3. The Water Cooler Effect: Why a Friendly Chat Is the Most Important Part of the Work Day

After you get sensing data, what should you analyze first? You could go after org charts, meetings, or compensation systems, but investigating something basic would be better, something fundamental to what it means to be human—such as water. For example, look at the ubiquitous source of water in offices across the world: the water cooler.

Buying a water cooler is the single most important investment a company can make. Well, maybe that’s overstating things a bit, but not by much (and no, this is not a commercial plug that Poland Springs^®, makers of the finest water coolers available, paid for me to put in this book).

The reason that water coolers are so important isn’t just that they slake our thirst for cool, refreshing H₂0, but rather that they create a nexus of social activity in the workplace. Water coolers are where you bump into people in the office that you haven’t seen in a while, and they’re where you gossip about coworkers or talk about last night’s game. They serve a crucial social function that desks and meeting rooms can’t provide.

Although this discussion focuses on water coolers, really any watering hole in the office has similar effects. Coffee machines, kitchens, cafes, and recreational areas provide a similar environment that can greatly enhance social connectivity in the workplace.

Sadly, in the vast majority of companies the water cooler is an afterthought, relegated to some corner where there happens to be a spare power outlet. The location of the water cooler isn’t a topic of discussion at the higher levels of management, and usually is decided based on where there’s enough space to put it, rather than using it to facilitate interaction.

This practice indicates a broader problem in workplaces. Companies rarely think about the things that aren’t formal aspects of work, but spend years crafting org charts, setting up IT systems, and planning organizational strategy. They should be spending time doing those things. They’re critical components for every major company in the world. The point is that communication and collaboration also need some attention.

We collaborate by passing information to one another. We can communicate using e-mail, phone calls, or talking to each other face-to-face, but no one in the history of work has ever created an organization where people can collaboratively make a product with no communication. In some sense, you could argue that if people are “collaborating” on a product and not talking to each other, then they are actually making separate products that work in tandem. This is equivalent to a person who makes light bulbs and a person who makes a lamp. You can’t use one without the other, but those people don’t need to talk to each other to each make a fully functional product.

That’s not to say that you need to talk with everyone at your company—far from it. If you work in a company with hundreds of thousands of employees, there’s no way you can meaningfully communicate with everybody. You could send an e-mail to everyone, but you’ll end up consuming hundreds of thousands of seconds for them to read that one e-mail. If everyone did that, no one would do anything but read mass e-mails all day. You could even try to communicate over the phone or face to face with as many people as possible, but at most you could have a meaningful conversation with about 100 people a day, which means that in five years at an average Fortune 500 company, you could have one five-minute conversation with every single employee.

Clearly, those two extremes don’t make sense. So is it just the amount of communication that matters? That’s also not the case.

Let’s say you want to spend one hour of your day interacting with people. Reasonably, you should be able to use that time to talk with around 10 people. Which 10 people should you talk to? If only the amount of communication mattered, then you could pick 10 people at random from the company and speak with them every day. Obviously, in very large companies, that’s pretty much a complete waste of time.

There’s no way to start a conversation beyond idle chit-chat for people who don’t share a common background. That’s not to say chit-chat is always bad, but one hour of idle chatter every day probably doesn’t make you any new friends or useful connections. You need to spend time nurturing relationships, finding a group, and communicating with your team so you have people with whom you can discuss deeper issues.

Leading thinkers and researchers are somewhat split on what this group should look like. You could have a core group of people who are all tightly connected with each other, or you could cast your net widely and talk with people who run in very different circles. Welcome to the cohesion versus diversity debate.

Most of the time when I mention diversity, I’m not talking about it in the demographic sense. For the purposes of this book, diversity refers to your social connections; that is, do you only talk to people who talk to each other, or do you talk to people who are in very different places in the network? A diverse network looks like a star, with many different lines emanating from you at the center, while a cohesive network looks more like a web. Figures 3.1 and 3.2 provide visual representations of what these different networks could look like.

The benefits of a cohesive network are best understood by looking at basketball, which is very much a team sport. Even if teams are able to cobble together players who look good on paper, there’s no guarantee that talent will translate into actual wins on the court.

Depending on whom you ask, LeBron James is one of the five best basketball players to ever play the game. His immense size, speed, and accuracy make him one of the most dangerous players on the court. He’s just as comfortable mixing it up with seven-footers around the hoop as he is pulling up and hitting a three-pointer from deep outside. Almost single-handedly he made the Cleveland Cavaliers, a perennial loser in the NBA, into a basketball juggernaut. However, he was never able to close the deal in Cleveland and win a title, in large part because the roster never had any long-term star players besides LeBron.

Chris Bosh was one of the most coveted big men of the offseason. Bosh does not have the same all-time great status as LeBron, but he is a huge scoring and defensive threat, putting up impressive numbers in his seven NBA seasons and making the All-Star team five times. Like LeBron, he had languished on the Toronto Raptors, barely reaching the playoffs and never making a serious run at an NBA title.

At the same time that Bosh and James were entering the market, the Miami Heat were fighting to keep Dwayne Wade on their team. Wade is considered to be one of the best basketball players of all time, slightly behind LeBron in the overall mix, with a playing style that combines blinding speed and deadly accuracy. He had already won an NBA title by teaming up with Shaquille O’Neal in 2006, but had struggled to maintain playoff relevance after Shaq’s performance started to decline.

Miami wasn’t content to just hang onto Wade, however. They wanted to create a basketball powerhouse, a dynasty that would deliver a run of championships to South Beach. During the offseason, the stars had aligned where they were able to pull off the deal of the century by signing LeBron, Bosh, and Wade to long-term contracts.

