Lessons from Coinbase’s Wild Ascent: Four Rules for Scaling
Management

Lessons from Coinbase’s Wild Ascent: Four Rules for Scaling

Engineering leader Varun Srinivasan had a front-row seat as Coinbase scaled to meet an explosion in demand for cryptocurrencies. Here, he shares the four org design rules they relied on to gear up for hypergrowth.

In early January 2017, Varun Srinivasan and the rest of the engineering team at Coinbase returned from the holiday break to an unexpected present as the new year got underway.

The price of Bitcoin was skyrocketing, pushing past $1000 and showing no signs of slowing down. And while it was an exciting development for those who had placed early bets in a nascent industry, Coinbase was struggling to keep up with the blistering pace of the cryptocurrency’s growth and the explosion in demand that followed. User traffic was setting new records every day and customer support struggled to keep up as a barrage of requests bogged down response time. Srinivasan’s team was pulled into rooms to put out fires left and right, in one instance working almost two days straight just to keep the site up.

After coming up for air, Srinivasan checked in with a mentor for some encouraging words to keep him going — and got a reality check instead. “He asked me: ‘How are you going to fix this? How are you going to grow the team and systems fast enough?’” Srinivasan says. “And that’s when it just hit me: this is what hypergrowth feels like. And we’re not even remotely ready for it. We were only looking one square ahead of our pain points but in reality, we needed to think six to 12 months ahead and work backwards to design the processes that would get us to the scale to handle this growth.”

Fast forwarding to today, it’s clear that Coinbase has made considerable progress on that journey. Now a Director of Engineering, Srinivasan has had front-row seats as the company has swelled past 500 employees, opened new offices, overseen the safe exchange of $150 billion, and rolled along with the punches of every wild price swing in the cryptocurrency market. In this exclusive interview, Srinivasan shares how they pulled it off, walking through how Coinbase shifted its mindset and codifying four org design rules startups can lean on in the ascent to hypergrowth.

SPOTTING AND RESPONDING TO THE ONSET OF HYPERGROWTH

Sometimes hypergrowth is a ship on the horizon. As sales forecasts climb and revenue ticks upwards, all signs point to a scale that startups can see coming in the distance. This may give founders at least a little time to prepare for the move from the blade up to the handle of that hockey-stick curve. But other times it’s a bolt from the blue, a surge in customer demand that leaves teams, infrastructure and leaders unprepared — and quickly underwater.

Srinivasan’s experience very much mirrored the latter path. “The most surprising thing about hypergrowth was just how rapidly it overtook us at Coinbase. SaaS businesses can have a pulse on when things start picking up and forecast growth pretty well. But when you're operating a consumer business, sometimes things just explode. In Coinbase’s case, it had a lot to do with broader market conditions. But it could come from unlocking a viral loop,” he says. “That’s what makes preparing for it so hard. Hypergrowth wasn’t something we could have banked on, but it requires thinking ahead and planning for a longer time horizon, which you can’t afford to do as a really early-stage team. Before coming to Coinbase, I founded my own company and I know firsthand that everything is a struggle during the zero to 10 employees phase. You're worried about product/market fit, not running out of money, motivating the team and ordering lunch — you really only have the bandwidth to look two or three months ahead.”

Even for the startups that have moved out of that earliest stage and found more solid footing, hypergrowth requires a complete shift in perspective. “Now our engineering team has grown to over 100 people, but for my first two years at Coinbase, we were very lean and efficient in our operations. When I joined we were an engineering team of 20. We were always talking about our burn rate and tried to keep spend top of mind, so we hired conservatively, only adding a couple engineers a month at our peak,” says Srinivasan. “The surge of interest in cryptocurrencies in 2017 made it clear that this approach wasn’t going to keep working. We had to change our mindset entirely or growth would turn into this undertow that would pull us down. What makes or breaks companies is their ability to react in that phase. You need to move beyond the short-term optimizations that accompany a frantic search for growth and mature into thinking and operating very differently. Our problem wasn’t simply sourcing more engineers. It was completely revamping our hiring process, because there’s no way it would scale to hiring 100 more people.”

