Applying a scientific approach to scaling

There is no single “right” way to build your company. There are some universal truths (e.g. people really do matter!), but most processes, tactics, and org structures are unique to each company and its stage of development. What worked for Airbnb in 2015 won’t work for Brex; Facebook’s org structure doesn’t apply to our company; and Square’s processes won’t match our team and org.

That being said, at Brex, we have applied one universal principle to nearly every challenge we’ve faced: We use the Scientific Process to constantly improve how we operate. In doing so, we’re continuously evolving (and hopefully improving); we’ve become Darwinian in our approach to ideas, org structures, processes, and more. While we don’t believe our exact decisions will work for your company, we do believe the Scientific Process is a helpful framework for nearly every challenge a company faces.

While we do not believe our exact decisions will work for your company, we do believe the Scientific Process is a helpful framework for nearly every company challenge.

So what exactly is the Scientific Process?

When we talk about applying the Scientific Process at Brex, it means embracing an iterative, hypothesis-driven approach to company challenges. First define the problem; then form a hypothesis, execute, and assess.

We’ve changed enormously over the past 2 years. Originally, we were organized by ‘service,’ then ‘business,’ and now ‘goal.’ Two years ago, we had one financial service (Card), whereas today we offer three (Card, Cash, Capital). When I joined in the Summer of 2019, we had 100 employees; we just recently crossed 450. And we started with one office, expanded to four, and have since evolved to remote-first. Had we not iterated through these changes, we would undoubtedly have failed; our systems and processes needed to change in unpredictable ways as our company evolved. As I expand on below, our organizational structure, in particular, needed to change nearly every six months. The org structures and processes that enabled us to be successful at 100 people and one product do not work in a world with 450 people supporting multiple products.

Applying the Scientific Process to Brex

There is no playbook for managing a company through this type of growth; there is no objectively great way to manage a company facing our specific challenges. And similarly, no company is going through exactly whatever your business is facing today (with exactly the same team, timelines, fundraising circumstances, competitors, market dynamics, etc.).

So instead of replicating what seems to have worked elsewhere, we would strongly recommend applying the Scientific Process to your scaling challenges. Here’s how we use it:

1. Problem: Clearly define the specific problem, condition, or gap to be addressed or improved upon
2. Hypothesis: Take your best guess at what will help your company overcome the given challenge (often informed by what others have faced and shared)
3. Execution: Implement solutions wholeheartedly (half measures lead to poor conclusions)
4. Results: After a defined amount of time, reflect on (1) what worked and (2) what failed
5. Conclusion: Keep what worked, and form new hypotheses re: how to fix what failed

Repeat, repeat, repeat

So instead of replicating what seems to have worked elsewhere, we would strongly recommend applying the Scientific Process to your scaling challenges.

As you can imagine, this process produces lots and lots of change. Many companies are afraid of change — and oftentimes change is extremely painful. You might hesitate to shift reporting lines, alter processes, or switch decision makers; when change happens, you as a leader are admitting you made a mistake– sometimes a very costly one. At Brex, our mistakes have resulted is mishirings, misfirings, and missed goals. Change is often viewed as a bug, but from my experience at Brex, it’s the ultimate feature. Change is the only way you can remedy your problems. Only by creating a company that embraces change can you ensure that you’re consistently better today than you were yesterday. Only by being a leader that embraces change can you ensure you’re putting your company in the best possible position to succeed.

Change is often viewed as a bug, but from my experience here at Brex, it’s the ultimate feature.

The Scientific Process in action

To make this Scientific Process more concrete, below is how Brex’s org structure has evolved over time:

Problem Statement

Brex’s success will be driven not by what we’ve already built, but by what we will create over the coming years. So, we need our Engineering, Product, and Design (EPD) teams to drive Brex forward. They are not currently operating with the business-oriented autonomy and accountability we need to be successful. This problem is resulting in delayed timelines and misprioritization.

Experiment 1 (August 2019)

Hypothesis 1

We will increase EPD quality and development velocity by creating autonomous, business-oriented product teams.

