Amazon web services sign hero.jpg

AWS’s Michael Arena: How Organizational Network Analysis Works in the Real World

Michael Arena is the VP Talent & Development at Amazon Web Services (AWS) and author of Adaptive Space: How GM and Other Companies are Positively Disrupting Themselves and Transforming into Agile Organizations. He’s also a leading expert in organizational network analysis and founding member of the Connected Commons consortium, co-managed by thought leader Rob Cross and i4cp.

Michael ArenaMichael, who will also host the i4cp Connected Commons 2020 Summit (October 20 – 21) at Amazon’s headquarters, sat down with us to walk through practical applications of network analysis based on his time at Amazon, General Motors, and other organizations.

Kevin: One of the findings from our recent research on culture renovation was that high-performance organizations are much more likely than low performers to identify hidden influencers in the planning stage of a culture change initiative. Based on your extensive experience working with social networks to drive innovation and other change initiatives, what advice do you have for companies that are looking to engage these influencers in a broader change initiative once they've been identified.

Michael: It's an interesting question. The first thing I would say is that it's more than just identifying influencers. I view this a little bit like, find the potential in the network or find the high-potential individuals. So the question is, "influencers for what?" Network analysis can certainly help you to find people who are uniquely positioned in the network, like bridge people or high energy people. But then I would ask what they are passionate about and what can they influence.

I'll give you an example: I can remember a study where we were trying to influence diversity and inclusion practices inside General Motors, and there was one group of people that was uniquely positioned inside the network, mostly bridge people but also energizers. In an entirely different study we were interested in people that wanted to reinvent and sort of innovate in a grassroots way. That was not necessarily the same group of people, because their passions were different.

So, potential for what? My first response to you would be, “influencing for what,” because network position and context matters immensely based on that.

That's a great distinction. It's one of the things I noticed as I started digging into the Connected Commons research that you and Rob have done over the years – that the way you engage folks is driven by your purpose in doing so. Much of the network science research has been about engaging people to be more innovative or agile. I've not seen as much about culture influence and that's specifically what we called out in our study — this idea of identifying hidden influencers for a culture change.

In the context of culture, I would look at it from two different levels. I would ask, who are the people who can spread culture across the organization: That's one type of influencer. And who are the people who can continue to reinforce culture in inside a given cluster or a small group? That's a different network position.

That reminds me of the 4D model you developed and share in your book, Adaptive Space. The Discovery connections of brokers, the Development interactions of connectors, the Diffusion connections of energizers, and the Disruptive connections of challengers.

Completely, and this gets more specifically to your question. If I'm trying to introduce something new, I want to pull in those bridge people. The high-energy bridge people are phenomenal at spreading ideas. At General Motors we would bring those people in, and we were trying to change the culture. We were also trying to get people to share ideas more broadly as part of that culture change, and we would bring them in to seed an auditorium with high-energy bridge people around the topic of sharing ideas and becoming more innovative. We would bring as many as 200 to 300 of them into the auditorium, and then we would invite people up on the stage to do pitches. We found those people were predisposed to pick up those ideas and spread them across the organization—it worked remarkably well. We could connect those dots and show how things were seeded in these experiences. But the reason they were seeded was because of the influencers that we invited to be part of the audience, not part of the pitch.

One of the questions that keeps coming up in conversations with our members about this has to do with the appropriate level of transparency when engaging people based on network roles. Did you tell the people in the room—the bridging folks or the brokers—why they were there and that they had been identified as such?

Not quite in those terms, no. We may say something to the effect of "you know you demonstrated interest in this topic" or "as part of the broader community wanting to drive change." Rarely did I ever say “Hey you were a bridge person" or "you are an energizing bridge person."

Later, as we got more mature and this was a completely volunteer group, we openly shared those networks. They would actually be able to say, “Hey, it's pretty obvious why I got selected, I'm high energy and a broker.” But that that was not necessarily something we openly disclosed.

How does the nature of the roles that people play influence how you're thinking about structuring the event or how you're thinking about designing the event itself and the interaction? You've given one example, but can you say more about how you think about and plan these events?

This is really what I focus on in Adaptive Space—again, why I ask, "influencer for what?" There are times where you need to get people to go discover, and I pull a different group of people together to go do that. There are times where I need people to be locked into a two-pizza team and develop, and then I'm pulling together deeply trusted cohesive people to do that. There are other times where I need to get people to diffuse.

