create lasting ties between them, and the type of social intimacy that fostered nicknames.
That summer Facebook declared that all of its new employees—no matter how junior or senior—would undergo a six-week induction process when they joined. Boz was named “Bootcamp Drill Sergeant.” “The primary goal of Bootcamp is to get people up to speed on all parts of our code base while promoting good habits that we believe will pay dividends in the long term, such as fearlessly fixing bugs as we come across them, rather than leaving them for future engineers,” Boz explained in a written message to the staff, posted on Facebook. “A small number of rotating senior engineers serve as mentors and meet with the new engineers regularly to coach them on how to be more effective at Facebook. The mentors review all the bootcampers’ code and even hold office hours to answer any basic questions that engineers might otherwise be too timid to ask. Senior engineers from across the engineering team also give a bunch of tech talks on a broad range of the technologies we use from MySQL and Memcached to CSS and JavaScript.”18 A crucial part of that training process, Boz added, would be a rotation program to show the new recruits the entire company. “Instead of assigning engineers to teams arbitrarily based on a small amount of interaction during interviews, bootcampers choose the team they will join at the end of their six weeks.”
However, the new recruits were not just being asked to just learn technologies such as MySQL. “Bootcampers tend to form bonds with their classmates who joined near the same time and those bonds persist even after each has joined different teams,” Boz explained. Or to put it another way, the Facebook managers were trying to use Bootcamp to do not one, but two, things. First, they were organizing the company into discrete project teams, dedicated groups to perform tasks. This was essential because the process of writing code requires intense collaboration on specific projects or processes. A company such as Facebook needs silos, in the sense of needing specialist departments and teams, simply to get its work done. Project groups were needed for focus and accountability. But the second aim of Bootcamp was to overlay those project teams with a second set of informal social ties not defined by the formal department boundaries. This, it was hoped, would prevent the project teams from hardening into rigid, inward-looking groups, and ensure that employees felt a sense of affiliation with the entire company, not just their tiny group. “Bootcamp [can foster] cross team communication and prevent the silos that so commonly spring up in growing engineering organizations,” Boz said. Facebook was both creating the preconditions for silos and instilling systems to break down those silos.
IN THE AUTUMN OF 2010, Jocelyn Goldfein graduated from her own Bootcamp19 and took charge of a small team of engineers working on a project known as “News Feed.” This was a tool on the social media platform that had first been launched back in 2006, which essentially enabled a Facebook user to see “news”—or posts—from friends, grouped into a chronological sequence.20 By 2010, the tool was wildly popular with many Facebook users and there did not appear to be any commercial imperative to update it. However, it was a core mantra of Facebook, like other technology groups such as Apple, that to survive the company had to constantly remake its most successful products. “If we don’t disrupt ourselves, someone else will disrupt us,” Goldfein observed.21 So Goldfein’s team dove into the challenge of improving News Feed. It was a complex computing problem. To make News Feed work well, Facebook needed an algorithm that would automatically select the most important details of all the news (or feeds) that a Facebook user was receiving from their friends, and present them in an easy-to-read manner. Initially, Facebook had done this task by just listing all the news that a user had received, in crude chronological order, on their computer screen. That worked well in the early days when the site was small and Facebook users did not receive many posts or messages. But by 2010 the site had exploded so dramatically in size that users were drowning in news. Posts that were very important, or a “MLE” (Major Life Event) in Facebook parlance, could get drowned out by trivia. News about a death might be given the same weight as a picture of a kitten, and then almost disappear from view if there were hundreds of kitty snaps.
