Link to podcast episode: EL 41: Using People Analytics to Improve Productivity & Happiness | with Ben Waber
JESSE LAHEY: Welcome Leaders! Google recently moved into the number three position of the most valuable firms in the world. Google is said to have the only truly data driven HR function. They call it People Analytics. One of the data driven techniques they use is Social Sensing Technology. This is a technology that is showing conclusive evidence that slight changes in behavior from changing when you take breaks to changing what lunch tables you sit at can make you and your team happier, healthier, and more productive. To help us learn some of these lessons, our guest is Ben Waber, author of the book People Analytics: How Social Sensing Technology will transform Business and What it Tells us about the Future of Work. Ben is a visiting scientist form the MIT Media Lab and he is the President and CEO of Socio-Metric Solutions, a social management firm that uses social sensing technology. Before I jump into the interview, I want to ask for your patience with a microphone issue we have. You will hear that the microphone that Ben is wearing is doing some scratching and bumping noises form time to time. We did our best to adjust that but you can still hear it occasionally. Ben, before we jump into things we can learn from this specific technology, will you help us understand the broader field of people analytics?
BEN WABER: So as I talk about it in the book, people analytics is, you know, it is not a new thing. People analytics is fundamentally about trying to use data to understand the people that you work with and to understand the environment around you. The data we used to use for people analytics originally was just pure human observation, you know, you would see things or you would see people behaving in certain ways or interacting in certain ways and based on those observations you would change your behavior, change the way your organization ran and those sort of things. Then we started to get a little more sophisticated. We started to use surveys. We started to ask people specific questions about what was going on and based on those answers we would change the way, again, the way the company worked. It has been very, very recently that this area has undergone a really fundamental change and it starts, really, with the advent of e-mail. What’s interesting is you think of e-mail and you think of all of the digital tools we have at our disposal as tools and you think of e-mail as a way to communicate. You use a computer to write reports and do other things, but what they are also doing is generating a lot of data. They are generating data on who actually communicates with whom, when they do it, how they do it. You have instant messaging data, phone call data. All of a sudden you have a digital record, an objective record, of what is going on in the digital world and over the last couple of decades people have started to mine that data to come up with really interesting ways to think about ways to manage and organize people. If you look at companies like Google and Google’s HR department, which isn’t called HR, it’s called People Analytics, and they do a lot of this. They do a ton of internal e-mail mining, internal instant messaging mining, to look at how are people collaborating in the workplace? The important part of this is that it is missing a huge part of the puzzle, which is what is actually happening in the real world.
JESSE: Right, because if I send you an e-mail and we have a conversation through e-mail , that is missing all the parts of my personality and your personality that are going on around that and I probably have, maybe, more important relationships that have nothing to do with e-mail.
BEN: And that is sort of the point. It is not that, well first of all, you use e-mail for certain things and you use face to face for certain things. So if you want to talk about something more personal, maybe you want to talk about your kid’s birthday this past weekend, right? You are probably not going to send an e-mail having a conversation around that. Maybe you will do it over IM but odds are you are going to do that face to face. You also do this very complicated stuff face to face. If you want to have a conversation around what the company’s R&D budget should be for this next year, you could do that over an e-mail but you would spend hours composing the e-mail, thinking about it and the other person would have to spend similar amount of time thinking about it then writing a reply and the whole discussion would take weeks before it would get resolved. Whereas most people have those kinds of conversations face to face because we can, again, in real time respond to subtle cues such as facial expressions, tone of voice, you know, and if you look at big decisions and big conversations that companies have, the vast majority of them happen face to face. You don’t really see many corporate mergers happen over e-mail. You will get the CEOs come together, they will have meetings, they will go out for beers afterwards and those are the interactions. Again, if you look at the data that are most important for driving productivity, job satisfaction, creativity, and all the things we really care about.
JESSE: So what do you do about that? You are not going to, necessarily, record all those conversations and then try to analyze what people are saying and talking about. So what can you do?
BEN: What is interesting is we already have a lot of centers at our disposal that are actually measuring this physical stuff that we are talking about. So when you think about it, when you go into work there are a few sensors that you have on you all of the time. The first one is your cell phone which is in your pocket. It might be your personal cell phone but your cell phone does have a lot of sensors that can tell you who is around you and how you are moving around. It has the microphone. Because it is in your pocket, it can only do limited things in terms of determining who you are actually talking to. Over a three month period I can actually figure out who your friends are but, again, that is not very predictive of how productive you are, how happy you are. I really need to know who you are talking to for that. We have this other sensor that we wear all the time and that is actually your company ID card. Most of the company ID cards today you use them to tap into a door. They have a little ID chip in it that is a radio and is a senor. If you put little RFID censors in the ceiling, you could actually figure out where people are. Besides being creepy, this tells us a lot about how productive people are and how happy people are. Again, you need to know who is talking to who.