The NBA makes pulling off these massive deals routinely difficult due to its salary cap structure. Each team has only a certain amount that it can spend on player salaries, with few exceptions. This required the Miami Heat to shed nearly every player they had who were not named Wade and who had a significant salary attached. To understand the depth of these cuts, of the four players who started more than half of the Heat’s games in 2009–2010, only Wade remained at the start of the next season. Including substitutes, these roster changes removed players responsible for 47% of the minutes played in 2009–2010.

So despite the fact that this was a massive acquisition of top-shelf talent, at the start of the season more than half of the Heat’s players were new to the team. This was a group that was not cohesive, one that would have to learn quickly in order to succeed.

The Big 3, as they came to be known, of LeBron, Wade, and Bosh, started the season with great fanfare. LeBron announced his signing in grandiose fashion in a primetime show on ESPN titled The Decision, proclaiming that he was “taking his talents to South Beach.” The ensuing media firestorm had LeBron defending his ego-fueled television appearance and Cleveland Cavaliers fans burning his jersey in the street. At the event where they were officially introduced to Miami fans, the Big 3 promised, “not one, not two, not three, not four, not five, not six, not seven” NBA titles, but eight.

Their first campaign for the NBA title got off to an inauspicious start. After their first 17 games, the Heat could only muster an unimpressive record of 9 wins and 8 losses. This had many calling for Coach Erik Spoelstra’s head. After all, how could a team with so much talent lose to a string of mediocre teams?

Those in basketball circles, however, were not fazed. ESPN analyst Hubie Brown stated: “Any coach realizes that when you bring six or seven new people to a team that is expected to play at a high level of execution, it takes time.”¹ The team was just not playing in sync; but this is something that develops over time. Players need to spend time in practice working with each other to be able to anticipate how they will respond in different situations. They need to understand the small signals that people pass between each other to convey their intentions without others knowing. All of these small things add up to big results on the floor.

Not surprisingly, things finally did start to gel together for the Heat. After their 9-8 start, they rattled off 12 straight wins, and they rode this success through the playoffs into the NBA finals, where they met the much older and heavy underdog Dallas Mavericks.

The Mavericks were the anti-Heat. Where the Heat were young and a team cobbled together less than a year earlier, the Mavericks were a veteran team, led by perennial all-star Dirk Nowitzki. In stark contrast to the Heat, the Mavericks team remained mostly unchanged from the previous year, bringing back 10 players who were responsible for 80% of the team’s minutes in the 2009–2010 season.

When you looked at the individual stats, however, the Mavericks shouldn’t have stood a chance. On a per-game basis, the Big 3 of James, Wade, and Bosh outscored the top three players on the Mavericks by a significant margin, and as a team the Heat scored more points and gave up fewer points than the Mavericks. Strange, then, that many pundits and media outlets, from Bleacher Report to retired NBA legend Charles Barkley, picked the Dallas Mavericks to win. Amazingly enough, in the end it was those old Dallas Mavericks who triumphed.

This is a classic case of individual talent versus team cohesion. Miami was still improving, still getting to know each other on the floor. Dallas made up for its lack of heavy star power with a team-oriented approach. The general consensus was that Miami needed to play together more, to develop team chemistry.

The consensus was right. Miami won its championship the next year. Incidentally, that season the players who returned from the 2010–2011 squad were responsible for 81% of the team’s minutes the previous year, slightly more than the former champion, the Dallas Mavericks.

A major benefit of cohesive networks is that they create high levels of trust within the group. This trust comes directly from the structure of interactions in these networks.

Let’s consider a simple cohesive network, with four people who spend the vast majority of their time talking to each other. Suppose that one of those people wants to mislead someone in the group, maybe telling that person that she wasn’t invited to an important meeting when in fact she was. In a non-cohesive network, or a dispersed network, it’s entirely possible that no one would discover the deception.

This occurs because when you want to expose a lie, you need proof. Most of the time people are not looking for proof because the assumption is that everyone more or less tells the truth. Unless you are mistakenly forwarded an e-mail or bump into the deceiver at exactly the wrong time, you’ll be entirely in the dark.

In a cohesive network, on the other hand, you receive a constant stream of information about your close contacts. People in your network are telling you about their work, what they did with other people, and so on. Because people in a cohesive network spend the majority of their time with each other, most of this information will be about your other close contacts. Similarly, you’re constantly giving out information about yourself and what other people in your network are doing. If you’re trying to lie, this means you have to constantly maintain the lie if you’re asked about it. It also means that your lie will be thoroughly spread throughout your tight-knit network. If any one of those people discovers your deception, then everyone will know.

Very quickly you would be punished by the group, either through a stern verbal rebuke or by being ostracized. This makes the potential downside of lying very high, and especially for major issues, successfully deceiving the entire group would be extremely difficult.

Not surprisingly, when you’re able to be open and trusting with a group of people, there are powerful psychological benefits. Stress in particular tends to be much lower for people who have cohesive networks. Job satisfaction also tends to be much higher for people in these groups.

The supportive effect of cohesive networks is analogous to its impact on trust. Instead of a lie, think about what happens when you tell someone in your group that you’re having a bad day. One of them could even notice that you’re feeling down. That information will quickly spread to everyone with whom you spend time. Pretty soon you’ll have people consoling you, offering to take you out to dinner, or even just giving you a break on some of your work so you can take it easy.

