Not long after James Everingham joined Instagram as the head of engineering, results came back from the employee satisfaction survey that’s conducted every six months. The marks were pretty good, but one problem spot caught Everingham’s eye: the low transparency score.
This took him by surprise, as he felt that the executive team was an open book as long as it didn’t violate someone’s privacy or confidentiality. Even after reflecting more and digging further into the data, Everingham still didn’t have any theories.
Ultimately, it was a conflict over a business decision that brought it into focus. After careful deliberation, leadership had made the difficult call to move one of their teams from their Menlo Park headquarters to the New York office. Shortly thereafter, Everingham pulled the team’s manager into a room to give him a heads up. The manager was rightfully shaken up, but it was his response that stuck with Everingham. "I’m really disappointed,” he said. “This is the first I'm learning about this. You didn't involve me in this at all. You were not being transparent."
For Everingham, this was when the lightbulb went off. Transparency is a persistent, thorny problem because we’re not all on the same page about what it even means. To him, transparency was about building trust with your people while giving them better data to make better decisions along the way. But for this manager, it was more about involvement. Everingham soon realized that boosting Instagram’s straggling transparency scores meant figuring out what a common definition was. And that wasn’t all — he quickly recognized that shedding light on the decision-making process would also become a sizable component of their efforts to solve the transparency equation. The engineering group would need to identify who was making decisions, how those decisions were made and why they were being made.
For the next six months, Everingham made boosting transparency and defining a decision-making process a priority — and it paid off. The subsequent survey showed that Instagram’s leadership transparency score increased dramatically by over 10%, making it one of the highest across Facebook.
Inspired by his talk at First Round’s Founder Summit, Everingham shares here precisely how Instagram fixed its transparency dilemma as it scaled engineering from 100 to over 600 people in less than three years. He details the steps in their systematic approach and gives helpful tips for defining transparency, educating others and building a thoughtful process and framework to support it. Now Head of Engineering, Blockchain at Facebook, Everingham is taking this lesson and system to his new team and role — as should any leader of a growing team.
At a foundational level, most agree that transparency is an important element of any company.
“It’d be unusual for someone to say, ‘Hey, you know what would be really fun? Let's start a company and not really tell our employees anything. Let them figure it out along the way,’" says Everingham. “It’s intuitive that that’s not the best way to do business. That’s because in the corporate world, people don’t like the element of surprise all that much. Imagine if you went to work one day, and your employer abruptly said to you, ‘Hey, we ran out of money and can't pay you, so thanks, go home.’ You'd be more than a little upset. But what if instead your employer came to you and said, ‘It looks like we're going to run out of capital in about six months unless we raise some money, and I wanted to give you a heads up.’ Then you would not only be able to make a better decision for yourself personally, but you'd be more motivated to do what’s best for the company in the interim.”
In addition to empowering everyone to make the best decisions, cultivating transparency can help solve a lot of the other problems that startups typically run into, from building trust and reducing politics to allowing people to be heard and maintaining velocity, all while expanding rapidly.
But actually making sure transparency exists is much harder than simply saying “Let’s be transparent,” especially as a startup grows. That’s because transparency often happens by default in the early days. By virtue of having a small group in a cramped space with an all-hands-on-deck mentality, the entire team naturally has a lot of insight into what’s going on. But as a team adds more people and sprawls across offices, that visibility starts to dissipate. A layer of management forms, meetings move behind closed doors, communication becomes less frequent and fewer folks are involved in making the most important decisions.
As Everingham found, simply having transparency as a value isn’t enough — leaders need to go several clicks deeper if they’re serious about opening things up. He did that at Instagram by uncovering his team’s definition of transparency, enacting a standard framework for how decisions would be made and who would be making them, and ensuring every action was soundly anchored in a set of common operating principles.
When Everingham first discovered the discrepancy in how transparency was defined, he started an investigation to unearth why takes on the concept diverged.
“During my 1:1s and skip-levels, I started asking everyone what transparency meant to them,” he says. “After a long stare and some thinking, people would usually give me some variation of ‘Well, it's keeping us in the loop,’ or ‘It means you're not hiding anything from us,’ or ‘It's involving us in things that affect us from the start.’ Those are all very different takes on transparency. I realized that without a common definition, you’ll disappoint someone no matter what you do. You’re also potentially opening up the door for someone to weaponize the concept against you, leveling accusations that you aren’t being transparent.”
Don’t confuse transparency with involvement.