What we have done is essentially created the next generation ID badge, which have a number of sensors in it, which again, they don’t record what you say and they don’t have your name attached to the data, but they are looking at who talks to ho and how people talk to each other, looking at your tone of voice, your volume modulation, how quickly you speak. It turns out those things are extremely predictive of outcomes we care about in terms of how persuasive you are, how interested you are in a conversation, how happy you are, and how stressed you are. One of the ways to think about these features is to think about watching a foreign film and turning off the subtitles. You don’t have any idea what they are talking about but you get the sense that this guy doesn’t like this guy, these people are having a heated discussion, and that is what we are picking up on. Not to mention the fact that the computers are very bad at picking up the content especially when you don’t know the context. We have been able to show with this kind of data we can do pretty amazing things.
JESSE: If a company is measuring this kind of information, what good can that be used for?
BEN: There are a lot of things you can do with this data. I think importantly when we work with companies and our, sort of, ethos around this data is companies don’t really have a good business case to look at individual data. What they should care about is what makes people productive? What are the things that make people happy? How can I change the company so that we are collaborating in a way that I think is effective and so that everyone is happier at work? Nowhere in that equation is there anything about the individual. We give the individual access to their own data but that is something that is their own property and the company actually has no rights to that data. So there is an interesting distinction there; but what you can do with this data is really understand what are the actual drivers of success at your company. It is something that is sort of weird to think about, the way that we traditionally manage collaboration, which fundamentally companies are about collaboration. You are bringing people together who, when they are working with each other, are doing things that they couldn’t do by themselves. The way we typically manage collaboration is through charts and formal processes.
In an org chart you will say “If you need to talk to somebody, we are going to draw a line on the org chart. So this is your boss and these are the people who work for you, and that is how we are going to manage collaboration.” The issue is that the kind of work that we do today has just gotten so much more complex than it was even a couple of decades ago. The ability to manage that complexity and the increasing need for very complex collaboration through an org chart is extremely difficult and essentially impossible. What you can do with the sort of data that we are collecting now and the sort of analysis we are doing, is really see…first of all what is the actual org chart? How are people actually collaborating? Then you can actually understand how you can change that. What are the ways that people collaborate that are most effective? It might be something, say, hardware engineers … you might have an intuition that hardware engineers are working on these very individual tasks and it only make sense for them to talk to their boss. They shouldn’t be really talking too much to each other, but that is just an intuition. The fact is that intuition is often right, intuition is also very often wrong and if you look at the problems we have in a lot of business today, it tends to be because we relied too much on intuition. What you can do with this data is test that. You can look at the data and say, you know what? What collaboration patters do make people effective and maybe it is what you thought it could be but it may also be something very different or it could be something you had no intuition about. So just uncovering this hidden side of work that there was just no way to look at this stuff before.
JESSE: So one thing I am confused by…so you are saying the individual data isn’t really tied to anybody, or it is not available to the company. I suppose it is available to you as the third party analyst so you are able to say we are measuring these teams and this team here is interacting more and their performance is better. Otherwise, how do you say what is the long term value of whether they are interacting or not interacting?
BEN: Yeah, that is right; we do have access to individual data, although it is important that their name is not directly in our database. There is a process that we go through that is partly manual to actually connect up productivity data to the behavioral data we collect. Companies do get to look at team level data. If it is aggregated then you can’t identify a specific individual and then we do give teams and individuals themselves access to that kind of data. You have to be very careful whenever you aggregate something. Whenever you show data about how people are behaving, you have to make sure there is enough data included in those statistics so that you couldn’t say, ok, this is Bob. No one should be able to know that except for the person themselves.
JESSE: Now at the companies where you are doing one of these studies, the employees are aware of what is going on. You are not monitoring people without their knowledge. Do they find this creepy or intrusive?
BEN: An important thing to mention you do this on an opt-in basis. In the last year whenever we have rolled out badges and we have gotten over 90% participation in every project we have done. I think one reason we get that kind of buy in … there are two good reasons, one is the whole roll out process that we have. To your point, if I just went into a company and I just pulled out these badges and said “here, wear this.” I wouldn’t participate but I don’t know anyone who would participate under those circumstances. So the key is to be really transparent about the data that we collect and what we do is go into companies and send out material basically showing previous results that we have collected, we show the actual database tables of what we collect and the second step here is not just showing people what we do and what we collect, but we give people consent forms so that this is, again, the opt in part. We sign the consent forms along with the participants that say we will not share your individual data with the company, and, actually, the contracts we have with our clients, they sign away all rights to individual data.