You might be surprised that these interactions don’t have to focus on work. In regards to trust and stress, in fact, it’s probably better if people in these networks talk about their private lives in addition to their work life. This adds depth to the relationship, further enhancing trust and decreasing stress.

Some workplaces have taboos against talking about private lives, in a bid to keep things professional, but this makes a false distinction between work and home life. Having problems at home can affect you at work, and vice versa. When people go home, they think that discussing work matters is perfectly normal, but for a variety of reasons the opposite is often not true. However, not sharing information makes it difficult for people to support and work with each other effectively. Without having access to all the information, people at work might assume that you’re lazy even though you’re exhausted because your husband is in the hospital and you’ve been taking care of the kids yourself.

Sharing good news through these cohesive networks is also easy. If your work involves passing things off to people with whom you rarely speak, you probably won’t see the fruits of your labor. This is particularly true in industries that don’t make physical things that you can point to. If you’re a programmer, for example, you might be writing one function of one program that is part of a larger system that makes up a piece of software.

By simply following the specs, you can make a passable piece of code, and if you really put your nose to the grindstone, you can produce some elegant code. After that code is integrated into the larger project, though, you probably won’t have anyone thanking you for the extra work that you did. No one knows who was responsible for that particularly effective code. However, if you are in a cohesive group that is working together on this part of the project, you’ll be personally thanked by grateful programmers who can appreciate how much work you put in.

This appreciation is another reason why job satisfaction tends to be higher for people in cohesive networks. After all, being happy at your job is much easier if you’re working with people who help support you and also care about the work that you do.

Digging into this further, cohesive groups don’t just have psychological and trust benefits; there is also a significant impact on communication effectiveness. As people in a cohesive network spend more and more time with each other, they also start to share communication shortcuts. In effect, they’re developing their own language. This is not a language in a formal sense, but one in which people share common assumptions and are familiar with the same concepts.

Developing a common language is something that people do with every group with which we communicate. In your family, when you refer to “Uncle Bob,” everyone knows whom you’re talking about. Someone who walks off the street into your house wouldn’t be able to engage in a discussion about Uncle Bob. She would lack that basic understanding.

This happens at a more complex level at work, and is particularly evident when you first join a company.

An internship I had at IBM is a good example of this phenomenon. When I first started, I had some sense of the work that was going on at the company, but I didn’t know about all of their activities. Needless to say, I was confused when someone came up to me and said the following:

“Are you on beehive? We could also use same time if you don’t want other people to know.”

If you’re an IBMer, you know what I’m talking about. Beehive is an internal social network at IBM, like an internal Facebook. Same-Time is the instant messaging program that IBM developed and uses internally. You can imagine my befuddlement when I was first confronted by this statement, but after a few weeks I was able to get some of the company lingo down. The point is that employees don’t just magically know this common language when entering a company. There’s a process of assimilating that information.

Company lingo is only one part of this common language. People also frequently refer to past events and organizational initiatives that you might not be familiar with. These events are typically unstated, so you find out about them almost by chance, or when one of your coworkers takes the time to explain each of these events to you. However, these events are so culturally embedded, so inherent to the organization, that realizing that outsiders don’t understand them is often difficult for employees.

In your everyday life, this can occur when you mix different groups of friends at a party. As the comedian Jim Gaffigan eloquently put it: “Don’t be alarmed when you hear me speaking in a British accent.” In his case, he was implying that one group of his friends knew him as British, whereas another group thought he was American. Although this is an extreme example, different groups of friends have different common events to draw on, and this can often be the cause of awkwardness when you have to constantly explain references that your friends make.

These references also take the form of common assumptions that people continually make. You might not hesitate to e-mail out a report to your whole group before your boss has a chance to take a look, but in some companies that would be frowned upon or even cause for dismissal. When talking about our work, we also make assumptions about our audience’s perspective.

For example, when I was a graduate student at MIT, the idea that sensors and computers will be integrated into everyday objects such as wallets and light bulbs was very natural. We talked about these things every day, and people had long built prototypes that showed these applications were not only feasible but also very compelling. I always found it strange when a company would visit the lab and we would have to spend an hour explaining why/how you could do this. To us at MIT, it was self-evident, but to the outside world, the assumptions that we had made weren’t clear at all.

A common language helps you predict how others are going to respond to you. Back during my time at MIT, a “deadline” was a fairly nebulous term that meant “have this completed in the few weeks around this date.” If I took this same attitude when I worked at Hitachi, however, I would have quickly been shown the door.

If you’re at a large company, having a common language is critical. Even at an institution such as MIT, I was able to walk into a meeting with people from a different department and almost immediately be on the same page. Getting this language proficiency can take weeks, even months when entering a new company. In fact, in the business world the general rule is that when you hire a new employee, you have to expect him or her to be unproductive for three months. It’s not that these new employees don’t come in with the necessary skills; it’s that they lack the language to communicate with others.

This same problem can pop up on a project where you’re collaborating with people from different parts of the organization. For a project to run smoothly, you have to get up and running quickly. If you have to spend the first three months of a six-month project working through communication issues, you’re going to be in trouble.

One of the problems with the shared context people develop is that the underlying assumptions might be wrong. Research in Motion’s (RIM—the company that makes the BlackBerry smartphone) declining fortunes are a vivid example of this. Even with the wild success of the first iPhone in 2007, RIM clung to the flawed assumptions that everyone wants physical keyboards, that cellphone apps were a “fad.” These assumptions had been built over years of proven success. Everyone at RIM knew that the form factor of their phones was second to none. Everyone at RIM knew that you had to build better and faster hardware, but you could throw in the software and app ecosystem as an afterthought. Everyone at RIM was wrong.