At the outset, it appeared that all Everingham and his team had to do was get together to agree on a common definition of transparency, a seemingly easy fix. This is the definition they came up with:
Clear and consistent communication
Clarity on how decisions are made
Honest and clear feedback
Admitting when we are wrong
Even after nailing down a definition with the team, Everingham quickly sensed that there was more work to be done. “We were already practicing three of these, but we weren't set up to even support that second statement,” he says. “This is what unlocked it for me. I realized it all came back to clearing up decision-making. When you think of transparency, you usually default to the communication aspect: telling everyone what’s happening or admitting when you’ve made a mistake. But when folks say that things aren’t transparent, what they’re probably getting at is that decision-making isn’t transparent. It’s the feeling that decisions sometimes roll on down from the lofty perch of the leadership team, seemingly out of nowhere. Instead, pull back the curtain on how decisions are made, putting some process and principles behind it so it’s not this mysterious black box that’s ripe for speculation.”
The fix for clarity on decision-making at Instagram came down to two things: Using frameworks and principles for the decision-making process and establishing clear roles with single accountable owners. “We had to make our decision-making understandable, consistent and repeatable. We also had to ensure people understood who was the single person making a particular decision. It was about standardizing our algorithms and guidelines,” says Everingham.
So Everingham and his team set out to find a framework. They settled on a simple RACI model for assigning roles and making decisions. Commonplace in consulting, it’s a tool that lends itself well to the world of startups. It works by identifying and assigning who is responsible, accountable, consulted and informed before every project or decision. “My favorite illustration of this that I’ve seen is applying RACI to Star Wars,” says Everingham. “If you’re a geek like me, then it’s perfect. Essentially, in the effort to destroy the Jedi, the bounty hunters were responsible, or doing the actual work, while Darth Vader was accountable, the stormtroopers were consulted and Darth Sidious was informed of what happened at the end.”
Of course, the issues Instagram applied the RACI model to were slightly different. “We used it for our most important initiatives and we actually would always start out by defining the 'A' first. For example, when trying to figure out which database architecture we were going to use moving forward, we singled out our CTO as the accountable decision-maker,” says Everingham. “The ‘R’ group was responsible for doing the work and putting together a recommendation on what we should do, which included gathering input from a group of engineers that were consulted. There was a final presentation to walk the CTO through their thinking, where the ‘C’ group was also invited to hear the decision being made. Finally, we actively identified who needed to be informed of the decision and communicated it out to every affected group.”
In a bigger company, it’s clear how the RACI model can help cut down on chaos. But is it necessary for a startup just getting off the ground, where reporting lines are simpler and teams leaner? “Instagram was a relatively mature company when we even started this process. And it's just plain harder to do this stuff when you get older and bigger,” says Everingham. “This may seem like pointy-headed management stuff now if you're a new founder and just starting, and that’s valid. Usually you’re able to sort most of this out instinctually at the beginning. But I firmly believe that it'll be a lot easier if it's already a part of your culture by the time you need it. There are just so many ways for decision-making to go wrong, whether it’s too many owners, too much of a focus on consensus or an inability to capture or articulate the decision before you. Going through a whole process may seem cumbersome, but you almost always uncover something that needs to be cleared up. And even if you’re not running into any of those problems just yet, I promise it’s coming. It’s a good team muscle to build.”
There are also other approaches startups could adopt of course, from a matrix for growing teams to a framework for difficult decisions. “It didn't really matter which decision-making framework we used, just that we picked one and then everyone knew what algorithm we were going to apply to everything,” Everingham observes. “The biggest consideration for us was that we wanted a single decider model. We wanted people to understand exactly who was making decisions, so you can only have one 'A' or one decision-maker.”
You can have more decisions than decision-makers, but if you have more decision-makers than decisions, that’s when you run into problems.
Everingham shares three strategies that Instagram’s engineering team uses to figure out who would be doing the deciding:
Map it out for all to see.
To bring the RACI model to life on a practical level, Everingham and his team turned to another acronym: RAM, or a responsibility assignment matrix.
“We needed to highlight the 'who' behind every decision or project and make sure that there was only one common understanding, so we went through this exercise of building a RAM out in a spreadsheet, putting the functions or teams across the top, the tasks vertically and filling the names of the owners in the boxes,” he explains. “It’s about who makes the call. Some of them had more than one name assigned, which illustrated a lack of alignment. Some of them didn’t even have any names, which highlighted a clear gap. I found it to be a great exercise to go through as an organization. It reminded me of the classic Eisenhower quote, ‘Plans are useless, but planning is indispensable.’ On its own, a RAM is fairly useless, but the process of building it and clarifying owners is essential to aligning your company. It’s a framework that outlines responsibility and makes visible how a decision will be made, for everyone to see.”