It doesn’t matter how you get there, it’s how you keep up with hypergrowth that defines you.

To help startups making a similar leap from keeping the lights on to powering hypergrowth, Srinivasan has drawn from his experiences at Coinbase to develop four rules for outfitting companies for hypergrowth.

RULE #1: SHAPE YOUR ORG CHART TO MOLD YOUR PRODUCT

To cope with hypergrowth, often the next logical step is to start scaling up teams to build the capacity to handle that pressure. But adding more individuals without rethinking how groups interact can lead to growing pains that radiate throughout the org chart. As teams balloon in size, organizational complexity and subdivisions are introduced to help wrangle and streamline an increasing set of moving parts. But increasing communication overhead and conflicting priorities often slow decisions and prevent startups from retaining speed as a habit.

At Coinbase, Srinivasan saw these side effects firsthand. “When we were launching our new cloud deployment pipeline on our platform, we were around 80 people and engineering was a 50-person group newly divided into six sub-teams. About a month into the project, I walked into a huge meeting and it was absolutely crazy,” he says. “The room was packed to the brim, but it was hard to understand the specific problems that we had gathered there to solve or what each group’s responsibility was. One team thought they were shipping everything on all platforms in a month, while another group thought that they were shipping just one component on a single platform in two months. Our new structure had left us a bit disconnected.”

For Srinivasan, this instance was the living embodiment of the warning that you ship your org chart. “Variations of this concept have been floating around for awhile, popularized more recently by Steven Sinofsky but dating all the way back to Conway’s Law in the 1960s. It’s the notion that products are ultimately shaped like the structures of the teams making them. In other words, when you create an org chart, you’re shaping your product because it’s the sum of all the teams’ missions and the resources you put against each of them. It all needs to add up to what you want the company to achieve,” says Srinivasan. “Make sure your org chart always reflects your priorities, even if it means you have to reorg or wade into the unknown waters of virtual teams for the first time as you grow.”

To ensure org charts add up to the outcomes founders are trying to achieve, Srinivasan outlines the most common issues that startups should take care to avoid:

  • Overlapping missions: Additions to or shufflings of the org chart often result in multiple teams being charged with similar mandates, which can create a lot of tension. “In my early days at Coinbase, we had a risk engineering team and a data science team that both owned preventing fraud. They wanted to attack the problem in different ways and it wasn't clear who got to make the final call. Things would often escalate upwards and it caused a lot of strain,” Srinivasan says.
  • Conflicting mandates: Teams can also be tasked with goals that are entirely at odds with each other. “Back in 2016, the engineering team was shipping features to increase revenue in the U.S., but the business development team was measuring their progress by the international percentage of revenue. If you’re looking at the whole picture, you would clearly see that the engineering team’s success would result in the business team’s failure, but in the fog of hypergrowth it’s easy to miss those connections,” says Srinivasan.
  • Resources in the wrong places: Another danger is over-indexing on a strategy and failing to revisit your structure to course-correct later on. “I've seen a lot of companies get very aggressive on certain goals to the detriment of others,” says Srinivasan. “For example, if you overhire on your growth team, but you’re losing users to the competition over features, you're in a tough spot. Not only do you have resources in the wrong place, but if they do their job right, user retention will get worse. This type of misalignment happens more often than you think. Companies aren’t thoughtful enough about this resourcing upfront, and when they realize it they're often too slow or too worried to reorg and fix it.”

To maintain decision-making velocity and solve for these org chart math problems during hypergrowth, Coinbase relied on these two strategies:

Comb through every team, every quarter.

The root of this issue often comes down to leaders not looking at their organizations closely enough to spot these mistakes and ensure their org charts reflect what they want to ship. Nowadays, Coinbase makes sure this doesn’t happen through quarterly exercises in which a single leader evaluates each team’s mission, deliverables and metrics with a careful lens.