There are actually huge privacy issues in the US. Technically companies could legally own this information if we didn’t have this agreement. A company could look at individual data and that is something that just has to change in terms of privacy issues in the US. Until it does, we are providing those guarantees. I think the other important thing that participants see is really the value of the data in this kind of analysis. When we worked with Bank of America’s Call Center Operation for example, these call center employees have really been managed one way for the past 50 years, since the inception of call centers. When you are on a team of about 20 people, and if any one person on your team is on a break, then no one else is on a break. The reason is in the 60’s you might have only had a hundred people in a call center, so if 20 people took a break at once it would really be difficult to handle all the calls coming in. Of course, we know today, they have thousands of employees but that is the way they have been doing things so they kept dong it that way. But the employees realize that occasionally my lunch time overlaps with the people I work with, and that is a really great 15 minutes when I am able to vent a little bit and have conversations around it and whenever they would bring that up to management they would say whatever…we don’t’ really do it that way. There is really no way to show that this is actually an important part of their day and that this is something that needs to change. With the data that we collect, all of a sudden now you can demonstrate the value of those sort of activities. The idea that you could actually ask the question; “How much money does the company make when two employees actually eat lunch together?” You can actually start to make those calculations. There was no way to do that before. Demonstrating how valuable informal interactions are and how much they contribute to not only your own job satisfaction but to the output of the entire company. It is something that most people see the value in that. They want to help the people they work with and they see this as a concrete, very important step towards making that happen.
JESSE: Can you give us another example of a change that a company has made after going through one of these projects?
BEN: One of the companies we work with was configuring these multi-million dollar hardware systems for customers. So these are data systems and they had hundreds of these engineers who would physically configure the systems. Most of these people have Master’s Degrees but they were paid based on how quickly they could turn around these configurations. So a sales person would call them up and say “Hey, I got company X and they need a server that does X, Y, and Z.” So these engineers had to figure out to do these things: What sort of a system do they need? How do I put it together? How do I price it out? All of these things a simple one could take 5 minutes but a complicated one could take 8 hours. What was interesting is that they were paid…their bonus was determined on their individual performance. So it was based on how many of these tasks they could turn around. What was interesting, though, is they had the resumes of the people so they knew how much education, how much experience they had, and when they looked at that the people who had the most experience weren’t necessarily the most productive. They couldn’t really figure out what was going on. So we went with the badges and we looked at how these people are really collaborating.
Importantly we could actually look at who did they talk to when they were working on one of these tasks. We knew exactly when they started it and when they ended it. What we saw was very interesting. There was this network of people who you would go to when you worked on a task. There were 4 people in the center of this network. What was amazing is if you talked to one of the 4 people at the center, you completed the task you were working on in about a third of the time that you would expect. One of the reasons this was so surprising is that form the company’s perspective, if you were spending your time talking to someone else, then you were essentially wasting your time. Because you should be just working on your task; what the stat is showing is that you are going to people to get advice. Again, it makes sense and it is very intuitive and if you ask people they would say “oh, yeah, well I didn’t know how to do this one thing and I knew this guy over here, he knew how to do that and he helped me out.” Ok, makes a lot of sense, but when you look at the individual output of these engineers, what you saw is they were really middle of the road. These experts were not significantly different from average because they were spending so much of their time helping the people they worked with. Obviously, this is a really big problem. What that means is the people who are actually contributing the most to the overall output of the company are being compensated no more than your middle of the road engineer. Something very surprising is where we actually helped them is to change the way they actually compensated employees based on this data.
JESSE: Yeah, that is fascinating. I think back to some experiences in life at work and I think of either myself or some of the other people who were really well known as great problem solvers and a lot of their time could get taken up by co-workers. As your reputation grows as a great problem solver, they come to you with their problems and get you to solve them, which, of course, you are happy to do because you want to help people, but then in the long run it can hurt your opportunities for advancement and compensation because you are spending so much of your time on other people and their priorities. Of course, you want to be a good citizen and be helpful to people but you sort of have to balance that out too.
