Adrian Cockcroft

The subtitle for this is kind of the inside story of the Netflix culture deck. So I'll refer to that a little bit. It's also about systems thinking, and one of the things about DevOps — DevOps is supposed to be a systems thinking approach. So what do we mean by systems thinking?

I'll take some quotes from a book that is a really deep book on the history and what's going on in systems thinking. And also another deep writer on systems thinking, Scott Adams, who wrote Dilbert. He wrote a book on the personal approach to systems thinking that he uses, which is actually quite interesting.

And I'll refer to the Netflix culture deck and the Nordstrom culture decks. There's some interesting things there.

So what's all this about? Why do we care about systems?

Well, most people are struggling with trying to get more innovation in their company. And a lot of the time people say, "Well, we need better people, right? We need people that can be innovative." And what you're trying to do is increase talent density. And the problem is that most company cultures are actually systematically reducing talent density.

So if higher talent density feeds innovation, why do so many company cultures drive away talent? And what can we do that's different about that?

So I'm really talking here about the people problem, largely, in getting these transformations underway. And part of this feeds off a conversation I was having with Henk Kolk, who's the chief architect at ING. We were talking at the GOTO London conference a month or two ago, and he was saying this is the problem he's struggling with. So I wanted to pick up this area and talk a bit about why it is that culture and the people all interact together.

So starting off with Netflix — the freedom and responsibility culture is a document that was released in 2009. And then in 2013, Sheryl Sandberg said this was a very important document. So where did this come from?

I'm going to give you a little bit of the background story about how this happened. So I joined Netflix in 2007, and this document obviously wasn't out, but they were trying to explain the culture to me during the interview process, which was kind of complicated. Half the interview, I was like, "What are you doing? That sounds different. How does all this work?" And it sounded fascinating to me.

And what they found was during the interview process, the first hour or two of the interview, either people would be fascinated or horrified. It divided people one way or the other, and the horrified people — okay, just walk out and forget the rest of the interview, we're done. And the fascinated people were mostly not believing that this was real. Is it just a story that you're doing this, or is it for real?

So I was one of the ones that got fascinated. I joined and started trying to figure out this culture from the inside. At the quarterly management offsite — so I went to all these offsites — if you think about seven years of quarterly management offsites at Netflix, I don't really need to go to an MBA, right? It's like I lived inside one of the all-time MBA case studies from the inside, and all the MBA case study write-ups are all wrong because they're done from the outside, but you can actually see it if you live on the inside for that long. So this for me was one of the attractions of joining Netflix. It was an incredibly powerful experience in figuring out a lot of these ideas of management. And you'd see an idea come up and over the following few years, how it played out. So being there for long enough to see this happen was also really interesting.

But one of these events, there would be pre-reading and there would be a few emails before we went out to the offsite, which was just a one-day meeting, basically, some hotel somewhere. And Reed sent out, "Okay, I'm thinking of publishing this deck that I've written about the culture. Take a look at it. Turn up at the meeting with your comments scribbled all over it. We'll sit in lunch tables, we'll go to the tables, and we'll each do little group discussions. And at the end of that, after lunch, we'll report back what everyone thinks. And you contribute back your fixes, and I'll look at the whole thing and decide whether we should release this." Which is a pretty controversial thing to release. And also, what should be in, what should be out, and all that kind of thing. So it was a collaborative editing by the entire team, but it was written by Reed to start off with. With a lot of input from Patty McCord and a lot of other people at the management level. So it had been discussed at the VP level and the exec C-staff level, but this was now everyone down to directors at this offsite.

So we were looking at this thing, and it kind of made sense, so we put it out there. And the most immediate effect it had was on the interviewing process. If you're going to come and interview at Netflix, like during the telephone screenings, go read this document. It's listed in the jobs section of the Netflix website for a reason. Go read this. If you turn up and you haven't read it and you don't have good questions, it's sort of an instant interview fail anyway. It's like, you're not paying attention, you're not serious about this. If you read it and you're horrified, you won't apply for a job. If you read it and you're fascinated, you'll share it with your friends, and maybe they'll apply for jobs. So it turned into a viral job marketing thing, and it still is. I'm trying to remember how many views it has — millions of views. Three million or nine million, something like that. I'd have to go look at SlideShare. But for a company culture document to have millions of views over a few years is huge.