So anytime somebody is asking me to run a network analysis I ask the question, "for what?" What are you trying to solve? What is the challenge? In my mind the architecture I'm thinking about are these four social architectures, and I'm always asking questions like, "Are you trying to discover?" If you're trying to discover or learn something new, I try to pick people who are cross-functional.

I can think of the MAVEN story that you've heard me talk about many times. In that particular case, what we did was pull together deeply trusted, highly credible connectors for a temporary moment of approximately six weeks and to go out into the marketplace and study urban mobility. The reason that was important is that, first of all, they become believers as they work together. Then they go back to their day jobs of whatever trusted cohesive pocket they were in, and they become influencers of a new strategy in their pockets.

They bring Discovery back to the local group. They're able to influence differently there because they're now deeply convicted believers of something they learned in the marketplace and they're now influencing in five or six or seven different pockets in the organization. Those pockets are critical to getting this new strategy deployed. That's how this plays.

I'm constantly thinking, in Discovery, I want people who are from different pockets but deeply credible to go out and learn something new. I want them to become believers of something because they've experienced something new. Then I want them to come back and influence locally because those local groups are required to get the new strategy in play.

You initially select them based on their network position and other factors, but ultimately they end up becoming influencers because of their ownership in the process. Would you describe that as a design-thinking process that you employed with them?

From a Discovery mechanism standpoint we would use design thinking very often, but it was never a good idea to pull 10 engineers together and send them out in the market because you're already starting from a point of commonality and common ground—and what I want is something that's disruptive. If I pull two engineers, two salespeople, a marketing person, a manufacturing person, and a finance person, then what I get is a diversity of thinking. But I also get multiple influencing points when they are dropped back into their home rooms.

That’s a great illustration of the value of a structurally diverse network that Connected Commons research has called out. Can you think of other examples where you've convened groups of people using different methodologies (other than design thinking)? We've talked about liberating structures for example, to generate kind of dialogue and involvement you were after.  

We've used all kinds of different methodologies. Agile development teams, where we bring people back and give them six-week or four-week sprints, to take a concept or an insight that they learn maybe through design thinking, but it doesn't have to be—could be through market analysis for example. The methods don't matter as much as how we seed the network by leveraging the methods.  

We've used multiple liberating structures. That's really where the tipping forward stuff grew out of. We've used hackathons, where we bring together a diverse group of people who are working together in a 24-hour blitz to design a new minimum viable product of some sort. I've used all those different structures.  

Another key insight from the Connected Commons research is around collaborative overload. When applied to the stakeholder engagement question that insight offers strong guidance to take the collaborative demands that are already placed on the people in these key network positions into account when you're forming these groups and project teams. Since many of the most influential people are also the most overloaded, I'm curious how you have managed to engage these folks without burning them out or further overloading them. What kind of strategies did you put in place to manage that dynamic?  

I can think of a couple different ways that we've done that. With big challenge projects that would always have the senior executives most likely to be overloaded, we would pair them with a more junior high-potential person inside their organization. Because what I didn't need from the executives was their tactical abilities to build and create something—what I did need from them was their sponsorship and influencing abilities. That was one tactic or one approach that we used.    

There were other times where we ran collaborative overload analysis to try and push decision points out. I can think of a time where 15 percent of the network was making something like 65 to 70 percent of the decisions. We would lock those people down and force them to look at all the decisions they're making, and try to get them to diffuse those decisions deeper into the organization. That's another tactic that I can think of.  

In that kind of scenario, when you're really helping people structure their work, is there a point at which you are actually transparent about the network analysis to indicate the data shows they are overloaded with decision-making demands? 

In that case we were. I would not show the entire network, but I would say, "Hey, you were in the top 15 percent or the top 5 percent of people from whom others wanted greater access to your agenda." And then we, HR, would have a one-on-one conversation with them. We rarely would show the bottom 85 percent where they were—we would just say they were in the other 85 percent. 

With that privacy piece in mind, how did you handle that on the front end? In announcing the network survey or organizational network analysis (ONA) work, did you set policy around what you were and were not going to share?  

One hundred percent. That all begins at the beginning. There were two core contracting components. Number one was, what are you solving for, because we're not going to just build fancy interesting network charts. If somebody told me, "I want to communicate better," or "I want to collaborate more," that would never be a sufficient answer. We would always ask why. Why is collaboration important? Are you sure you want to collaborate more? Maybe you want to collaborate less. Are you sure you want to communicate more, because maybe you're communicating too much already.  

I would always flip it back to the leader and ask her why is communication important? First of all, communicating more always wigged me out because it was almost never the core issue.  