Here’s Srinivasan’s advice for startups looking to replicate their process:

  • Get a head start. “Startups need to develop this habit early on, right around when you hit 50 people. Under that you're still small enough to figure out challenges and quickly resolve them by just talking to each other. But once you cross that threshold and engineering starts to divide into multiple teams, things start getting complicated and this becomes a really helpful exercise. In the early days at Coinbase, we didn’t think carefully enough about mapping teams to opportunities upfront, and sometimes we lost precious weeks or months building things people didn’t want. At the very least, this process will force a conversation about priorities upfront, and you'll make the necessary hard cuts,” he says.
  • See how it all stacks up. “Make sure to avoid overlaps and minimize dependencies so that the sum of every team’s goals adds up to the company mission. The goal is to look across the entire organization and see if the bets that you're making are tracking effectively,” says Srinivasan. “For example, when we kicked off an initiative to launch Coinbase in new markets, it became obvious that we weren’t making enough progress. Using this exercise to dig deeper, we found that while it was the backend team’s number one priority, the frontend team actually had a totally different mission and was shipping other projects focused on growth. It’s about ensuring that partner teams can co-set goals and look at each other’s bandwidth.”
  • Pick the right perch. When choosing a leader to lead this exercise, make sure it’s someone that rolls up to the founders and has enough visibility across the company. “The COO is often a good choice. Of course, when you cross a thousand people it's no longer possible for one human to conduct the entire exercise, so then you can subdivide the responsibility between VPs,” says Srinivasan.

Use virtual or matrixed teams to make a change.

Whether it’s a rigorous meeting cadence, pushing desks together or actually taking your pen to the org chart, Srinivasan notes that it’s important to create the virtual teams or structures that enable rapid communication and alignment when launching big company-changing projects.

Here’s how they did it at Coinbase:

  • Make it the priority. The first step is to set the goal for all to see. “This can be as simple as saying, ‘This is the most important thing for the company right now and this needs to stack rank above everything else,’” says Srinivasan. “It's about forcing that conversation and clarity around what’s important upfront — you don’t want to find out that people are confused later on as things are going out the door.”
  • Spin up a new group. “When we wanted to quickly launch Ethereum to our customers in 2016, we turned it into its own separate three-month project, temporarily consolidating a couple of the teams under one mission. We matrixed in people from other functions and had them join our engineering standups so we could be as closely knit as possible,” says Srinivasan.
  • Get a leader to step up. To make a group truly feel like a super team, it helps to have one individual take charge. “On the Ethereum project, we had one person that naturally started operating as the technical lead, functioning like an organizational pacemaker. He said ‘I’m going to own this, I’m going to create a document that goes from day one to launch and re-encapsulate everything every team needs to do,’” says Srinivasan. “That simple act of interviewing everybody and identifying that giant dependency list was critical for bringing the team together. It just wasn't happening when we had all these cross-functional pieces. Each group had their own leads and they were all working at different paces.”
Varun Srinivasan, Director of Eng at Coinbase
Varun Srinivasan, Director of Eng at Coinbase

RULE #2: PUSH STRATEGY SETTING DOWN THE CHAIN OF COMMAND

In the early days, many founders prescriptively tell their engineering teams exactly what to do and when, either out of a conviction that they know the right answer or out of a fear the group won’t work as effectively without strict guidance. But in Srinivasan’s experience, that isn’t something leaders should pack in the move to hypergrowth.

“Teams going through hypergrowth need to learn the difference between strategic goals and tactical goals. As you push for growth, goals should become less tactical and more strategic,” he says. “When we were a smaller startup, we were operating very tactically and our goals were very specific, such as ‘Launch in the EU on this date.’ This worked because the team was small enough to see the big picture and understand how it all connected. But as we got larger, we had to shift to a much broader, strategic goal such as ‘Get 100 million users,’ which gave the team more leeway to figure out the best way to make that happen. At a certain size, leaders don’t have enough context to go as deep, so they need to trust teams to get it right. And it works, because as the organization grows, there are more people thinking about your problems and they're going to come up with better ideas,” says Srinivasan.