So it went out there, and actually a year or two later, we had another pass through it, and we did a revision and tidied it up and did some updates. And then they also decided that there needed to be another document to describe the business model, because this document was really aimed at everybody in the company, but for the more senior managers there were some things they want to know about the business. So there's a second document describing the business model, which was even more controversial. It sort of reveals the secret sauce of how Netflix thinks about things and what markets it's going after.

So that was interesting. So what is in this document?

I gave this talk in France, so here's a hint: if you go to a country, quote somebody from that country at the beginning of your talk — it butters them up a bit. So quoting a French philosopher or writer to the French was a good policy.

From The Little Prince: "If you want to build a ship, don't drum up the people to gather the wood, divide the wood, and work and give orders. Teach them to yearn for the vast and endless sea." That is a systems approach. It's not a top-down, fine-grain management, do-this, do-that, we've-got-rules approach. It's — make sure you like movies and you want to show them to everyone in the world. Make the movie-watching experience and the TV-watching experience better. Care about that, and figure out how to make that better in every way. And that's kind of the cultural system that they put in.

Most of the staff meetings start with five or ten minutes going around the table: what have you watched on TV? Any movies, any TV shows, anything like that. It's in the culture, right? That's the thing we should be caring about, and you do that every week. So there are these reinforcing things that a lot of the teams use.

So what is this systems thing? This is Jamshid Gharajedaghi — I don't know if I'm pronouncing that right, but I'm going to meet him next year because I sent this slide deck to him and he said, "Oh, you should come and talk to my students." He's a professor somewhere on the East Coast.

Systems Thinking: Managing Chaos and Complexity, a Platform for Designing Business Architecture. That sounds useful, that sounds interesting. And what we're talking about here is purposeful systems. You've got companies — what is your company's purpose? What is your group's purpose? Everything has a purpose. So this is this purposeful systems, purpose-driven approach. If something doesn't have a purpose, you kind of dig in — maybe that's something you shouldn't be doing, or you haven't figured out what it's for yet. So think about that.

And then this thing: plurality in three dimensions. What is he talking about? Function, structure, and process.

Let's start with process. If you have a single process and you lock everything into that process, then it can't evolve. The point about this systems thinking approach is you're building organizations that evolve — they're designed to evolve, they're designed to add new capabilities and new techniques as you grow and as you cover new markets. And Netflix is very highly optimized for high evolution. It brings in new ideas all the time. It's willing to try new things. Whereas the sort of traditional, stodgy old enterprise, you get locked in. "We're going to be efficient. We're going to have one release process, and everyone will follow it, and it will have a document." What was Mark Schwartz's document that we talked about for the Department of — at DOES last year? Some thousand-page document to describe why it would take a year to make the slightest change to a system. That kind of thing.

So that's process. Plurality of process means that you have a purpose — the processes aren't going in random directions, the processes are oriented around a purpose, the purpose of releasing code to the site, say. But you can have multiple — that's the purpose, but the process for getting there can be adapted and change and evolve over time.

And then plurality of structure. You've got the hierarchy — the business hierarchy with the founder at the top and a bunch of people underneath, everyone reporting up in a structure. That's the traditional business structure. And then you've got a producer-consumer structure, where this group produces something that's consumed by this group, and it's consumed by this group, and the final consumer is the customer. That's another structure across your company. And if you say one is dominant and the other one doesn't matter, then you're not playing this game right. There's a plurality of structures. There may be other structures in the company, but you have to understand that you're building a system that has multiple structures, and again, these structures evolve over time, and they're designed to evolve over time. If you get too rigid into your hierarchy and every few years do a big reorg and everyone gets upset because they've just about figured out how to manage their manager or something — that's a problem.

And plurality of function means that things aren't just used to do one thing. Why do you do this? Well, there isn't just one reason for doing it. There may be different reasons for different people. If you look at the open source program that I basically helped create and run for Netflix, it was done for many reasons. Why did you do that? Well, there were four or five different reasons. There was a plurality of function there. It was designed to help hiring. It was designed to help stop us getting dead-ended into a technology trap by being the only people to do things one way. It was designed to make it so that you can't talk about cloud anywhere in the world in any form without mentioning Netflix at least once every half an hour somewhere in the presentation or blog post or whatever it is. So that kind of technical PR. And it was designed to take the internal people that we had and make them into more valuable people. These are all different functions that are totally unrelated, and they're different weights to different people, so there is no single reason.