BEN: Right and I think that is a very important point. What I would argue is … I think it is interesting, when you think about interruptions, most of us hear interruptions and we think interruptions are bad, and obviously there is a certain time when you are at your desk and someone comes over and says “Hey, I got a bunch of cat videos to watch. Why don’t we watch them together?” Maybe that is not the most effective use of your time, although there is some social bonding that occurs there that is potentially important but not incredibly important around cat videos. The important thing is if you spend a part of your day helping someone you work with 5% more effective that is actually a huge boost for them. Even if you are getting interrupted all the time and your own individual output suffers, then that really should be your job. If you can spend your entire day and make 20 people 10% more effective, then you should be doing that. The issue is, like you brought up, is we aren’t compensated that way. We are not evaluated on how much we help other people. One big reason is that you couldn’t’ measure that before. I couldn’t show how many minor actions with other people really helped them and with this kind of data you can start to do that.
One of the reasons is, especially in the US and the west in general, we have this real basis towards productivity as individual. You know, you like to point to your inbox and say “Look, I finished all these e-mails today.” Or you point to a stack of papers and you say “Look I wrote that report.” That means you are productive. Whereas, interactions with other people, having conversations with other people, you can’t point to that. The data clearly shows…what is really important, what is ultimately important, isn’t your own individual output, it’s what the whole group outputs. When you look at what groups produce and the end result that the people on the outside world see…the people outside the organization see…the stuff that predicts those results is all about communication patterns. It’s all about collaboration. What we should start to do is not only value that more, sort of qualitative it different, but quantitatively measure that and start compensating people actually based on that data.
JESSE: That is fascinating. Now so far we have been talking about how organizations form a large organizational structural level can use people analytics to improve but let’s talk a little bit about the individual contributor or manager. What they can learn from the things that you have been discovering. For example, you have a chapter in your book that says “should I stay home and work in my pajamas?” How do you answer questions like that?
BEN: It’s funny when you think about remote working and teleworking. There is a lot of qualitative …there are a lot of articles out there of respondents talking about why flexibility is good or why people should always be in the office because collaboration is so important. Whenever I hear stuff like that based upon anecdotes, I really get annoyed and say “Look, we really have data on this. Let’s look at what the data says is relevant.” What’s interesting is there is a big difference for people who have been following , again, a couple of months’ ago Yahoo had this whole big thing come up when they decided that they were no longer going to let people only work from home, that they actually had to come into the office. There is a big debate about that, about flexibility and it being fair to parents and everything like that. But what did that decision actually mean in terms of job satisfaction? In terms of retention? In terms of productivity? What did that decision actually mean? We could argue around what the appropriate trade off is but we need to understand quantitatively, what is the trade off? So, for example, if you are trying to decide whether you should work from home or going into the office; there is a big difference between never seeing the people you work with and working from home a couple of days a month.
The data is quite clear that working from home a couple of days a month is not going to adversely affect your productivity and more importantly if you have something stressful to take care of, say your kid is sick or your sibling is getting married and you have to prep a few things before the wedding, if you don’t take care of that stuff, you are going to be extremely stressed at work, you’re not going to be very effective and you’re going to like your job a lot less. So, again, the data is very clear that that makes a lot of sense. It’s when you start staying home 3 days a week, 4 days a week, never coming into the office, that’s when you start to see some really negative effects. We looked at IBM, at a bunch of programmers, some of which were co-located in the same office, and some of whom were remote and what we did is looked at first of all whose code depended on whose code. Some people think that programming is a very individual task and the programmer is the guy who sits in the corner drinking Mountain Dew and typing on the computer all day and actually it couldn’t be father from the truth. Programming is something that is extremely collaborative mostly because your code depends on thousands of other people. If you actually don’t talk to those people, the people who write the code that you use, that is where the bugs pop up. It is 12 times more likely that a bug will occur if you don’t talk to a person that your code depends on.
So what we did is say “ok who is talking to who?” Whether they are co-located or remote, and we aren’t even going to look at face to face, we are just going to look at electronic communication. How much of the time do you satisfy these dependencies when people are co-located and how much of the time that happened when people were working remotely. What was amazing is if you look at co-located groups again, using a very low bar, we look at people who communicated once about a dependency. About 55% of the time when people are co-located they communicate once. That changes to about 45% when people are remote. What’s important to note is when you don’t have this communication, it takes you 32% longer to complete code. So, if you back this out in Yahoo’s case, this means their average productivity and output saved, just on that very basic analysis, $150 million a year by having people come into the office. That is a huge affect. There is actually more there if you look at the average percentage of communications about a dependency, people who are remote only communicate an average of 8 times. People who are co-located communicate an average of 33 times; again, digitally, not even looking at each other face to face. There are a few things that this says. For one thing for various projects we have to be remote. Again, if you start a project in India, you need somebody in India on the ground. A great study by the University of Michigan showed that when you have people who are remote that at the beginning of a project, flying people together to meet face to face made those teams significantly better than those who never met face to face. There are diminishing returns for this.