But collaboration was always an interesting question. Where do you need to collaborate more, and for what? OK, so you want to collaborate more because one pocket has a lot of insights and the others don't have access to it. That's a good reason. Let's talk about the business reason as to why that's important.  

So, getting super crisp on a single purpose was always modus operandum #1. We would never do one that was, “Well, I want to collaborate better and I'm trying to figure out how to create better working teams.” We always had a single core focus so that it wasn't just an interesting chart on the wall. That's number one.  

Number two, we would then ask the question, “How transparent do you want to be with this?” We have three contract groupings:


  1. We could be fully transparent, and we did do analyses that were fully transparent. I can talk about that in a moment.
  2. We could be confidential. Confidential meant that we would run the analysis and my people analytics team would have access to names, but nobody else would. We would hold the data confidentially. And the reason we would do something like that is because there could be a time where we would reach back out to the top 10 percent and say, “Hey, you were recognized as a top 10 percentile person, would you be willing for us to disclose your name and share your information?”  
  3. And we could be anonymous. In the anonymous case, basically what we would do is we would contract to re-code each individual. We didn't do many of these actually, but we would use an identifier and store it off premises so that it couldn’t be misused or abused. I would say that those were super rare.


Confidential was the norm and transparent was one that we would sometimes do. We were super crisp around these three classifications, so that the survey invitation would highlight the intention, and therefore oftentimes what the data wouldn't be used for—such as not using it to evaluate performance, but rather to think about better collaboration structures. We’d also specify how your identity would be used. We were fierce on contracting.

The purpose established in the contracting phase would drive the decision-making around the confidentiality and dictate how you communicate to the survey population?

Yes, and that was all very much specified in the invitation to each participant.

What’s an example of an analysis where you were fully transparent?

I would say that as you earn trust with these things, that they don't get misused over time. We evolved toward being more and more transparent.

The one case that I wanted to share with you is the GM 2020 work, which was a voluntary community of ultimately almost 8,000 people. We would refresh that data set every year. And we created an open platform where you could click on your network position and see exactly who you were connected to by name. You could see it sorted by department and you could actually use that to go find the people you needed inside your network of volunteers, in order to activate whatever strategy or idea you might have. So that was like the most extreme case where we actually made it almost like open source, where anybody could click onto any other node and find out exactly who was where.

We obviously had to disclose that's what we were going to do before. But there was like minimal to no risk with that because these were volunteers in that community. It wasn't like their day to day job. There was very little risk.

I know you are trained in design thinking and are well-versed in a number of other methods for engaging people in innovation and disruption, but what guidance do you have for companies that are interested in doing some of this work but don't have that capability?

I think it's fine to work with a consultant, but I also think that we make these methodologies more complex than they need to be. If you are going to go the consultant route, I'd say go find a good broad-based consultant who has done everything from design thinking to liberating structures to agile scrum and bring that person in to help develop the assortment of plays.

But my experience has actually been more about unleashing that potential from within, like the whole GM 2020 thing. We just brought people in as volunteers and they helped figure it out—they helped design those things. They helped create what design thinking can look like. We invented our own solutions—for example, we did something we called a code lab. Tipping forward was almost like a Shark Tank pitch for the masses. Most companies already understand those kinds of things.

I think the far more critical piece that I would coach an organization around is what I was talking about just a few moments ago: getting really clear about what your intent is. What is your business intention? And then start to think about the social architectures and/or roles within the network that are most critical to pull together for each of those.

With design thinking, I would find a certain group of influencers in the network to do that. If I were putting people into small development teams and they were building minimum viable products, I would bring a different group of people together. In that case I might want a bunch of engineers together. If I were talking about diffusing, I would bring yet a different group of people together. I think the far more important question is, “Who do you invite?”

So you have to ensure the approach is a fit for both purpose and role.

One hundred percent. Yeah. Hundred percent.

You mentioned that you did some things with ONA at GM related to D&I. We're seeing increased interest from our members in this idea of measuring inclusivity by looking at the network, but there are some legal risks it seems. Can you share your experience with that at GM?

What we mostly looked at was gender diversity.

Generally, if you give people a choice of who they interact or work with, there is usually pretty strong homophily. Especially from a gender standpoint, where a network is bifurcated or clustered.

What we were interested in was how do we use things such as ERGs, not to create more homophily or cohesion within a diverse group, but to actually spread that out across the network. Those ERG or affinity groups are awesome and important in a certain stage in order for people to feel empowered and build local trust and have an in-group kind of conversation. But if you stop there, it could be dangerous, because what you end up doing is creating deeper relationships within a singular group.