Remove bias, especially your own.
Assigning decision-makers requires some careful consideration. Simply selecting the best framework isn’t enough — in order to successfully implement it, bias needs to be removed from the equation wherever possible.
“Bias often leads to bad decisions. Going back to the New York team relocation example from earlier, it’s fair to ask why we couldn't have just involved the team’s manager in the process from the outset,” says Everingham. “But I felt he couldn’t be a part of the decision, because I already knew what his answer would have been. He would have of course said, ‘No, don't move my team.’ It's self-interest bias in action. We couldn't involve him in that decision if we wanted an unbiased outcome that would work for the whole company. This is part of the line in the sand you need to draw between transparency and involvement. Looking back, of course there are some things we could have done differently. We could have headed this off by being more clear about our priorities as a leadership team. For example, if we had emphasized the importance of having on-site senior leadership or reducing interdepencies between teams, the manager might have had more context ahead of our decision.”
But to mitigate bias leaders also have to take a look in the mirror. “I also do my best to remove myself from the decision when I know that I will be biased,” Everingham shares. “It’s essential to set that example, otherwise it trickles down and suddenly everything looks suspect. For example, I might want to have an office next to my house, full of all of my friends and family members as employees. Yet even if those were the right decision, if they were the best people I could have hired and that was truly the best location for our company, people would still be questioning my motives. It's hard to build trust across your team when they know that you're biased in making that decision.”
Be courageous and reorg.
For Everingham, the broader problem of shedding light on who owns decisions is intrinsically linked to how teams are organized.
“Too many leaders shy away from reorganizing their teams for fear of the churn, anxiety or signals it can send, opting to instead work within the limitations of the existing structure or apply short-term patches. But I’ve seen teams cave in on themselves due to bad design or a simple outgrowth of the original blueprint,” Everingham says. “A lack of clear role definition is incredibly disruptive and actively inhibits transparency. Ownership starts to drift, becoming incredibly murky and slowing things down. Teams can end up working separately on the exact same issues, frustrating everyone and wasting time. You also have senior managers getting pulled down into conversations they don’t need to be in or more junior folks not being looped in when they need to be. When I joined Instagram in 2015, you could really feel some of this in the engineering group. Things weren’t working at the numbers we had at the time, much less at the scale we were hurtling towards. So we reorged, carefully defining the metrics, responsibilities and team leads for engineers, all with the intention of eliminating opacity at every turn.”
What’s more is that Everingham found that this restructuring and clear detailing of responsibilities had added benefits outside of boosting transparency. “Since we transitioned, the biggest learning is that when people clearly understand what’s expected of them, they’re more likely to surpass it,” he observes. “For example, the Instagram Stories features was concepted, built and launched to half a billion people in a little over three months by a small team. I don’t think that would have been possible in our previous setup. It’s breathing more innovation and speed into our engineering team. That’s what’s led me to believe that managing isn’t about giving your team ideas or interfering — it’s about making sure the organization is a well-oiled machine that is set up for success.”
But even if a team figures out how it’s going to make decisions and who is going to make them, a piece of the transparency puzzle is still missing.
“The last part of our journey was figuring out what we're aiming to improve when we're considering a course of action. What are the top outcomes you’d like to see? You need to isolate the ‘why’ behind what you do, agreeing to a set of guidelines that can be a north star for every action you take,” Everingham says. “For us, this meant coming up with a set of principles that we published. To me, principles are ideas or rules that explain or control how something works. 'Be 100% goal-driven,' is a very basic example of what a simple business principle might look like. That means any time you plan something, you need to have a very explicit goal. So with that in mind, our team got together and proposed more than 20 outcomes to strive for, such as ‘build an inspired workforce.’ Then we narrowed it down and prioritized it into our top five from there.”
These are the high-level organizational principles Instagram’s engineering team came up with, as well as what they really mean on a practical level:
Move as fast as possible. We should be continuously shipping value to our users, so speed matters more than cost efficiency. That means we have to minimize dependencies so everyone can move quickly.
Build clear accountability with the fewest decision-makers. We need to avoid trapping responsibility and slowing down action in the cracks that fall between teams on cross-functional decisions or product features. That means we need to 1-1 mappings between product and engineering.
Write simple and clear KPIs. We need to all have a clear understanding of what we’re working towards and what success looks like. That means all groups need to have measures.