I think there's a problem that we typically fall into: to make something simpler, we will boil it down to a single thing. There's a single reason for doing things, there's a single structure — that's simpler. But the problem with that simplicity is it locks you into a structure which is fragile and rigid. And that rigidity is part of the problem. And if you manage by managing down to a single simple one-liner, then you've got a problem because it's not evolving.

So what we're talking about in this book is how to build systems — the systems thinking approach — which creates an evolutionary architecture. It looks confusing, it looks chaotic if you're used to looking at the single thing. But when you understand what's going on behind it, it makes perfect sense. If you think about something like a fractal, there's this one very short equation that creates an incredibly complex structure. But until you know what that equation is, you don't realize that there is an incredibly simple underpinning to a fractal picture. It's that kind of thing. Until you learn to see the roots of these new systems, they look complicated and they look pretty crazy at times.

So this is another slide from the Netflix culture deck. "With the right people, instead of a culture of process adherence, we have a culture of creativity, self-discipline, freedom, and responsibility."

Okay, so this is a very clear statement. We're not trying to make everyone get rewarded for following the process. We're trying to give you the autonomy to do the right thing for the purpose of your group and the purpose of the company. But then the hard part of this is the top line: the right people. Where do these right people come from?

If only we had a room full of the right people, we would be doing better. So I was doing a talk about Netflix culture a few years ago, and I got this question: "We can't copy you because you have all these superstar engineers. We don't have those people." And I looked around the room at the name badge and it said, "We hired them all from you."

So you had these people. In fact, you probably still have people like this in your organization, and you're systematically getting in their way and annoying them to the point where they leave and come and work at Netflix.

You just look at the people that Netflix is hiring — some of the best engineers in the world want to work at Netflix because they see it as a place where they can be productive. And what that's doing is reinforcing this talent density. But what's also happening is these people are getting driven out of good organizations.

So now we're all friends of Gene Kim here, so you've all read The Phoenix Project, because that's kind of why we're here to some extent. And there's this character called Brent. He's this pivotal, central person who's the only person that can fix everything, and everything goes through him, and he's getting burnt out. You've got to find in your organization these people — they're really easy to find. They're the Brents in your organization, the people that can get anything done, and because of that, they're just buried in stuff.

You have to figure out how to take them out of frontline responsibility, put them up above it, allow them to say no, give them the ability to say no to doing things so they can stop and think. And they will be the best architects, the best people to drag you into the future. They know how it works. When you send them to conferences to explain how things work, they will attract other people. They will basically turn into the core of your — whatever you want to call it — your distinguished engineers or your fellows or whatever. You need a program that takes these people and turns them into the core technical conscience for your system. But then you have to give them the ability to say no to random requests for work, control their own schedule. Maybe they have some responsibility, but it should never be to the point where they're burning out. They've got to be the people that go to conferences and find new ideas and bring in all of this new evolutionary thought as things go forward.

That's a relatively plausible thing to do. I was at Sun, I was a distinguished engineer at Sun, and one of the ways you could tell somebody was a distinguished engineer was that people had heard of them. That was the point. So there's James Gosling and all those kinds of people, and I snuck in there because I wrote a book — as was mentioned — and so all the other DEs had got my book, so they'd heard of me, and that got me in, basically. But if you think about that, what you're trying to do is create this group that can think and that can steer you forwards, and that is outside the regular management chain to some extent. You've got to have this technical conscience. So that's one way forward. It's possible to do at big organizations, and a lot of people do actually have something like this.

But always look for that Brent-like character and think: should that person be pulled out so that they can fix the big system rather than just being the point of contention?

The other thing about firefighting — and stopping firefighting — is that the people who are doing firefighting should be writing tooling to automate things away, but they're too busy to do that because they're firefighting. And if you think about concentrating all of the things that go wrong into a small number of people in operations, you're concentrating the heat. There's no wonder you get fire there. Like you've put a big magnifying glass up to the sun and you've got a hotspot. If you spread the responsibility across the whole organization, the heat is very, very dissipated — there's little spots of heat. But if every developer's carrying a pager, that is a systems thinking approach. And that systems thinking approach causes you to spread the work out. It causes people to get a feedback loop that says, "If I break it, I know how to fix it, and I'm on the hook to fix it, and I will eventually build things that don't break."

So let's look a bit more at the Netflix culture. These are the seven things. "Values are what we value" — it basically says these are not intended to be empty claims. We do actually take care of this, and the company culture is one of the biggest things that is taken seriously at Netflix. High performance, freedom, context not control. This is instead of micromanaging. If somebody consistently micromanages in a management role at Netflix, they'll probably be let go and replaced by somebody that knows how to give context and get out of the way.