I'll talk about it from a network standpoint: what you really want to do is also  create like mixing chambers where you're basically breaking up that homophily. Some of the things that we did at GM were to try to get diversity into the core of the network. I just wrote a blog on this, but what we want to do is get past thinking about diversity numerically. That's the first stop. Once we get past questions such as, “have we brought in diverse candidates?” what we really want to do is bring diversity into the core of the network, because that's where influence happens.

This is why I think things like diversity in staff roles, core influencing roles, or critical roles is so important from a talent perspective because what you're really trying to do is have diversity at the influencer level, which is usually at the center of the network based on the way network diagrams are formed. Let's just say you were 40% female and 60% male. If all the women are in the outside clusters of that network, you're fooling yourself if you think your diversity strategy is working. What you really want is a 40/60 split in the core, or maybe even a 50/50 split in the core of the network, because that's going to continue to put diverse groups at the center of influence.

This reminds me of something Amy Edmonson talked about at the last Connected Common Summit. She was taking about how homophily is effective and that groups with high cohesion function well, and yet, paradoxically, we also know that individuals with a structurally diverse network tend to perform better and innovate more. When do we want homophily and when do we want diversity, and is it purely utilitarian and based on what we're trying to accomplish?

There is always a balance between cohesion and brokerage. You never want just brokerage, or just cohesion. The reality is that you want oscillation between the two across time. There are times where I want very cohesive pockets—as I said earlier, there are times where I want engineers all together.

It's harder for me to think about this from a diversity standpoint because I might always want some degree of diversity, and I could still have cohesion. But if I had radical degrees of diversity with really unique people with diverse experiences in one team, then you never get the cohesion you need to move fast. Cohesiveness drives speed and decision-making. There are times when I want speed, and setting homogeneity aside, having a group that has more cohesion, has worked together, and is of like mind is a good thing in that case.

If they stay together too long, then you have this huge risk of social cohesion that could actually create cultural lock-in and is equally as dangerous. So it's always this tradeoff between lock-in and speed, and I think if you if you use oscillation as a mechanism to constantly reshuffle the deck, then you get the benefits of cohesion and you get the benefits of bridging connections at the same time.

If there's a natural tendency towards homophily and cohesion, especially in underrepresented groups, maybe the opportunity is simply to more proactively encourage the other side of it—the structural diversity or the brokerage.

One hundred percent, and I think that's where you're being very intentional about movement and creating some oscillation in the network, like moving people from a very cohesive pocket to a new pocket so that they maintain those connections but also have to forge new ones.

I know you've been doing a lot of interviews with the release of your book, but what haven't I or others asked that you wish you did get asked so that you could share with people?

One thing we haven't talked much about ,and that Rob Cross speaks to a lot, is the whole concept of energy and emotions in the network. The basic level of analysis inside a network is the dyadic relationship that you and I have with each other, and how you emit emotion and/or energy to or from me. I just think when you're talking culture that's critical because that dyadic relationship is the way that ideas, culture, and energy spread. I know we didn't talk a lot about that, but I wouldn't want to miss it because it's critical. It's essential to driving change inside an organization. And it could stifle change because you know those dyadic relationships are always critical in nature and shut down. It's a little bit of the emotional safety stuff, right? Low psychological safety shuts down creativity, whereas  high-energy, positive emotional exchanges can generate creativity. Both are important, by the way. But as you might imagine they’re important at different stages in the development process. I want high energy, strong positive emotion emitted early. And later, like whenever I'm pressure testing an idea, that more critical challenging mindset is more critical, more important.

So as the architect of experiences like this where you're thinking about all these things and trying to match the right people with the right methodology and purpose, are you incorporating data from outside the analysis, like your subjective experience of people, for example?

Completely. Early on, I try to invite the advocates and follow the energy. Bring in the believers because the person with the best idea might not even believe in their idea yet. They need people to play with. They need to find a friend or a handful of people who are believers just to get the idea off the ground. Once we've gotten it off the ground and we feel like we're getting movement, then I want to bring in the challengers, the devil's advocates, to make the idea better and to pressure test it. If you bring in the devil's advocates too soon you may prematurely shut down an idea. But if you ignore them, you can foolishly charge forward with a stupid idea. So being able to think about how to cadence those things in the process, and thinking about how to cadence those things from a cultural spread standpoint is critically important.