The Culture Code: The Secrets of Highly Successful Groups

by Daniel Coyle

  • Close physical proximity, often in circles

  • Profuse amounts of eye contact

  • Physical touch (handshakes, fist bumps, hugs)

  • Lots of short, energetic exchanges (no long speeches)

  • High levels of mixing; everyone talks to everyone

  • Few interruptions

  • Lots of questions

  • Intensive, active listening

  • Humor, laughter

  • Small, attentive courtesies (thank-yous, opening doors, etc.)

  One more thing: I found that spending time inside these groups was almost physically addictive. I would extend my reporting trips, inventing excuses to stick around for another day or two. I found myself daydreaming about changing occupations so I could apply for a job with them. There was something irresistible about being around these groups that made me crave more connection.

  The term we use to describe this kind of interaction is chemistry. When you encounter a group with good chemistry, you know it instantly. It’s a paradoxical, powerful sensation, a combination of excitement and deep comfort that sparks mysteriously with certain special groups and not with others. There’s no way to predict it or control it.

  Or is there?

  —

  On the third floor of a shiny modernistic building in Cambridge, Massachusetts, a group of scientists is obsessed with understanding the inner workings of group chemistry. The MIT Human Dynamics Lab is a humble set of offices surrounded by a riot of workshops and offices that contain, among other things, a British telephone booth, a mannequin wearing pants made of aluminum foil, and what appears to be a miniature roller coaster suspended from the ceiling. The lab is run by Alex (Sandy) Pentland, a soft-spoken computer science professor with bright eyes, a bushy gray beard, and the easygoing assuredness of a country doctor. Pentland started out his career studying satellite photos of beaver dens, establishing a research method that never really changed: using technology to reveal hidden patterns of behavior.

  “Human signaling looks like other animal signaling,” Pentland says as we sit down at a coffee table in his small homey office. “You can measure interest levels, who the alpha is, who’s cooperating, who’s mimicking, who’s in synchrony. We have these communication channels, and we do it without thinking about it. For instance, if I lean a few inches closer to you, we might begin mirroring.”

  Pentland leans closer, raises his bushy eyebrows, and opens his eyes wider. It’s a little disconcerting when I find myself doing it too, almost against my will. He smiles reassuringly and leans back. “It only works if we’re close enough to physically touch.”

  Pentland introduces me to a scientist named Oren Lederman, who, as it happens, is in the midst of analyzing a group working on the spaghetti-marshmallow challenge. We walk down the hall to Lederman’s office to look at the video. The group consists of three engineers and a lawyer, and their tower is coming together nicely. “This group’s performance is probably better than the MBAs but not as good as the kindergartners,” Lederman says. “They don’t talk as much, which helps.”

  This is not just Lederman’s opinion—it is fact. As we speak, a river of data from the group’s performance is rolling across his computer screen, including the percentage of time each person spends talking, the energy levels of their voices, their speaking rates, the smoothness of turn taking, the number of interruptions, and the amount each person’s vocal pattern mimics the others. Lederman has captured this data using a small red plastic device the size of a credit card that contains a microphone, GPS, and an array of other sensors.

  The device is called a sociometer. It samples the data five times per second and wirelessly streams it to a server, where it is rendered into a series of graphs. These graphs, Pentland informs me, are only the tip of the data iceberg. If they desire, Lederman and Pentland can equip the sociometers to capture proximity and the percentage of time each participant engages in face-to-face contact.

  All in all, it is the kind of real-time, deep-dive data that you could imagine being used to measure presidential polling results or a golf swing. But this is a different kind of game. The sociometer captures the proto-language that humans use to form safe connection. This language is made up of belonging cues.

  Belonging cues are behaviors that create safe connection in groups. They include, among others, proximity, eye contact, energy, mimicry, turn taking, attention, body language, vocal pitch, consistency of emphasis, and whether everyone talks to everyone else in the group. Like any language, belonging cues can’t be reduced to an isolated moment but rather consist of a steady pulse of interactions within a social relationship. Their function is to answer the ancient, ever-present questions glowing in our brains: Are we safe here? What’s our future with these people? Are there dangers lurking?

  “Modern society is an incredibly recent phenomenon,” Pentland says. “For hundreds of thousands of years, we needed ways to develop cohesion because we depended so much on each other. We used signals long before we used language, and our unconscious brains are incredibly attuned to certain types of behaviors.”

  Belonging cues possess three basic qualities:

  1. Energy: They invest in the exchange that is occurring

  2. Individualization: They treat the person as unique and valued

  3. Future orientation: They signal the relationship will continue

  These cues add up to a message that can be described with a single phrase: You are safe here. They seek to notify our ever-vigilant brains that they can stop worrying about dangers and shift into connection mode, a condition called psychological safety.

  “As humans, we are very good at reading cues; we are incredibly attentive to interpersonal phenomena,” says Amy Edmondson, who studies psychological safety at Harvard. “We have a place in our brain that’s always worried about what people think of us, especially higher-ups. As far as our brain is concerned, if our social system rejects us, we could die. Given that our sense of danger is so natural and automatic, organizations have to do some pretty special things to overcome that natural trigger.”