These assumptions caused RIM’s smartphone market share to plummet from a dominant 43% to an abysmal 12%. Why wasn’t RIM able to quickly pivot and churn out an iPhone killer within a few months? After all, RIM had been making cutting-edge phones for years. Its technological innovations in software and hardware were unparalleled in the industry. The underlying issue was that RIM never questioned these basic assumptions, never allowed their context to evolve. With no data on which to base our assertions, people and companies are much more likely to go with their instinct, with what feels natural. At best, this leads to mixed results.

Enter the flip side of cohesive networks. They’re also bad at a lot of things, especially when taken to extremes. As shown in the BlackBerry example, when you’re in a closed-off network, discovering new information is incredibly difficult. Cohesive groups are also poor at influencing others. Because they’re very inwardly focused, reaching out to different stakeholders to affect substantial change is difficult.

Diverse networks, on the other hand, are good at precisely the things that cohesive networks are not. They’re structured to help us break out of old habits and change our perspective.

Breaking out is really about doing new things and meeting new people, and the benefits of that approach have been well documented. So, if breaking out is mostly positive and cohesion is mostly positive, which is the right approach? The simple answer, although many researchers would have you believe differently, is that it isn’t all or nothing. You can have a cohesive group that you spend most of your time with and have an extended network that you’ll tap into occasionally to get new information.

Cohesiveness versus diversity is one of the most hotly debated subjects in social science. Without going into the nitty-gritty details, this debate kicked off in earnest in the 1970s with Mark Granovetter’s seminal work, The Strength of Weak Ties. This paper showed that when you were looking for a job, the weak ties, or the people that you don’t talk to very often, were the most important relationships to have. The more weak ties you had, the easier it was to find a job.

Later on this research started to bleed into the study of organizations, with people touting the benefits of weak ties even within companies. Other researchers, such as David Krackhardt, fired back with their own research, showing that in many cases weak ties in fact led to poorer performance. Krackhardt studied a firm that sold computer systems to business clients. He asked employees to fill out surveys about their different networks: who their friends were, who they went to for advice, and so on. It turned out that people who had very cohesive networks, especially networks where people were both friends and were relied on for advice, had much higher performance than those with weak ties.

The debate is by no means settled, but the pros and cons on both sides mean that people need a way to understand how the balance should shift in different companies. Different circumstances call for different patterns of interaction, but precisely defining when and how to shift the balance isn’t possible with surveys. It is possible, however, with Sociometric Badges. These badges can shed light on exactly how organizational initiatives, such as the all-important purchase of water coolers, play into these different types of networks.

The same cannot be said for blue-collar industries. Their mentality hasn’t changed much from the days of the industrial revolution, with a focus squarely on efficiency and time management. Efficiency is very different from productivity. Conceptually, increasing productivity by 5% increases the size of the whole pie by 5%. When increasing efficiency by 5%, the size of the pie stays the same, but the slice of the pie devoted to worker pay shrinks by 5%. This increase is often much less valuable than increasing the size of the pie.

The reasoning for a focus on efficiency is that there is an assumption that people in certain jobs can’t really be more productive. Packing workers are a prime example of this. Let’s assume people are stuffing boxes as fast as they can, and on an average day an average worker can churn out 100 packed boxes. Let’s say I discover a new way for people to pack boxes, and now the average worker can pack 105 boxes in a day. If the company only has to pack 105,000 boxes a day, then instead of 1,050 workers, the company would only need 1,000 workers to pack boxes.

Call centers work in a similar way. In a modern call center, often a few thousand employees sit in one huge room answering calls from customers. The company wants employees to answer the greatest number of calls in the least amount of time possible. If it takes you less time to answer a call, then the company can hire fewer staff and become more efficient in general.

So how do you figure out how to answer calls more quickly? For a long time a lot of companies tried to solve this from the top down. Executives and managers would listen in on calls and work out strategies that could be disseminated to the rest of the organization. This method was okay, except that sifting through the millions of calls employees were making to discover successful strategies was often very hard. After all, you can’t just organize calls by completion time. A problem that is inherently hard will take longer to complete than a call that involves an easier-to-solve issue. So, by trial and error, people sift through this data and sometimes come out with valuable insight. Other times they’re left empty-handed.

As this practice indicates, call center management has not changed much since the 1960s. Call centers back then were organized a lot like small factories. They had about 100 people on the phones organized into teams of about 20 people, normally based on different specialties. When one person on a team was on a break, either for lunch or for a coffee break, no one else could be on a break. The reason was fairly straightforward: If 20 people went on a break at lunch, keeping up with demand would be impossible.

Fast forward to today. Now with thousands of employees, modern call centers no longer necessarily have to choose between call load and team break times. However, that’s the way things have been done for decades, and there’s no real incentive to change—or so everyone has been led to believe.

Working in such a call center is stressful—monumentally so. Your entire day is people yelling at you for stuff that’s not your fault. When you finally go on break, no one you know is taking their break. There’s just a total lack of social support. It’s no wonder that turnover in call centers is 40% per year.

Turnover isn’t just a problem for the people who leave and have to find another job. It is psychologically draining to the employees who remain, because seeing your colleagues burn out and quit leaves you with one less person to talk to, one less person to go to for advice. Monetary effects follow as well. Every time a veteran employee leaves, the call center needs to spend months getting a new person up to speed. Not just on formal procedures such as how to answer the phone and how to use the computer system, but also acclimatizing the person to the culture of the organization (see Chapter 2). So not only is working at a call center psychologically difficult, but employees constantly head for the door, deflating morale and adding enormous cost to the company. Typically, companies spend 25% of a veteran’s yearly salary to hire and train a replacement.² All of a sudden, the toll this environment takes on workers has real economic significance.