Pay top of the market. I'm going to talk a bit about the compensation system, because I don't often talk about these things, but Netflix has an extremely strange and unique compensation system, and I want people to understand why it's done that way, because it's part of this feedback loop that causes talent to stick around.

Here's another culture deck — Nordstrom Technology: "Good judgment in all situations." And when it comes down to it, the thing about good judgment is that if you give somebody context and they have good judgment, you can reliably get a good outcome given that context. If you give them the wrong context, they'll do the wrong thing. So the manager's job is to give everyone the right context and make sure people are exercising good judgment and are held responsible for their results. That's the responsibility part. And then basically, if people keep doing the wrong things, it's probably either the manager's giving the wrong context or they're hiring the wrong people. So there's this sense that if a team is dysfunctional, don't fire the team, fire the manager.

So this is another piece of the Netflix system: the compensation. There are no bonuses in the system. There's no "I'm going to wait until the end of the quarter to lay this person off so that they get their bonus" or the end of the year or anything like that. Everything's paid every month, and all of the stock vests the same day. So there's nobody at Netflix who has golden handcuffs on to stay. That's actually an interesting poison pill for people that might want to acquire Netflix — the staff could just all leave and keep their stock.

Somebody said, "What would happen if such-and-such company bought Netflix?" I said, "I'd leave. But I keep my stock." So basically, you're free to leave. What that means is, because you're free to leave, you end up with a culture of only the people that actively want to stay being there. And this is a bit backwards in some respects, but by making it easy to leave, people are more willing to stay. And it means that if somebody really shouldn't be there, there's a very generous severance package. It's like, "No, you're the wrong fit for this role for what I need at this time, and you need to go somewhere else." So that system works. People stay because they're passionate about their work.

Now you've got a system where you can go to anybody in Netflix — some people I've never met before, I could go across the other end of the building, somebody I'd never met before, and be confident that they had good judgment and they were passionate about their work, because it's a system that causes only those kinds of people to be there. It's not 100%, but compared to most companies, it's way at the good end of the scale on this. And that leads you to trust other people — it's a very high-trust organization because of that. And getting trust into an organization is very asymmetric. It's hard to build trust, it's easy to destroy trust. It's one of those asymmetric things.

So going back to Jamshid: "An organization's success is the product of the interactions amongst these five basic processes of throughput" — getting stuff done, that is — "decision-making, learning and control, membership, and conflict management." Does your organization have an understanding of what its conflict management process is? Is that handled in some known way across the organization?

This is a very deep book. It's got a lot of history and a lot of good thinking in it. If you read this book and you want to discuss it, I'd be happy to talk through it.

All right. So there's another thing you might have heard about recently called Holacracy, mostly because people are making snarky comments about Zappos. But Holacracy is a systems-based management approach which is in some senses a little bit like Netflix, in some senses a bit like what's in this book. It's an interesting take — it goes a bit further than Netflix does. Netflix has this thing where it sort of absorbs ideas without ever giving them labels. Because the Netflix DVD business — any Six Sigma analyst would have just had a field day. It's beautiful. The Netflix DVD shipping system is one of the most perfectly tuned Six Sigma projects ever. And the Netflix development systems are all agile, but with a lowercase A, and you don't have to do particular scrums or stand-ups if you don't want to, but they do in some places. And then you've got a management system that takes some of the ideas that are in common with systems thinking and Holacracy, but doesn't give it the label. You just sort of do it and do what makes sense. So everything is basically very fluid.

The point about this though — constitutionally derived power. What does that mean? If you think about, say, a startup — that's the easiest way to think about it. You've got the founders, they're the kings of the startup. It was their idea, they own it, and they probably own most of the stock at that point. And then they have a bunch of people that work for them going down, which they can hire and fire. That is something like a feudal system. You've got the king, and you've got the courtiers, and then you've got the workers, and at the bottom there's some serfs.

Holacracy — to move a typical startup to Holacracy, you create a constitution, and it's like a revolution. The CEO gives up power to the constitution. And in fact, then the entire organization has this role which is sort of a bit like a CEO, but it's like the difference between a king and a prime minister. Like if you look at a country like the UK or Australia — Australia just had this, they just changed their prime minister, that's just like flipping out a president. That wouldn't happen in America because it's a different system. But it's like whoever is currently leading the party inhabits the role of prime minister, which is where the executive power is. The executive power is in a role, and people inhabit the role. So the whole point of Holacracy is about creating a system of roles and people inhabiting different roles. And those roles — you can be in multiple roles, and those roles are not sort of handed down like birthright, like divine right of kings and all this kind of stuff. And the hierarchy is very different, which is why people find it confusing.