Founders and leaders need to loosen their grip and give up a piece of the strategic decision-making process to gain speed and flexibility for the startup to grow into its clothing. You should always inspect and understand the strategic thinking your team comes up with, but it’s about empowering someone else to do it rather than just taking over yourself. Figure out what kind of strategic thinking you can push to other parts of your organization and do it before you hit 50 people — that’s where it tends to break down,” he says. “To start, you can give up the strategy setting in an area you're most familiar with to someone who is in a growth role, so you can still inspect or deep dive if need be. Or you can delegate the area you're least familiar with to an expert, although I would caveat that this doesn’t work as well for product and sales — it’s hard to outsource those until you're already good at it."

During hypergrowth, strategy needs to be broad so teams have enough room to figure out the tactics that make sense. Otherwise you’ll never move as fast as you need to.

RULE #3: THINK IN 3D TO VISUALIZE WHAT’S MISSING

“If you design your organization by just looking at org charts, you're going to get it wrong,” says Srinivasan. To anchor this assertion, he points to an unexpected example from the world of cartography.

“Most people rely on maps because they’re convenient, even though a globe is a more perfect representation of the world. Because maps are 3D projections onto a 2D plane, they give you an incomplete perspective of the world. So when you fly from San Francisco to Dubai, you may be surprised to find yourself going close to the North Pole because it doesn't look that way in a map. The visualization tool doesn’t fully capture reality,” says Srinivasan.

Humans take 3D structures like globes and organizations and squish them into a 2D format because it's easier. But you lose a lot of nuance in translation.

“Org charts work in the same way. They’re flat renderings that miss the nuances of how teams relate to each other. They represent two dimensions: hierarchy and functions. They show you who reports to whom, which is helpful for identifying decision-makers and knowing where to go to get alignment. The functional aspect is also useful because the relative size of each group shows what the company is investing in,” says Srinivasan. “But org charts don’t show you the reality of how teams work — or don’t work — together. Most people look at org charts and ask ‘Have I staffed these teams well? Do I have the right sub-teams in product?’ But there are so many other things to consider. How are teams going to work together to get your most important launch out the door? Are the right parts of the organization communicating frequently enough? What cross-dependencies are you missing?”

The effects of this missing dimension of org charts are particularly acute during a growth spurt. “The clarity of ownership begins to drop dramatically when you’re going through hypergrowth. There’s new teams, new managers and a whole wave of new projects. It’s the combination of different perspectives, thinking in 2D and 3D, that can help founders capture all the dynamics and the challenges hypergrowth brings,” he says.

Here are two strategies Coinbase used to create different lenses to view their org in 3D:

Map out how the work gets done.

As another tool in his belt, Srinivasan uses “work maps,” or structures of how people need to work together to achieve critical objectives.

“In addition to the org chat, I use work maps to capture each of the big strategic things that we want to ship this quarter and then list the set of teams or individuals that need to work together to make it happen. You’ll almost always discover something interesting,” he says. “Maybe you see one person showing up in all of these lists and you realize, ‘Okay, this person's on three of the most important things, how is she going to get this done?’ Or you notice that two teams need to be working closely on this upcoming project but they’re actually sitting on entirely different floors and it's hard for them to work together effectively,” says Srinivasan. “You still need an org chart, but if you look at reporting structure, the work that needs to get done and the people needed to get it out the door, then you'll have a more complete picture of how your organization actually works together.”

Sample org chart and work map

Get permanent with your structures.

Work maps can help uncover deeper structural issues that may necessitate a reorg.

“It’s almost certain that you'll need to reorg once or twice as you scale up, but most startups wait too long on this front,” says Srinivasan. “I think people often don't reorg fast enough to account for changing priorities during hypergrowth. In the early days, you can get away with or work through your org chart limitations, so it’s easy to think that carries over. But when you're in this high-pressure environment, you need teams to be focused and empowered.”

