April 5th, 2022
In May 2021, I walked into my manager’s office (virtually of course) and resigned from the only place that I’d ever worked. For almost three years, I’d spent my time building NLP models and machine learning solutions as a member of Capital One’s Center for Machine Learning. I built models that helped call centers, read social media posts and answered questions all at a scale and rigor that only exists at Fortune 100 company.
As of February of 2022, we’ve built (and rebuilt) countless models, deployed a dizzying amount of infrastructure and written enumerable lines of code. Just like my previous job! But also... not at all...
So, you may find yourself in the same shoes that I was in 9 months ago, or maybe you prefer my co-founder Michael’s pair. Regardless, one thing is clear, the job is all about behavior. There are some worth keeping, some you must leave behind and some you didn’t even know you needed.
And the associated caveats...
Time is a precious asset for a founder, and while time is still an important asset at a larger organization, the practice of building automation for the most repeatable processes is incredibly valuable and imperative for a startup. Investing in easily extended and repeatable infrastructure deployments saved us countless hours of tinkering in the console. While the initial investment may have been time consuming, we knew this would be something we’d use often. (Most startup founders will tell you they release ALL the time, we are no exception).
With that being said, there are moments where some tech debt and elbow grease are the superior option. The golden rule for us came down to the concept of repeatability. If you’re going to do it again...and again...and again, make the investment!
There’s this unique phenomena among founding teams where you’re incredibly tight knit, burrowed deep in the foxhole together, yet, can often find yourselves operating in deeply individual silos. The reasoning behind this phenomena is that there is simply so much to do, that you need to divide (and communicate) and conquer.
As a result, it’s absolutely necessary to construct a system that provides checks and review opportunities due to the relative independence of work. Review each others code, implement a change management process and require approvals before builds and deployments. Going a step further and implementing automated checks (unit testing based status checks, code scans, etc) will all ultimately help you reduce the number of times you release defective code to production. You’re not going to eliminate releasing products that have a couple bugs here and there, but catching even a single defect through your review systems makes the entire process worth it. Anything to make the customer experience better.
One thing that happened at previous roles, almost to a fault, is planning paralysis. Often due to the scale of the initiative, and the number of participants involved on the project, it was important to deliberate and thoroughly plan out new initiatives.
Planning is an important practice even at the start up level, but there’s one major mindset shift that you must remember at all times: you need to plan at the speed of light. The value that planning provides holds true for a startup as well: you avoid obvious speed bumps, better understand the level of effort and can more effectively prioritize. However, you’ll quickly find that in the early days you gather so much game changing feedback from your customers that even the best laid plans can quickly be destroyed. So, keep planning by all means, but just remember that speed is your friend and the best thing you can do for your company is to deliver product to customers and gather feedback.
To a degree, there’s this convenience at a larger organization, especially one with product market fit. The customer is well understood, their needs and pains defined, and the objective well known. Previously, working on a new project or use case began with a clear definition of the problem at hand, and an obvious eventual consumer for the developed solution.
You are offered no such stability with a small startup. Rather, you must quickly adapt to the experience of existing in the unknown. The art of product building at this stage is trying to light as many candles as you can in this pitch black room, each new candle, a new insight leading you towards product market fit.
Existing in this unknown has only one path out. Deliver fast. Collect feedback. Iterate.
This is a hard shift to establish, and one that I struggled with. Coming from a larger organization with projects with well defined scope, it’s easy to build confidence that you know what to build. This mentality must be avoided and eliminated in the startup game. Only once you accept this can you create opportunities to listen to your customer, as many as possible, and trudge towards product market fit.
Your first 5, 10 customers are not customers. They’re not buying your product, they’re purchasing your team and your vision. This relationship is fundamentally different (and for good reason) than anything found in industry. As a result, the way you deliver, the way you interact, the way you reveal and the way you collaborate has to be completely different. Some things we learned:
Be transparent. If something isn’t working, you’ll be surprised how much your customer can help. They bought the vision, they know the pain, and their perspective is something that can, at times, move mountains for you.
Go above and beyond. This sounds obvious, and sounds like something that you’d expect/want at any company, at any size. But it doesn’t happen. We lived and died by the statement of work, but at this smaller stage, you have to remember this relationship means more. So...do more.
There’s this comfort with delaying delivery that creeps up in larger organizations. It’s not a negative - the desire to deliver perfection is often the motivation. To avoid defects and satisfy requests, there were many times we delayed delivery in an effort to squeeze in that last update. This mentality will kill your startup.
Why? The pursuit of perfection reduces, if not eliminates feedback opportunities. And, if there’s one thing we’ve learned in the last year, it’s that feedback is the lifeblood of a start up. It’s uncomfortable, but learning to live with that tech debt, with that flaw and being okay with the fact that you’ll get to it will let you iterate faster, deliver quicker and ultimately, build better!
And so, the marathon continues. It’s been a hell of a transition, and somethings are still changing, but we’re going to keep running!