One reason we know how much this costs down to the dollar is that call centers are some of the most quantified organizations on the planet. Call center managers measure how quickly people complete calls, how many times they put people on hold, how many mouse clicks employees use during a call, and they even record call content to analyze what went right or wrong. Employee breaks are similarly measured and planned. Precise breaks are allocated to individual employees to ensure maximum uptime on the phones while complying with federal work standards.

The issue of breaks is interesting because in the modern company they have traditionally been viewed with disdain or at least passive disapproval. In many companies, always looking busy is important. Not surprisingly, taking a break to schmooze by the coffee machine or water cooler can lead to negative assessments by coworkers. The common impression is that even the appearance of talking to a colleague about non–work-related activities means that you don’t care about your job and must not be working hard enough. This perception turns the coffee area into a barren wasteland, perfectly clean, perfectly stocked, and completely free of socialization.

I’ve been to many companies where eating lunch at your desk was standard practice. Rather than communicate with coworkers, people felt more comfortable surfing the web and looking at cat videos on YouTube while slurping last night’s leftovers. Not that looking at cat videos isn’t hypnotic and fun once in a while, but I haven’t yet seen a study that proves that extended cat video viewing is correlated with higher productivity.

Lunch is one of the most important times of the day, not only to physically recharge our batteries but also to take some time to network and communicate with others. This idea has been explored in depth in books such as Never Eat Lunch Alone by Keith Ferrazzi, which details how people who eat lunch with others advance faster in their careers and perform better in general.

Although to some having lunch with others might be viewed as an additional burden and detract from a much-needed physical break, this strategy ignores a perfect socialization opportunity. For those of us who simply can’t make it through the day without reading a bit of trashy news or something else distracting on the web, taking that break separately from lunch is best. You can then preserve your opportunity for socialization but also take some time for yourself.

Both of these activities are part of your work. Communicating with others is work, as is resting your body. The benefits of both have been tangibly demonstrated time and time again, although mostly in the blue-collar workforce. Even at the height of the industrial revolution, when factory workers could be thought of quite literally as cogs in a machine, they still took breaks. Even Taylorist (see Chapter 2) managers who maintained strict work plans and provided only a few minutes each day for breaks realized that workers were much more productive if they had time to eat and take care of biological necessities. This cold, hard reality forced the hands of these calculating factory owners, and this influence is still felt today.

An interesting study on the benefits of breaks took place in a meat packing plant in the U.S. Initially, this factory implemented standard lunchtime and bathroom breaks for its employees. Like call centers, breaks were staggered so that no one on a team would be on a break at the same time. Unfortunately, turnover and fatigue concerns plagued this plant. Researchers showed that creating breaks that were long enough to provide for cohesive interactions significantly reduced employee stress and might help stem the tide of defections and re-energize the workforce.³ Other research on factory rest periods devised a cryptic name for this kind of break: “Banana Time.”

The exchange of work-related information is a second direct benefit of breaks. Beyond the physical respite a break provides, it can also create a platform for ideas. People who have the same job but otherwise don’t communicate have the opportunity to compare notes at a high level. Technical terms that might be missed by those not on the front line can be freely exchanged during these breaks. No translation is needed because these people represent a cohesive group of workers.

This type of interaction is fundamentally different from a meeting, which filters ideas through management’s idealized view of work. Much like the Toyota production system provided through formal channels, breaks create an opportunity for front-line employee feedback to spread through informal channels. In the original Banana Time study,⁴ information sharing during breaks allowed new techniques to percolate through the worker network until these practices were recognized at the company level. These bottom-up innovations were then integrated into training programs and the standard processes of the company.

A complementary third benefit of breaks addresses the mental fatigue associated with these jobs. Previously, workers did not have the opportunity to vent to coworkers or socially support their colleagues. They were alone on the packing floor, grinding away for hour after hour, alone with their thoughts. This by itself can be quite taxing, because staring at raw meat for an entire day can do things to one’s appetite, not to mention one’s state of mind. Throw in personal problems at home or other things eating away at you, and one can easily see the strong negative impact on productivity and mental health.

On a break these topics are open for discussion. Rather than have to bear all of this stress and physical toil alone, people can complain to coworkers and vent a little steam, or bring up personal issues that have been weighing on their mind and ask for advice. The alternative is to let these problems fester, to let stress at work and at home build up until it’s unbearable. Like a pressure cooker left on the stove, it’s only a matter of time before this situation explodes, causing people to quit and remove themselves from the whole situation.

These personal discussions go hand in hand with work-related conversations. One is not necessarily more important than the other, but both types of conversations have tangible benefits. Work-related discussions transfer relevant information between employees and can lead to new innovations. Social conversations create trust, build rapport, and relieve stress.

At different times one or the other type of conversation can be more effective. If deadlines are looming and there’s pressure to deliver results, sharing work-related information is probably more helpful. If workloads are piling up or it’s a particularly stressful time, then social conversations will be more useful. By mixing and matching these conversations appropriately, people can achieve powerful results.

This makes it all the more puzzling that in spite of these benefits many companies have been pushing to reduce breaks. This is especially true in blue-collar work, where companies stick to government-mandated minimums and make sure employees know that they’re on the clock whenever they’re not doing physical labor. This mirrors the trend in white-collar work discussed earlier, but with the much higher probability of burnout in physically demanding, high-stress roles.