By looking at both the work map and the org chart, Coinbase was able to rearrange its structure to accommodate the realities of hypergrowth. “By the time we grew to a 300-person company, we had many distinct businesses, some of which operated like an early-stage startup and some of which operated like a growth startup. There were cross-dependencies across functional groups, such as legal, compliance and support, and they were getting pulled in so many different directions with competing priorities,” says Srinivasan. “In the early days, when you're a single-product company of 50 to 100 people, you may largely be aligned across your functional groups, only spinning up working groups occasionally. But if you notice that some cross-functional initiatives are always challenging and the root cause is communication, think about grouping them together into a more permanent structure. What we found at Coinbase is that the as we grew large enough, we wanted to make a working group permanent for a year or even multiple years. That's when we shifted to more of a business units model, reorging to a structure where we had Business Units with Core Teams and Service Teams in supporting roles. We found that matrixing people into cross-functional teams and giving them a single priority eliminated tension and gave people focus.”

Here’s an overview of how that works in practice:

  • Business Units are cross-functional product teams that build things for a specific set of customers. “These groups typically have engineers, product managers, marketing folks and a general manager,” says Srinivasan.
  • Core Teams are deep functional experts working in non-product domains that matrixed into business units to support them with functional expertise in achieving their goals. "The Data team has data analysts, engineers and machine learning experts, some of whom matrix into Businesses Units, and others who work as a functional group to build a machine learning platform,” Srinivasan says.
  • Service Teams build internal products or platforms that help Business Units scale faster by solving the highest common denominator problems. “The Infrastructure service team built Odin, which is how teams at Coinbase ship code to production,” he says.

“For companies such as Coinbase or Google that have a plethora of products, the business unit structure becomes important early on. Once you start having different sets of customers, it’s tough to optimize across them unless you have your teams organized around those groups. The evolution of this structure down the road when you're a over 10,000-person company is something like Amazon, where a group operates as a fully self-sufficient single unit with its own profit and loss metrics,” says Srinivasan. “At Coinbase, we’re kind of halfway there in that evolution, relying on that dotted line matrixing where you work with other groups but still report into your own functional areas. But the unit as a whole works together, sits together and is a very close-knit team culturally. We're not quite at that scale where we can split up HR and all those other functions.”

RULE #4: PUT PEOPLE OVER STRUCTURES, COMBINING TEAMS IF YOU HAVE TO

While tinkering with organizational schematics to match the hypergrowth phase, startups should spend more time considering the individuals slotting into those structures.

“During hypergrowth, you need to create new teams and new leadership positions and you’ll naturally have some gaps that are hard to fill,” Srinivasan says. “Often founders feel pressed to promote someone early because there's a perceived gap in the org chart and this person has the most domain knowledge. But what if they’re not a good fit? Forcing a square peg into a round hole is not the answer. Promoting an ineffective leader will hurt your company a lot more than having an imbalanced org structure. Great leaders think about organizations in 3D and create working structures that compensate for the org structure, building around the skillsets of capable employees.”

Map your organization to the skills of your people, instead of trying to fit individuals into an org chart.

“It’s better to unify two groups under a strong leader than to create the 'right' org chart with the separation you think you need, even if it’s a really unusual combination,” says Srinivasan. “At Coinbase, Dan Romero was the head of business operations but he ended up running the support organization as well. While it’s an atypical pairing, it was valuable because he was able to go in with his lens of efficiency and optimization to help Support scale. Or maybe you have a really strong product manager, but they're not really the type of person that can represent the team at an executive level, so you have your eng leader step in and take over product as well,” he says. “A lot of startups start splitting those functions up at a certain size because, especially for engineers like myself, people want to see a clear delineation of responsibility. But it can actually work out better when you have a leader to represent both groups, operate them effectively and eventually work on finding a replacement.”

How these group-merging leaders emerge.

