The controversy surrounding Yahoo’s recent decision to bring employees working from home back into the office has sparked a lot of healthy debate, which is always a good thing. One of the more positive outcomes is that it has prompted a serious discussion that specifically addresses the impact and prudence of telecommuting by IT professionals.
One informed voice on the matter is Ben Waber, president and CEO of Sociometric Solutions, a visiting scientist at the MIT Media Lab, and author of the forthcoming book, “People Analytics: How Social Sensing Technology Will Transform Business and What It Tells Us about the Future of Work.” In short, in Waber’s view, telecommuting by IT pros is a bad idea.
Using Yahoo as an example, Waber indicated that the complexity of projects typically undertaken by IT professionals screams for co-location:
If you talk about Yahoo, they’re building software, they’re building technological solutions which are really, really complex. The idea—and something we’ve also collected data on—is that when you work with people on these complex projects, if you’re a programmer, for example, your code doesn’t just depend on you. It typically depends on the code of dozens of other people. The issue is that these dependencies are really complex, and sometimes really subtle. You might not even know that some of them are there. In one case, we looked at software teams where some people were working in the office and some people were working from home and from other locations, and we looked at how likely those people were to communicate about those dependencies if they were in the same place or if they were remote. We were scraping electronic communications data to get at this. Technologically, nothing stops you from communicating with people who have these dependencies. But when you were remote, you were about 8 percent less likely to have those conversations. Importantly, when you don’t communicate with someone who has a dependency on your code, the probability of a bug popping up is about 12 times higher—so there’s a really huge effect. And in general, having these gaps means that it takes you about 32 percent longer to complete code. When you’re talking about a company like Yahoo, you’re looking at, conservatively, hundreds of millions of dollars in lost productivity.
To expand on the point, Waber cited the case of an IT services firm that was configuring very complex, multi-million-dollar server systems:
They were configuring the servers based on really rough specifications—the customer really didn’t know what they wanted. They’re paid on an individual basis—they’re in a queue, and when they get to the front of the queue, they’re assigned the task, whatever it is. If it’s an easy task, it can take you five minutes; if it’s a harder task, it can take you up to eight hours. What’s interesting is people were paid based on their throughput—how many of these tasks they could turn around. That’s what determined their bonus.
They thought something weird was going on, because when they looked at people’s skills on paper, and how much training and experience people had, the people with the most experience weren’t necessarily the highest performers. So they had a feeling that there was something about the way people were interacting at the office that was one of the biggest predictors of their performance.
We went in with our sensor badges—we have these socio-metric badges that are sort of replacements for company ID badges—and we measured their interactions and related that to how quickly people were completing these tasks. First of all, it was extremely interesting that if we looked at all the individual metrics in terms of education, tenure at the company, all these things, all of those were orders of magnitude less predictive of how productive they were in their face-to-face interactions. It turned out that people who had these really tight-knit face-to-face groups—people that talked a lot to each other—these people were completing tasks very quickly. If they spend 10 percent more time talking with their core group of contacts, they would actually clear about $100,000 more in revenue for the company over a month—a decent-size effect.
We also looked at what people did when they were working on a task. The idea was that in having this tight-knit face-to-face group, in general, they were all competing with each other on productivity metrics, but they knew and trusted each other, and it was a lot easier to exchange tips—they spoke the same language because they talked to the same people all the time. But then when we looked at the individual tasks, we wondered, just in looking at who they were talking to in doing their tasks, could we figure out how long it would take them to complete it?
What we saw was that essentially, there was this network of experts where there were these four guys, and whenever you talked to one of these four guys, it took about one-third the time that you would expect to complete the task. The reason was these guys were the experts, and when you talked to them you’d complete the task very quickly. It makes a lot of sense. What’s interesting was that we removed the face-to-face interaction—you didn’t talk to these guys, you emailed or IMed them—these are very complex discussions, and it’s difficult to type stuff out that quickly. These guys had the same job title as everybody else, so you learned about them from being in the workplace and from talking to people at lunch or over coffee. If you looked at how productive those experts were, they were actually middle of the road because they were spending so much of their time talking with other people and helping them out. So the company changed the way they paid people to reward people who were spending time giving advice.
The idea is that this face-to-face interaction, and being in the same office, was such a huge part of that division. If all of these guys had worked from home, or even if just the experts had worked from home, it would have crippled the performance of the entire division. All of this is to say that flexibility is good and is important, but there are all these unplanned interactions. You’re working on a task, and there’s something you don’t know and you wonder who you should talk to about it, you bump into somebody and start to learn what his skills are—that knowledge is invaluable to do your job, and to help other people do their jobs. Digital communication still has a long way to go to support that kind of activity.
And therein lies the irony that a lot of people have cited in the case of Yahoo: an Internet company dissing the viability of virtual collaboration. I mentioned that to Waber, and he provided a helpful perspective:
It’s interesting—[Yahoo CEO] Marissa Mayer comes from Google, and I know a lot of the guys at Google. Google’s HR division is called “People Analytics,” and they do great stuff. If you think about how Google deals with this, they’re building these gigantic campuses all over the world so that they can have thousands of employees who can be together at the same time. When you think about it, if there’s one company that should be able to have people telecommute all the time, it’s Google—this is what they do. And yet they still spend hundreds of millions of dollars creating these incredibly centralized campuses so that they can have everybody together. Because they understand that, first of all, the coordination on just executing projects is extremely important. But then you think about all the other things: How do you create new ideas? If you bump into someone in the hallway who’s working on a completely different project, those interactions are extremely valuable, because when you talk to a guy who’s doing something that’s completely unrelated to you, that’s when you get that new idea. But those are unplanned interactions. Electronic communication is pretty good at planned interactions—you can set up a time to talk on the phone or over Skype or whatever—but once I’m remote, I don’t even know who I should be talking to, beyond my formal requirements. That’s the issue that a lot of these successful companies understand.