Interestingly, in call centers in particular, some people view turnover as a feature rather than a bug. This view treats people as replaceable cogs in the call center machine and argues that performance degrades over time, and therefore companies shouldn’t worry about people burning out. This perspective was laid out in a study by Catriona Wallace from the University of New South Wales,⁵ which showed that at one particular call center, productivity was negatively associated with tenure.

Without getting too much into the specifics, the idea is that people work less effectively as their stress increases. As already discussed, call centers are inherently stressful workplaces. Rather than attempt to devise policies to reduce this stress, under this “burn-out” strategy companies merely need to calculate when the loss in productivity due to stress becomes greater than the cost to train a new employee. At that point, it’s time to either get rid of that employee or crank up his workload to make it very likely that he’ll quit.

This calculation makes a number of assumptions about call centers and businesses in general that seem suspect. First think about how a call center employee would react to being treated this way by an organization. He would immediately realize that there isn’t a future for him in this company beyond a year of employment. Although that’s not necessarily something to thumb one’s nose at, it certainly affects how an employee will interact with a customer. Without the prospect of a continuing career or advancement, why would anyone go the extra mile for a customer? Simply keeping his head down and trying to deal with a difficult call by minimizing the time that he spends on it is much better than actually trying to solve the problem. This tactic decreases the time he spends on the phone, but it would have a huge negative impact on the customer experience and would very quickly start eroding a company’s customer base.

On top of this reduction in customer satisfaction, there would be far-reaching effects on morale. When everyone an employee knows is being chewed out and treated poorly by the company, he will start to think pretty quickly about changing jobs. Eventually this toxic environment would lead to an acceleration in turnover, forcing the company to rotate in new employees faster and further increasing costs. Wallace assumes that these employees are low-skilled enough to not have many job prospects, implying that they would stick it out at a job they hate rather than look for other employment. This premise might be right up to a point, but at the very least, people would start looking for other positions almost immediately. Because a fraction of these people would be able to find at least some work, turnover is bound to increase.

When these employees walk out the door, not only do they take with them their individual contributions, but also the tacit knowledge that they acquired about how to do their job effectively. These are the little tricks about work that make things just a bit easier, the pieces of knowledge that slowly spread across a workplace. These tricks could be something as simple as how to circumvent a laborious part of a software program or as nuanced as how to deal with enraged customers screaming at the other end of a phone line. If employees are turned over too quickly, not only do they have less time to uncover these tricks, but a much shorter window exists in which they can share this acquired knowledge. Again, this turnover would exert strong downward pressure on performance across the whole call center operation.

Lastly, and very significantly, turning over employees quickly means that the company sheds any potential future leaders before they even get started moving up the ladder. CEOs from companies as storied as McDonald’s, Goldman Sachs, and General Electric started in entry-level positions, rising through the ranks to become the captain of the ship. So not only are companies throwing away organizational performance in the mid-term, but in the long term as well.

This evidence is something that call center employees in particular have been searching for. How could they demonstrate to their employers that they should be able to take breaks with other people? Most of these workers understand all too well the stress of their jobs, and realize that if they don’t have tools to deal with it effectively, they will end up burned out and be forced to quit. The conditions were ripe for a project with the Sociometric Badges.

My MIT colleagues and I were approached by Bank of America (BoA) to study precisely this problem of burnout and call center performance. BoA had an interesting issue related to call centers. For some background, note that this company has one of the largest financial call center operations in the country, with thousands of employees stationed at call centers across the United States at all hours of the day.

As in other companies, BoA standardized its call center operation. Its call center in Rhode Island had roughly the same org structure as one in California, the same IT systems, and the same training programs. Employee demographics were also quite similar, with most having high school diplomas and a few with college degrees. Everything that could formally be put into place was the same—and yet, performance was different. Despite all of the similarities between employees in different locations, something about these call centers couldn’t be quantified with other methods.

The one possible cause of these differences in performance was culture. Some of these call centers must have different collaboration styles, different cultures, that cause them to be more or less productive. However, there isn’t a general understanding of what the term culture actually means quantitatively.

The study took place at one of Bank of America’s call centers. Our goal was to measure how people interacted and behaved to understand what was making people successful. These call centers typically have thousands of individuals, so instead of looking at everyone, we focused on a few of these teams. By concentrating on the differences across these groups in depth, we hoped to derive general lessons about productivity that could be disseminated across the company.

Notice that I haven’t mentioned anything about breaks in this study. From Bank of America’s perspective, breaks were not necessarily a consideration. Remember that call centers have been managed one way for more than half a century, and there wasn’t a feeling that changes in this age-old break structure would bring any lasting effects. However, BoA knew that something with this traditional model was amiss.

To study these teams in detail, we collected not only badge data, but also performance metrics, demographic information, survey data, and e-mail records.

The performance metrics were relatively simple. Essentially, an employee’s performance boils down to how quickly she completes calls on average. In the past some clever people had figured out that they could improve their metrics by hanging up on a customer right at the beginning (or middle) of a call. This would look like a short call from the company’s perspective, and so would improve your overall performance numbers. These calls were naturally removed from any calculation of performance for the purposes of the study.

The demographic information we collected was also fairly basic: tenure at the company, gender, age, and a few other features. This information was collected mainly to identify whether any aspect or combination of aspects of particular individuals might lead them to behave in a certain way. For example, you might expect that someone who had been at BoA for a long period of time would have a more tightly knit group of friends than someone who had been there for only a year.