So think of it as a much more, in some senses, democratic, constitutionally driven way of doing management, rather than the autocratic "I own this company and I will do what I want with it" kind of way of doing it. That I think is interesting. It's definitely worth reading this book. There are good ideas in it, although you probably don't want to go all the way, particularly if you're a big organization. I think it works at small scale pretty well, and Zappos is sort of the largest example of somebody trying to do it at larger scale, and they've been having interesting problems trying to get everyone's head around it.

So the summary. The summary of the freedom and responsibility part: as Netflix knew it was going to grow, the trouble with most companies as they grow is they add more and more rules, and then everything gets slower and slower and harder and harder, and then all the good people leave. So the thing Netflix was trying to get to was to grow bigger and have less rules by having higher talent density. So the system's feeding back on itself now, because everyone in there is committed and has good judgment — you don't really need rules. The vacation policy is "take some." The expenses and HR handbook is summarized as "act in Netflix's best interest." That's it. That's the only thing you needed to know. If you can't figure out what that means, you don't have good judgment. So it all goes back in a loop.

And also this thing — flexibility is more important than efficiency in the long run. If you keep tuning a system for efficiency, it locks down until it can't change. The most efficient systems are the ones that aren't changing. So by putting more and more efficiency into your operations side and driving them to cut cost and be more efficient, you're actually inhibiting their ability to change. And I think there was a quote from Nordstrom last year, one of my favorite tweets, which was like, "We changed our optimization from efficiency to agility." That's basically what we're talking about here. And that kind of, for me, summarized last year's DevOps Enterprise Summit. These are the people here who are trying to make that transition. You don't want to be wasteful, but you want to optimize for flexibility and agility first.

And then Werner said this: "You build it, you run it." This is the sort of foundation of DevOps for me. This is 2006, before the word DevOps came out. This is the basic architecture that we were just coming up with in 2010. We read this article, we figured out what it was, and it kind of aligned with the way we were operating anyway. So we went off and we did something that we later found out was called DevOps or some variant of DevOps. But the point was to give people the responsibility for the components of their system. So break the system into small components — and we could talk about microservices or whatever, but that's really just a tactic for getting a component that you can manage — and then you basically take that forward.

And then Conway's Law, which comes up every now and again: organizations designing systems will only produce designs that are copies of the communication structure of the organization. So this is why I say DevOps is a reorg for most organizations. If you're trying to build a microservices, DevOps architecture, agile architecture, you have to have an organization that supports that rather than fighting against it. If you have your stovepipe teams, then you can arrange something that looks like it, and you can say "I want to do this," but the teams will revert back to their stovepipes. So you have to break out of that and create an organization that is actually built as a network.

All right, so the last slide I've got. A goal is a specific objective that you either achieve or don't at some time in the future. And it's kind of annoying — if you overachieve a goal, you feel you were sandbagging, or you feel dissatisfied; and if you underachieve a goal, you are dissatisfied. It's actually kind of hard to be happy with goals.

But what Scott Adams says is you should have a system. Like his system for doing startups is to learn as much as possible. Whether the startup succeeds or not, he doesn't care really. He'll be happy if it succeeds. One of them is going to succeed eventually, and that's the way that good founders do good startups. They learn as much as possible, they build relationships, they're setting themselves up for — in two or three startups' time — they will succeed. That's the systematic learning approach, rather than saying, "I will put everything into this, and I will cling to it, and if it fails, I will fail with it, and then I'll go off and do something else." That's the wrong approach. This is a great book, by the way, and though it's humorous in places, it's actually a really good systems thinking book, but more on the human level.

All right, so one last quote from Jamshid, because I'm over time here. "We see the world as increasingly more complex and chaotic because we use inadequate concepts to explain it." You've got to understand the underlying equation behind the fractal — then it starts to make sense. You've got to understand Holacracy is about a constitutional approach to authority — then it starts to make sense. These are the underlying patterns that you need to figure out, and because when you understand it, it no longer looks chaotic and complicated.

So that was what I wanted to leave you with, and I'm out of time for questions, but happy to take Twitter chats, and I'll be around. I'm sorry I missed this morning, but I'm going to be around through Wednesday afternoon. All right, thank you.