Execution

We shifted from service-ownership teams (web, mobile, statements, integrations, users, etc.) to business-oriented teams (Card, Cash, and Capital). In making this shift, we moved EPD from a focus on engineering service quality to business outcomes.

What worked

Decision quality and development speed for Card and Cash increased materially. This change single handedly enabled us to move from supporting one business to multiple.

What didn’t work

EPD accountability continued to be unclear. A large amount of each team’s success was driven by ops, sales, support, and others. For example, we wanted the Card team to feel accountable for Card success (measured in transaction volume); however, Sales was able to most effectively drive transaction volume within any given quarter. There was very little EPD could do to move the transaction volume needle intra quarter. As a result, the shift did not result in EPD feeling accountable for business outcomes. Our engineering teams were building and executing, but they were not yet acting like owners.

Experiment 2 (Jan 2020)

Hypothesis 2

We can increase accountability by bringing cross-functional teams together and forming “business units.” These business units will act like steering committees with complete ownership over all outcomes.

Execution

We created cross-functional teams that together were responsible for the entire P&L for each of our businesses. After this change, there was a single person on product, design, engineering, ops, support, finance, and sales aligned to each business. This change required us to realign 6+ organizations and transform our planning and decision-making processes.

Our cross-functional leaders met weekly to run each business. For instance, the Card cross-functional team met every Wednesday, discussed business performance, and shifted priorities to improve outcomes.

What worked

There was one group that entirely owned the success of a given business, all working as a unit towards a shared goal. Never before had our cross-functional teams worked so well together. Each person within this group felt extremely accountable for the business’s success.

What didn’t work

We dramatically slowed decision-making. There was no single decision maker on a given issue. If issues were contentious, it could take weeks to reach consensus.

Experiment 3 (Current)

Hypothesis 3

Group decision-making is a bug, not a feature. We need to create an organization with:

• The autonomy that came from our independent business-oriented product teams, and
• The accountability that came from our “business units”

Execution

This hypothesis was a clear admission that we had failed with our “business units.” We unwound the cross-functional teams, shifting to independent metric ownership. From Ops to EPD, every team now had a mutually exclusive business metric. There were no longer “owners” for a business; there were owners for a core business metric. For instance, before, we aimed to have our Card team “own” transaction volume. With this change, the Card EPD team would own % spend on Brex, and Sales would own transaction volume.

Implementing this change was extremely difficult. Many people were asked to take on completely new roles (sometimes in new locations), and several teams were asked to wholly shift priorities and focus. Some folks moved to new teams, and sadly some people left Brex altogether. This type of change is hard, but necessary given the problems that surfaced during our “business unit” experiment.

What’s working

Still evaluating. Decision speed has increased materially. Metrics ownership is creating more clarity for teams, making prioritization easier than ever before.

What’s not working

Still evaluating. There is no longer clear ownership over a continuous customer experience (e.g. no one team owns the end-to-end Card experience). Therefore, no single team feels accountable for Card customer satisfaction (as measured via NPS).

Continuing to refine and evolve our scientific approach to scaling

As you can see, with each change, some things worked and some things didn’t. As soon as we identified the pieces that were not working, we started thinking through new hypotheses and potential improvements. After all, if you know something isn’t working, the only dumb thing is to keep doing it again and again.

As you can see, with each change, some things worked and some things didn’t. As soon as we identified the pieces that were not working, we started thinking through new hypotheses and potential improvements.

This Scientific Process has enabled Brex to scale and overcome a lot. It’s helped us change how we work in a remote world; it’s enabled us to evolve decision-making processes (becoming written by default); it’s helped us constantly improve our organizational structure; and so much more.

There is no single “right” answer when building your company, so make your best guess, evaluate, and iterate. The best companies embrace change; the best companies develop a collective growth mindset; the best companies are consistently better today than they were yesterday. More than anything else, the Scientific Process makes me optimistic about Brex’s ability to overcome whatever new challenges we’ll face as a company! We hope it helps you overcome whatever lies ahead.