These differences could be further explored with the survey data. Each year employees were surveyed by Bank of America about their stress levels, communication with managers and coworkers, and their overall perception of the company. Stress levels are particularly important at call centers because high stress is a precursor to turnover. If the bank noticed a particular team had higher stress levels than others, it would try to investigate the causes further. For our purposes, this stress data enabled us to investigate behavioral patterns that help mitigate these effects.

In addition to all of this data, we had a few other critical pieces of information. We knew when people took their breaks and when they got lunch. This was important because it enabled us to see what was actually going on during those break periods. Who were people talking to? What were they doing? By combining these different data sources, we were able to get to the heart of the problem: What is it that makes people effective?

The groups studied were located in different areas of the call center, which itself was basically a huge single room with rows of cubicles for the roughly 3,000 employees that manned the phones. Some variation existed in this physical layout, however. One of the groups had cubicle walls that were below eye level and desks that were about one meter longer than other groups. The other three groups all had high cubicle walls that completely blocked their view of the other people on their team.

When we introduced the badges to participants, management had some trepidation as to what the reaction of employees would be. In particular, the potential perception of the badges as a “big brother” intervention was a concern. After presenting the study plan and the technology, however, we got an extremely positive response. For years employees had tried to convince their managers that their interactions with other workers were important, but they had been unable to tangibly demonstrate the benefits. They felt that this was a huge opportunity for them to show that value.

Bank of America was also focused on showing value. Although they were interested in understanding what unseen aspects of call center culture were responsible for performance differences, the company was more interested in improving them. They wanted to identify specific changes they could make to the way that their people worked and measure the effects of those changes.

This study took place in three phases: The first phase consisted of the initial measurement of the call center teams. After analyzing the data, we would propose and implement changes in work processes at the call center. The second phase was a normalization period of three months, where we would wait for the changes we implemented to become part of the regular process. The last phase was a re-measurement, where we would precisely gauge the behavioral and productivity effects of these changes.

You might ask why the second phase was needed. After all, the impact of a substantive change in the way people work should manifest itself almost immediately after implementation. Everyone has experienced this type of effect at one time or another. A good example is the transition from elementary school to middle school. Although most students aren’t able to adjust to this transition immediately, after a few months they settle back into a state of normalcy.

Although behavioral dynamics such as movement and interaction patterns do tend to settle down, other changes can easily upset these dynamics. A major cause for disturbance happens when people are observed. This is experienced by many of us when we take our first driving test. The tester gets in the car with you, and all of a sudden your awareness changes. You might get nervous or suddenly forget months of parallel parking practice, and all because you’re being observed.

This phenomenon is known as the Hawthorne effect. This effect was formally identified in the early 1930s by researchers studying the Hawthorne Works factory in Cicero, Illinois.⁶ These researchers wanted to understand the effect of light levels on employee performance. They designed an elegant study where they would subtly change lighting levels each day and gauge the impact. At first they made the factory floor brighter, observing that performance went up very significantly. Next they turned the lights down, expecting to see a drop in productivity. Instead, they saw the exact same uptick.

Researchers were initially puzzled over these results. Then they hit upon the truth: The workers knew they were being observed, so they endeavored to work harder no matter the change.

In our Bank of America call center study, the three-month break period was put in place to avoid placing emphasis on any initial reaction the employees would have. Three months later, any change we made would have become normal practice. This would enable us to measure the actual effect of our intervention.

With this structure in place, we had only one small hurdle in front of us: deploying the badges.

With that in mind we visited this call center in the suburbs of a mid-size New England city. Standing up front were the badge team from MIT: myself, Taemie Kim, and Daniel Olguin. We were quite the international team, with each of us from a different country—the U.S, South Korea, and Mexico, respectively. People filed into the meeting room where we would present our plan for this project, and their eyes strayed curiously to the small box hanging around our necks.

We explained how this project would work. In a few weeks, we would give everyone who chose to participate a Sociometric Badge, but anyone who didn’t want to participate could wear a fake badge that wouldn’t collect any data. As far as what data the badge collected, we described the different sensors and their functionality in detail. They could also find a description on the consent forms that we handed out, required by MIT’s internal review board to ensure that people understand what they’re agreeing to. No individual data would be shown to managers, and their names wouldn’t be directly tied to the data collected on the badges.

When we returned, we had to set up the office for data collection. The wearable badge itself is quite easy to use. You simply flip the power switch and hang the badge around your neck. At night, you take off the badge, flip the power off, and plug it into a USB charger. To recognize location, however, we had to put base stations up around the office.

A base station is essentially a wearable badge attached to the wall. Every 10 seconds the base station sends out a ping over Bluetooth. When a wearable badge receives that ping, it can estimate the distance from the base station. By receiving pings from multiple base stations, we can triangulate the signal to figure out where someone is to within about one meter. In order to make these calculations, a researcher has to take a wearable badge and stand in different locations for a few minutes, measuring changes in the signal strength from the different base stations. This is essentially the same way that phone navigation programs work, using Wi-Fi access points instead of Bluetooth.

With the call center thoroughly badged up and prepared, the data collection could begin.

Examining the e-mail data, we saw something that at first was almost too perfect to believe. When we plotted the network of e-mail communication between participants, what we saw was almost link-for-link a mirror image of the org chart. There was practically no communication between peers who were on the phones with customers, or for that matter even team managers. All the real communication, if it was happening at all, was happening face to face.