So Goldfein’s team set about hunting for ways to create a more sensitive algorithm to rank information. It was intensive work, requiring hours at the keyboard, pushing the boundaries of coding. Or as a newspaper in India observed: “It wouldn’t be an exaggeration to say the task [of rewriting News Feed] is at the frontiers of computer science, since it uses artificial intelligence derived from the huge amounts of data FB has on user behaviour.” First, the team experimented with writing pieces of code that would group news according to clusters of topics that appeared among users. Then they tried treating the News Feed as if it were a newspaper, pl
acing MLEs at the top of any feed. But nothing worked well. “We could do a variation [of a code] in a week or two, but a theme [for the site] would take five weeks and there were three major themes in the course of four or five months,” Goldfein explained. Then the team tried another method called the “boulder prototype.” This algorithm picked out the biggest stories and assigned them “boulder status,” around which other related stories appeared as “pebbles.” This system worked better. “If you log on after a week, you get all the important stories since you last logged in,” Goldfein said, “[but] if you hang out on FB, you’ll see changed feeds. It has the magic element because you see the best stories first, but it feels less deterministic, which is something users like.”22
As the months passed, and the News Feed team spent hours at their keyboards, tossing ideas around, they became increasingly close. Indeed, to Goldfein the experience was so intense that she often felt as if she was working on a start-up. That was deliberate: Zuckerberg and the other Facebook managers wanted to give the separate coding groups as much freedom as possible to brainstorm, experiment, and develop ideas on their own, in a fast-moving and entrepreneurial way. They reckoned that the sense of independence was crucial to enable the company to develop quickly. But even as the team bonded as a quasi-start-up around a project, the Facebook managers also kept trying to pull them back into the wider group. Each week, Goldfein would meet with the other senior engineers to explain her progress. “Mark [Zuckerberg] was the hub around which we moved—we would meet once a week with him and explain what we were doing,” she observed.23 The Facebook staff often met up with the other members of their Bootcamp groups. Under the guise of conducting social events, they were being encouraged to swap ideas and news. “Facebook started doing the Bootcamps to solve a really different problem, which was giving people greater freedom of choice about where they wanted to work later,” Goldfein explained. “But this has serendipitously been huge for silo busting. The power of knowing at least one person in each silo is crucial for making the company work.”24
Goldfein also shuffled her own team. To do this she used a second Facebook ritual: the “Hackamonth.” This practice had developed out of the Bootcamp idea. “Hackamonth is basically the second part of Bootcamp. It’s effectively a rotation program,” Michael Schroepfer, the chief engineer of Facebook, explained. A small wiry man, he was normally known as “Schrep” inside the company.
“If you have been working on the same thing for 12–18 months we then tap you on a shoulder and say get to work on something else for a few months! Most people end up choosing something pretty far away from what they are doing.”25 As with the Bootcamp idea, the Hackamonth ritual had arisen partly by accident and partly by experimentation. Schrep and Boz had started it with the aim of keeping the engineers motivated by their jobs. Technology companies in Silicon Valley were growing so fast that when engineers got bored, they were liable to be poached by rivals. “When people actively choose to do something, they are going to do better work,” Schrep explained. “Passion is a force multiplier—and that offsets everything else.”26 However, once Hackamonth started, it became clear that this ritual could break down silos too. Moving people around prevented the different teams from hardening into inward-looking units. So the Facebook managers expanded the scheme. Facebook was born out of a culture that liked to constantly experiment with small steps and then chase after whatever worked, in an iterative way, irrespective of whether that was in the field of management practices, such as Hackamonth, or computer code. “About 50 percent of the people who go on Hackamonth switch teams and half return to the original team. But either way we think we win,” said Schrep.
The Hackamonth system had some big drawbacks. It was time-consuming to find places to put the engineers who wanted to move, and fill the gaps in their original teams when they left. Inevitably there was some duplication and overlap, if not waste. “All of these processes are painful and inefficient. Sometimes you have two or three people away from a team, one on Hackamonth, one on parental leave, or whatever. It would be much easier to just put one person into each team and tell them to stay there,” Schrep admitted.27 But he concluded it was a small price to pay to meet the goal of keeping the organization fluid and connected; it was crucial to have a bit of slack, or inefficiency, to breed creativity and give people time to stay connected.