  The key to creating psychological safety, as Pentland and Edmondson emphasize, is to recognize how deeply obsessed our unconscious brains are with it. A mere hint of belonging is not enough; one or two signals are not enough. We are built to require lots of signaling, over and over. This is why a sense of belonging is easy to destroy and hard to build. The dynamic evokes the words of Texas politician Sam Rayburn: “Any jackass can kick down a barn, but it takes a good carpenter to build one.”

  It’s useful to look at the bad apple experiment in this light. Nick was able to disrupt the chemistry of the groups merely by sending a few cues of nonbelonging. His behavior was a powerful signal to the group—We are not safe—which immediately caused the group’s performance to fall apart. Jonathan, on the other hand, delivered a steady pulse of subtle behaviors that signaled safety. He connected individually, listened intently, and signaled the importance of the relationship. He was a wellspring of belonging cues, and the group responded accordingly.

  In recent years, Pentland and his team have used sociometers to capture the interactions of hundreds of groups in post-op wards, call centers, banks, salary negotiations, and business pitch sessions. In each study, they discovered the same pattern: It’s possible to predict performance by ignoring all the informational content in the exchange and focusing on a handful of belonging cues.

  For example, Pentland and Jared Curhan used sociometers to analyze forty-six simulated negotiations between pairs of business students who played the role of employee and boss. The task was to negotiate the terms for a new position, including salary, company car, vacation, and health benefits. Pentland and Curhan found that the first five minutes of sociometric data strongly predicted the outcomes of the negotiations. In other words, the belonging cues sent in the initial moments of the interaction mattered more than anything they said.

  Another experiment analyzed a competition
in which entrepreneurs pitched business ideas to a group of executives. Each participant presented their plan to the group; the group then selected and ranked the most promising plans for recommendation to an outside group of angel investors. Pentland found that the sociometers—which tracked only the cues exchanged by presenter and audience and ignored all the informational content—predicted the rankings with nearly perfect accuracy. In other words, the content of the pitch didn’t matter as much as the set of cues with which the pitch was delivered and received. (When the angel investors viewed the plans on paper—looking only at informational content and ignoring social signals—they ranked them very differently.)

  “The executives [listening to the pitches] thought they were evaluating the plans based on rational measures, such as: How original is this idea? How does it fit the current market? How well developed is this plan?” Pentland wrote. “While listening to the pitches, though, another part of their brain was registering other crucial information, such as: How much does this person believe in this idea? How confident are they when speaking? How determined are they to make this work? And the second set of information—information that the business executives didn’t even know they were assessing—is what influenced their choice of business plans to the greatest degree.”

  “This is a different way of thinking about human beings,” Pentland says. “Individuals aren’t really individuals. They’re more like musicians in a jazz quartet, forming a web of unconscious actions and reactions to complement the others in the group. You don’t look at the informational content of the messages; you look at patterns that show how the message is being sent. Those patterns contain many signals that tell us about the relationship and what’s really going on beneath the surface.”

  Overall Pentland’s studies show that team performance is driven by five measurable factors:

  1. Everyone in the group talks and listens in roughly equal measure, keeping contributions short.

  2. Members maintain high levels of eye contact, and their conversations and gestures are energetic.

  3. Members communicate directly with one another, not just with the team leader.

  4. Members carry on back-channel or side conversations within the team.

  5. Members periodically break, go exploring outside the team, and bring information back to share with the others.

  These factors ignore every individual skill and attribute we associate with high-performing groups, and replace them with behaviors we would normally consider so primitive as to be trivial. And yet when it comes to predicting team performance, Pentland and his colleagues have calculated nothing is more powerful.

  “Collective intelligence is not that different in some ways than apes in a forest,” Pentland says. “One [ape] is enthusiastic, and that signal recruits others, and they jump in and start doing stuff together. That’s the way group intelligence works, and this is what people don’t get. Just hearing something said rarely results in a change in behavior. They’re just words. When we see people in our peer group play with an idea, our behavior changes. That’s how intelligence is created. That’s how culture is created.”

  They’re just words. This is not how we normally think. Normally, we think words matter; we think that group performance correlates with its members’ verbal intelligence and their ability to construct and communicate complex ideas. But that assumption is wrong. Words are noise. Group performance depends on behavior that communicates one powerful overarching idea: We are safe and connected.

  * * *

  * Not coincidentally, many successful groups have adopted the use of family-esque identifiers. People who work at Pixar are Pixarians, and people who work at Google are Googlers. It’s the same with Zappos (Zapponians), KIPP (KIPPsters), and others.

  In the early 2000s, some of the best minds in America were competing quietly in a race. The goal was to build a software engine that connected Internet user searches with targeted advertisements, an esoteric-sounding task that would potentially unlock a multibillion-dollar market. The question was which company would win.