This observation probably shouldn’t have been too surprising. E-mail is good for communicating rote information, and so what we were observing was top-down dissemination of simple directives to the front-line employees. To exchange tips on how to deal with customers or to vent about difficult calls, e-mail isn’t the appropriate communication medium.

Obviously, this result has implications for the future of call centers. Many companies, including BoA, have placed large bets on a distributed call center workforce. Buying desks and computers for employees, setting up a secure connection at their house, and letting them work from home is easy enough. With no commute for employees and no expensive office space for the company, this would seem at its face to be an ideal solution. The results from the study, however, indicate that if there is any value in communication between employees, the IT tools that are typically available to call center workers are woefully inadequate.

So what about face-to-face interaction? This was, relatively, a much more active communication channel. Employees on the phones interacted with three other employees on average (actually 3.06), and these were almost exclusively other people on their team.

This data became more interesting when we mashed it up with performance and stress data. As far as general statistics are concerned, on average it took employees 263 seconds to resolve a customer call, only a shade over four minutes. Given my personal experience (and frustration) contacting call centers, this seemed exemplary, but the picture was not all rosy. Employees were under a moderate amount of stress, with the average stress rating coming out to 3.07 on a scale from 1 to 5. This might not sound too bad, until you realize that this is their level of stress all the time. We expected a result like this given the high turnover typical in call centers, but these numbers point to the substantial problems affecting these workers.

We then got around to testing our hypothesis and correlating these different data sources. Given the numerous arguments listed previously, we expected cohesion to be positively related to productivity and to be associated with reduced stress. Our hypothesis was not only confirmed, it found that cohesion was far and away the single most important factor in regards to productivity and stress.

This point is difficult to understate. To get a sense for the magnitude of the importance of cohesion in worker productivity, cohesion was about 30 times more important than experience. Put another way, having a network that’s 10% more cohesive is equivalent to having an additional 30 years under your belt at a call center.

However, these positive benefits weren’t confined to productivity alone. Cohesion was strongly related to lower stress levels, albeit not to the degree that we observed with productivity. That being said, high cohesion was responsible for reducing stress by around 6%. Clearly, it is a useful weapon in the war against burnout and mental fatigue.

We examined other factors as well: total amount of face-to-face interaction, network centrality, degree, and so on. However, none of these features were significantly predictive. This seems to be a story about the positive effects of cohesion. The question was: Where did this cohesion come from and how can BoA increase it?

We had to devise a strategy to distinguish interactions that increased cohesion from those that decreased it. Essentially, our program went over each interaction and measured what the effect of removing that interaction from the network would have on the overall cohesion level. We then overlaid these interactions with location and time information to see what places were hotbeds of cohesive interaction. By looking at this “heat map” of interactions and varying it over time, we were able to investigate what activities were actually leading to these important conversations.

The results couldn’t have been clearer. Neither formal meetings nor people chatting at their desks encouraged higher cohesion. The vast majority of these interactions were happening away from the desks, during the brief periods of overlap between the lunch breaks of employees on the same team.

This was completely counter to the management dogma on call center operations. The story had continuously been one of efficiency, of aligning work schedules, of reducing interaction. But we had brought hard, objective data to this problem, and had uncovered the harsh reality that people were just doing it wrong. To run the best call center organization, a company needed to encourage cohesion, and to do that it had to align breaks.

After the results of this first phase were shown to managers, they were floored. Mostly they lamented all that the company had lost by managing these groups in the traditional style for so long without checking their assumptions. Still, we needed to test whether or not this was a causal relationship. We needed to move into the second phase, giving people on teams breaks at the same time and observing the results.

From the bank’s perspective, this intervention was free. It wasn’t giving employees more breaks, just shifting when they took a break. Because the company has thousands of other call center employees, shifting the load to other teams was fairly straightforward. This made it easier to convince BoA that aligning breaks was the right thing to do. Now all we had to do was wait three months to see the effects.

Few personnel changes occurred in the intervening period. Some people had left BoA, but overall the group remained more or less the same. We did note that the groups we had studied had low turnover. Over a three-month period, only about 3% of people had left the company, implying that the turnover rate on a yearly basis would be 12%, drastically lower than the 40% industry average.

This result was encouraging, and lent us confidence as we handed out badges for the final phase of this project. We were going to collect data for another four weeks, and at the end of it, we would have a definitive answer. Do breaks really make people more effective?

The results were clear-cut. Cohesion was up by 18% in the third phase, an extremely significant result. This is like adding 50 years of experience to each employee, a truly astronomical number. These teams had become far more cohesive than they were before, but at a certain level this result isn’t surprising. We had set up the environment so that this was almost bound to happen.

The impact on the bottom line was equally powerful. The performance increases associated with this intervention would conservatively yield $15 million in yearly savings on call center costs across BoA. Changing how people spent 15 minutes of their day yielded $15 million.

This result is truly astonishing. When we talk about performance gains of this magnitude, increasing performance by double-digit percentages, the normal reaction from companies is “Wow, we have to change everything to achieve that kind of growth.” What these results show, however, is that companies don’t need to think about these massive changes to achieve huge gains. If instead they can find the social levers that people are responsive to, and act on them in the right way, they are going to get big results.

Overall, the study showed definitively that breaks matter as a way to increase cohesion. Not only do they matter, but they’re also crucial to the effectiveness of companies in general and call centers in particular. Rather than thinking of information exchange and social support as something that only matters in the most creative and high-level jobs, this study shows that it matters even in organizations as seemingly straightforward as call centers.

So what’s the most important place in the office? It’s not your desk or the CEO’s office or a meeting room. It’s the simple, inconspicuous water cooler.