So, a few months after she had embarked on the News Feed project, Goldfein rotated one of her key engineers into another group. This second team was working on a separate project linked to the Timeline feature. At first it seemed like a loss for News Feed, since the battle to develop the boulder prototype was at a delicate stage. But then she realized that the rotation had produced a benefit too: the different project teams started swapping ideas. “Timeline owed a lot of its code to News Feed and we hoped that having [the News Feed engineer move teams] would give us closer lines of communication,” Goldfein said. “And it turned out to be very valuable . . . on a day-to-day basis nobody in the company has time to understand what everyone else in the different project teams are doing. But the key thing is to get this rich surface of community and information sharing, in whatever way you can.”
IN DECEMBER 2011, EIGHTEEN months after she had joined Facebook, Goldfein and the other staff moved from the office in Palo Alto into a brand-new campus building in a nearby district, Menlo Park28 By then the Facebook staff had risen above 2,000;29 breaching the Dunbar principle multiple times. However, as the ranks swelled, the social experiments grew apace. The Facebook managers were determined to use whatever social tool they could imagine both to create dedicated, specialist project groups and to prevent those little teams from ossifying into competitive silos. Architecture was one weapon in this fight. The new Facebook campus was based at a site previously been owned by Sun Microsystems, and its logo still sat on the back of the entrance address sign. Sun was yet another tech giant that had once flourished in Silicon Valley in earlier decades, initially as a freewheeling start-up, but later turned into a stodgy behemoth, plagued by silos. When Sun had owned the site, whose official address was 1601 Willow Road, its employees had been tucked into dozens of different buildings, which were subdivided into numerous small offices and cubicles. “It felt like a cattle pen!” laughed Schrep, who had worked there. “You didn’t have much contact with anyone at all.” However, when Facebook bought the site, Zuckerberg decided to christen it “One Hacker Way” and the address sign was painted over in blue and the image of a giant white thumb pointing upward was added on top, the “Like” symbol on the Facebook platform. Builders ripped out most of the internal dividing walls in the old Sun buildings and added whiteboards, bare pipe work, and graffiti walls. Insofar as there were meeting rooms, these were surrounded by walls made of glass, so that anyone could peer into anyone else’s room. Even Zuckerberg worked in the open plan space, visible to all. So did Sheryl Sandberg, the high-profile chief operating officer. Zuckerberg had a “private” office too. However, this was lined with glass and placed in the center of the campus, next to a walkway along which all the employees constantly strolled. A sign saying “Do not feed the animals!” was fastened to the window. “We call it Mark’s goldfish bowl!” quipped Schrep. “Everyone can see him.”
Then Schrep went further. He asked the architects to connect the higher floors of the separate buildings with walkways. These stretched high in the air above the Facebook walkways, painted in the same bright orange-red as San Francisco’s famous Golden Gate Bridge. Doors were installed at the end of these walkways that opened automatically when anyone passed; they were modeled on supermarket doors. Schrep’s goal was to ensure that the engineers never need to pause when they rambled around the building. “There is all this research out there which shows that if you can keep people moving and colliding with each other, you get much more interaction,” he explained. The space between the buildings was turned into attractive “rambling” zones, to encourage employees to hang out together in the balmy California weat
her. In the center of the campus an open meeting spot was created, known as “Hacker Square.” A couple of times a year Zuckerberg held “all-hands” meetings in this open space. He also held town halls (or “Q&A” sessions, as Facebook preferred to call them) every Friday afternoon in a vast cafeteria. Sometimes employees would throw unexpected questions at Zuckerberg. Other times the events seemed more formulaic. But either way, the symbolism was clear: the Facebook managers were determined to present the company to the employees as a single, open mass, where everyone could—and should—collide with everyone else, in a free-wheeling, irreverent way.
Hacker Square was also used to stage another of the experiments in social engineering, the hackathons. Every six weeks or so, several hundred engineers would congregate in the square around a piece of yellow machinery that had once been part of a crane, before later retiring to a large meeting room with bright orange walls plastered with inspirational posters.30 There they would spend all night working in small teams on coding problems; the idea was that the engineers would test out ideas together, or “hack,” to use the slang that computing experts loved to toss around. Being pushed together in a small space and asked to work intensively overnight was one way to unleash the creative juices.