  The overwhelming favorite was Overture, a well-funded Los Angeles outfit led by a brilliant entrepreneur named Bill Gross. Gross had pioneered the field of Internet advertising. He had invented the pay-per-click advertising model, written the code, and built Overture into a thriving business that was generating hundreds of millions of dollars in profits, as well as a recent initial public offering valued at one billion dollars. In other words, the contest between Overture and its competitors appeared to be a profound mismatch. The market had placed a billion-dollar bet on Overture for the same reason that you would have bet on the MBA students to defeat the kindergartners in the spaghetti-marshmallow challenge: because Overture possessed the intelligence, experience, and resources to win.

  But Overture did not win. The winner of this race turned out to be a small, young company called Google. What’s more, it’s possible to isolate the moment that turned the race in its favor. On May 24, 2002, in Google’s kitchen at 2400 Bayshore Parkway in Mountain View, California, Google founder Larry Page pinned a note to the wall. The note contained three words:

  THESE ADS SUCK

  In the traditional business world, it was not considered normal to leave notes like this in the company kitchen. However, Page was not a traditional businessperson. For starters, he looked like a seventh-grader, with large, watchful eyes, a bowl haircut, and a tendency to speak in abrupt machine-gun bursts. His main leadership technique, if it could be called a technique, consisted of starting and sustaining big, energetic, no-holds-barred debates about how to build the best strategies, products, and ideas. To work at Google was to enter a giant, continuous wrestling match in which no person was considered above the fray.

  This approach extended to the raucous all-employee street hockey games in the parking lot (“No one held back when fighting the founders for the puck,” recalled one player) and to the all-company Friday forums, where anyone could challenge the founders with any question under the sun, no matter how controversial—and vice versa. Like the hockey games, the Friday forums often turned into collision-filled affairs.

  On the day Page pinned his note to the kitchen wall, Google’s competition with Overture was not going well. The project, which Google called the AdWords engine, was struggling to accomplish the basic task of matching search terms to appropriate ads. For example, if you typed in a search for a Kawasaki H1B motorcycle, you’d receive ads from lawyers offering help with your H-1B foreign visa application—precisely the kinds of failures that could doom the project. So Page printed out examples of these failures, scrawled his three-word verdict in capital letters, and pinned the whole mess to the kitchen bulletin board. Then he left.

  Jeff Dean was one of the last people in Google’s office to see Page’s note. A quiet, skinny engineer from Minnesota, Dean was in most ways Page’s opposite: smiley, sociable, unfailingly polite, and known around the office for his love of cappuccinos. Dean had no immediate motive to care about the AdWords problem. He worked in Search, which was a different area of the company, and he was more than busy navigating his own urgent problems. But at some point that Friday afternoon, Dean walked over to the kitchen to make a cappuccino and spotted Page’s note. He flipped through the attached pages—and as he did, a thought flickered through his mind, a hazy memory of a similar problem he’d encountered a while back.

  Dean walked back to his desk and started trying to fix the AdWords engine. He did not ask permission or tell anyone; he simply dove in. On almost every level, his decision made no sense. He was ignoring the mountain of work on his desk in order to wrestle with a difficult problem that no one expected him to take on. He could have quit at any point, and no one would have known. But he did not quit. In fact, he came in on Saturday and worked on the AdWords problem for several hours. On Sunday night, he had dinner with his family and put his two young children to bed. Around nine P.M., he drove back to the office, made another cappuccino, and worked thro
ugh the night. At 5:05 A.M. on Monday, he sent out an email outlining a proposed fix. Then he drove home, climbed into bed, and went to sleep.

  It worked. Dean’s fix unlocked the problem, instantly boosting the engine’s accuracy scores by double digits. On the strength of that improvement and subsequent others it inspired, AdWords swiftly came to dominate the pay-per-click market. Overture’s effort, hamstrung by infighting and bureaucracy, faltered. In the year following Dean’s fix, Google’s profits went from $6 million to $99 million. By 2014, the AdWords engine was producing $160 million per day, and advertising was providing 90 percent of Google’s revenues. The success of the AdWords engine, author Stephen Levy wrote, was “sudden, transforming, decisive, and, for Google’s investors and employees, glorious….It became the lifeblood of Google, funding every new idea and innovation the company conceived of thereafter.”

  Yet that was not the strange part of the story. Because inside Google, there remained one key person for whom this incident didn’t mean much—for whom the events of that historic weekend registered so faintly that he barely remembered it. That person happened to be Jeff Dean.

  One day in 2013, Google adviser Jonathan Rosenberg approached Dean for a book he was co-writing about Google. Rosenberg wanted to get Dean’s version of the story, so he started in—I want to talk to you about the AdWords engine, Larry’s note, the kitchen—naturally expecting Dean to pick up on the cue and launch into a reminiscence. But Dean didn’t do that. Instead, he just stared at Rosenberg with a pleasantly blank expression. Rosenberg, slightly confused, kept going, filling in detail after detail. Only then did Dean’s face dawn with the light of recognition—oh yeah!

  This is not the response you would expect Dean to have. It is roughly the equivalent of Michael Jordan forgetting that he won six NBA titles. But that was how Dean felt and how he still feels today.