Scale to a much larger organization. We can’t assume our current structure will continue to work so we need to think ahead, sketching out what it will look like when we’ll be doing more at a much larger size. That means all organizations should have roadmaps.
Maintain a very high-level of quality and craftsmanship. Every part of the user’s experience needs to be meticulous and consistent, with the least amount of friction possible. That means performance, stability and code should all have their own owners.
And this set of principles and practical operating rules isn’t set in stone. “Anyone can suggest a change and if we all agree, we'll just incorporate it in. But the biggest advantage is that it functions as an accountability mechanism,” Everingham explains. “When we communicate a decision, anyone can refer to our principles and ‘code review’ us for correctness. Team members can check and hold us to the standards we’ve set for ourselves. For example, when we’ve announced an initiative, we’ve heard ‘Hey, that move just created a bunch of dependencies,’ or ‘Who's the decision-maker now that you split that work across two teams?’ It’s very helpful in keeping us honest and making sure transparency is a hallmark of the way our group operates, not just a concept we aspire to.”
The final step was to tie the what, how, who and why together to create a streamlined process to follow when a new decision or issue crops up.
“We refined and tweaked it until we were satisfied that it was solving everything we needed it to,” says Everingham. “But for this to work on the day-to-day level, it has to be a living framework that we revisit and constantly reference, instead of a guideline that sits on a shelf. That’s why we encourage everyone to look at our published process, giving input on what could be improved or calling us out when we’re not following it.”
Here’s what Instagram’s transparent decision-making process looks like — and how it works in practice:
Diagnose the problem. The engineering team noticed that web traffic was growing but they weren't investing enough in the web experience.
Organizational change proposal. The growth team manager suggested that he build out a team based in New York. The group didn’t need to be a very large team, and there were some good reasons that it made sense to do it from that location.
Identify RACI model and communicate roles. The growth manager took responsibility and as head of engineering, Everingham was accountable. Peer engineering managers were consulted, and the Head of Product was informed.
Audit action against principles. But when they audited the plan against the engineering team’s published organizational principles principles, it became clear that this proposal didn’t make sense.
Articulate tradeoffs. (Write them out.) Given the “scale to a much larger organization principle,” there would need to be a roadmap that showed how the new group could grow into a large team with an important scope. Additionally, the proposal only called for a front-line manager running the team, reporting to a lead across the country, which meant that there wouldn’t be on-site senior leadership.
Debate design change, considering alternative designs. By discussing the tradeoffs between different versions that didn’t violate the organizational principles, the team reached consensus on an organizational model that made sense.
Decide. With a new proposal in-hand, the final decision was made to build the new team local to Menlo Park. This allowed a local senior leader to play an active role in developing the team’s manager while keeping close tabs on the new effort. Other benefits included the ability to communicate more closely with important product counterparts, highlight long term career paths for the team’s engineers and take advantage of Facebook’s local talent pool.
Update responsibility matrix. Finally, the RAM was updated to reflect the new team, its scope, who the decision makers were, what they specifically owned and how the new effort would be measured. The entire process and matrix updates were then communicated to the rest of the company.
Most of us strive for transparency, but the hard realities of scaling mean that the principle is more elusive in practice than we’d like to admit. With focused effort and clear processes, Instagram’s engineering team was able to roll back the natural erosion of clear decision-making that inevitably accompanies growth. For other teams seeking to make the same leap, start by establishing a common definition of what transparency means in your organization. Adopt a decision-making framework such as RACI, mapping out responsibilities for all to see. Clarify roles and owners, removing bias along the way and reorganizing if you have to. Identify what outcomes you’re striving for in your decisions, codifying them as guiding principles. Lastly, line up a repeatable process that you can run every decision through.
“The biggest change we saw was that team members weren't confusing transparency with involvement. A lot of our win was simply educating the team on the difference and setting a common baseline. Now instead of throwing blame, people had a system and a language to engage in a productive debate. In the place of a blanket ‘You’re not being transparent’ statement, we could go deeper, respectfully probing by saying ‘What principles did you go by to make that decision?’ or ‘How does that change reduce dependencies rather than create them?’” says Everingham. “In some ways we were changing the DNA of the company. Transparency is an admirable goal to strive for, but you have to set up the right processes to support it or it won’t become your reality. Technology and teams may change, but the need for people to understand what’s going on and feel bought in never does. And even though I’ve led engineerings teams at Instagram, Yahoo and Netscape over the past three decades, I’m starting anew as we build out the blockchain team at Facebook, so I still need to spend the time getting it right. Transparency is a lifelong investment that you have to constantly recommit to.”