The idea as simple as if you have a team where certain people are remote, getting them together a couple of times a year and letting people get to know each other even just in a social context, that that’s extremely important. I think the way to think about it, you don’t miss the meetings, you still have the meetings, despite the fact that for those of us who have been in teleconferences with more than a couple of people, know how ineffective those are, but it’s where people get to know each other, going out for a beer after the meeting or going out to dinner, those interactions are extremely important and it helps you get a sense of the people you are working with. When you find yourself remote or you find yourself trying to figure out “do I want to make the drive into work today? Maybe I’ll just work from home.” If traffic is really bad and it’s going to take you 4 hours to commute in, you know, there are limits to this stuff. Try to think of this less as individual output and more as group output. It will probably change your perspective on it. Because what is missing when people are remote isn’t your individual stuff, it’s the interactions you have with other people which makes everybody happier and more effective.
JESSE: It’s interesting because that flies in the face of so many trends where teams are more decentralized, they are virtual. They may interact with each other all day long but they never see each other for months at a time. For example, let’s say I am the manager of a team and one of my employees comes to me and says “Hey, I’d like to start working from home one day a week or two days a week, and, of course, every situation is different, but on average you would say that is a bad idea It is one thing to say work at home a couple of days a month if you need to but this weekly business I am not going to go along with that.
BEN: Yeah, I think it is important to note that here are some roles that are certainly easier to do remotely. If you are a salesperson and you meet people outside the office, then it doesn’t make much sense to pull you into the office every day. For all the stuff we do every day, whether it is writing reports, producing media or writing code or building cars, or building airplanes, managing buildings, these things have gotten much more complex over time and as a consequence the need for communication has increased that much more. The tools that we have at our disposal today are just not very good at supporting those kinds of interactions. And so as a manager, when you are confronted with something like this, obviously there has to be some discussion around it, it’s not like it’s a hard and fast rule, necessarily, but you really have to acknowledge that if you are going to work from home a couple of days a week, everybody is going to be a lot less effective. Not just you. It wouldn’t be a question if it was just about you as an individual – sure if we work from home all the time, then probably individually I can produce a lot of stuff but it knee caps the growth of the organization. It also makes it…I know form some of my friends who worked at Yahoo when they would come in on a Friday and no one would be there, it is demoralizing. That is something that we, as a society, have to start valuing. In the foreseeable future, that kind of interaction is going to be extremely important. Maybe in the future when we have things like holograms, and we can physically reproduce environments, then maybe it doesn’t matter as much, but until then, as humans, we still really need this.
JESSE: One thing that seems contrary to me is when I think about people who are either introverted or ambivrted, for example, me, I perform very well when I spend half of my time when I am interacting with people and half of my time when I am working alone. That is just where I get my energy. If I come in on a Friday, in thinking about Yahoo, and the office is quiet because half the people are gone, inside I think “Oh, great, I am going to get a lot of my quiet work done today less interruptions.” But on the other hand, if I am getting so many interruptions throughout my day that I don’t get any alone time to have that kind of energy flow, it really de-energizes me and I think it may be even harder for true introverts. So how does that factor into this performance discussion?
BEN: I think what is interesting is the things that predict how happy you are at work, the things that predict how happy you and the group are, it’s not the raw amount of communication so it’s not just saying if we sit at the water cooler for 8 hours a day that everyone is going to be more effective. It is actually the pattern of communication. So it’s who to you talk to and how do those people talk to each other? How I that situated in the overall social network of the organization? That’s the stuff that is really important. Now what that means is, if you are an introvert, maybe you only spend 30 minutes a day talking to other people. That’s ok. The question is who do you spend that time talking to? Not just saying that more is better, but really saying you have to have the right interactions. Even if you are an introvert, again, fundamentally, we are in companies and we work in teams because together we can do things that we couldn’t do by ourselves. You need to have communication. You need to collaborate with other people. The extent that you need to do that depends on the type of work you do. Maybe it means that some introverts find that a little harder, however, I would say if you look at the data overall, it’s not really an introvert/extrovert dichotomy, it’s really are people spending their time effectively?
JESSE: Well the book is People Analytics: How Social Sensing Technology will transform Business and What it Tells us about the Future of Work. Ben Waber, thank you for joining us on Engaging Leader.
BEN: Thanks for having me.
Link to podcast episode: EL 41: Using People Analytics to Improve Productivity & Happiness | with Ben Waber