Srinivasan encourages founders on the cusp of hypergrowth to look around and identify these leaders proactively. “While it may seem like this combination of abilities is rare, if you look back at successful startups from Facebook to LinkedIn, there always a few people who’ve played this role. I’d bet that there’s a person like this at most startups,” he says.

Here’s a list of attributes to look out for:

  • Team player, not empire builder. “These people aren’t out to own other teams as a land grab. They put company first and recognize what’s most important. They go pick up areas that maybe aren’t something they ever wanted to do, but are important for the company to be successful,” says Srinivasan. “This is critical because in the long run, you want this person to step away from it, hand the reins to someone else and help with the next problem. They can’t be building their own fiefdom.”
  • Strategic communicator. “The person who over-indexes on communicating concepts and strategies to the rest of the organization is who you should tap to take on more. Look for someone that sees the forest for the trees and can be the force multiplier that pushes other people to think that way as well,” says Srinivasan. “During hypergrowth, everyone at the company will be motivated to work harder and pull late nights. But there's a narrower subset of people who are more strategic and say ‘Okay, we have 10 times the support volumes, how do we want to think about handling this in the long-run because we can't hire 10 times as many people.’”
  • Domain knowledge. “The leader that’s capable of taking over another group during the chaos of hypergrowth needs to have a certain level of operational knowledge in that function. They don't need to be experts by any means, but if you take somebody who fundamentally doesn't understand engineering and put them in charge that's very rarely gonna work out,” says Srinivasan. “They need to operate as an individual contributor in at least a limited capacity in order to work at a higher level. If you're bringing on a leader for your support org, have this person actually work the support lines for a day. You want somebody who really understands what it is to be in those shoes in order for them to really be able to run that org, even temporarily. Obviously individuals are going to spike more in the function that they rose up the ranks through, but get someone who recognizes that and who can create the counterweights as a support. Maybe they have some really strong eng managers or tech leads who they actively seek out and rely on to round out their own shortcomings as head of eng and product,” he says.
When hypergrowth comes, your leader isn’t the one focused on late nights or land grabs. She’s planning for and communicating about the long haul.

BRINGING IT ALL TOGETHER

Hypergrowth can be tough to spot in advance, but once founders find themselves in the thick of it, it’s important that they shift the organization’s mindset and design to brace their team for the wild ride. Given that startups ship their org charts, consider putting a single leader in charge of reviewing goals across the company on a quarterly basis to look for overlapping or conflicting missions. Use virtual or matrixed teams for key projects in order to ensure alignment and focus amidst the chaos of a growth spurt. Make strategies more abstract as you expand in order to give teams enough room to fill in the tactics underneath and move quickly. Use a work map as a supplementary tool to capture the nuances your 2D org chart is missing and consider reorganizing into business units as you grow into the clothes of a multi-product, larger company. Finally, don’t try to fit people into an org chart, even if it means you have to unify departments in unusual ways under a single strong leader.

“At the end of the day, one of the most important things to remember is that hypergrowth puts a tremendous amount of strain on your employees. Companies and people grow at different rates — some will step up to the plate and do phenomenally well, but you’ll also have others who are pushed into a role they can’t grow into fast enough. Identify these signs early, take remedial actions to address them and have the hard conversations. In some cases, you're going to have to say ‘Look, this role has changed dramatically, I know you're trying your best and I'm here to support you, but I think we need to hire a senior director for your group.’ Or maybe it’s just that their energy is drained, they’re burning out and you want to help them get back to a better place,” says Srinivasan. “Hypergrowth is very exciting, but it’s also scary. It changes the fundamentals and staffing of the organization. It’s really important not to sweep it under the rug because you want to retain all of these people that are helping your company grow. They've been there through the hard times, they've made sacrifices, so go out of your way to take care of them by trying to create a role that excites them. Looking after your people while being realistic about what the company needs is what’s going to propel you through all the permutations of your org chart.”

Image by Jorg Greuel / Getty Images.