Building the Astra Beta: Lessons on bringing your first product to life when taking on a big problem
Breathing life into your startup’s first product is hard. You may have first-hand knowledge of the pain point, and a clear vision for what you are striving for, but put plainly—how do you just get started? What is the guide for deciding what that first thing should be, and then how do you build it? In my experience, there isn’t a one-size-fits-all solution. At Astra, we tried just about all of the startup models, synthesized what we learned, and mapped out what we wanted to build. Then, we realized that the problem we were taking on was really big, our vision required innovative technology, and our very first product required users to willingly share a lot of sensitive personal data. Now what?
In the first article in this series, I shared the origin story of our company, Astra, and how deploying Deep Learning enabled us to get our idea off the ground. This article shares our approach to developing our first application, why we felt the beta needed to be real, and how we decided to develop it with Udacity Blitz — a partnership that solved our “Now what?” question and brought our app to life.
What we knew
In the months leading up to founding Astra, while we were just beginning to truly understand the power of Deep Learning when applied to personal finance, my co-founder Sam and I were also avidly researching our market. Our personal experiences with finance instilled a passion in us for the subject, but our research drove home the scale of the problem. Even though understanding these numbers is a daily exercise for us, it is still challenging to fully grasp them.
The numbers are staggering across a wide range of measurements:
- 46% of Americans can’t afford a $400 unexpected expense without using credit
- The US is projected to have a $137T retirement savings shortfall by 2050
- Household debt back has returned to an all-time high of $13T (remember what happened in 2008?)
This only made us more enthusiastic about taking on the problem. Our debt and savings numbers paint a dark path ahead, one that requires an immediate and impactful solution. With this research in hand, we knew that the problem of personal finance—and more specifically, optimizing cash flow—is a BIG one, and there should be urgency in addressing it.
So we had a general technology roadmap focused on financial intelligence, bolstered by learning we had done with Udacity. We had a lens into the challenges we all face with our finances. And we had some ideas about how we could start to solve this big, tough problem with an app — but that didn’t mean we knew with much confidence where to begin.
How to start?
Astra had a north star: fix our cash flow challenges to improve our financial health. But we didn’t have a clear path on how to get from mission to product. We read the standard array of startup books, watched a multitude of startup videos, and scoured the web for startup insights. Our conclusion was that whether it’s your first or your nth startup, the best model for how to make your product for your problem for your users has to be yours.
If that sounds like “you’ll figure it out along the way,” that’s partially true; however, low overhead tools exist that can help you define your path to that first product. We tried lots of these, from the Business Model Canvas to “getting out of the building” and going to TechCrunch Disrupt armed with a Typeform survey to gather more details about users who might test our still-to-be created app. Then we created a click-through prototype with ProtoIO to gather more specific feedback about our hypothesis for what would be valuable to users. These exercises were useful. We learned a lot about how our users interact with their money, but we didn’t really have a Minimum Viable Product.
What we learned
What we did have was an idea of what that could be and a pair of somewhat contradictory lessons from our startup process research:
Learning is the essential unit of progress for startups.
—Eric Ries, The Lean Startup
First, it’s more important to be engaged in the process of continually learning than over-revving about the thing itself. The canvas, the survey, and the prototype were all mechanisms for advancing our learning. That was more valuable than knowing for certain that anything along the way was the MVP. Lesson: Be reassured when you are making progress.
I don’t believe in statistics. I believe in calculus.
—Ben Horowitz, The Hard Thing About Hard Things
Second, as a founder there comes a time when you have to commit to what you are building and believe there is a deterministic future in which it comes to life. What is the thing that you going to spend your energy, or your money, or your investors’ money, developing? You can’t just answer that you are learning. The decision comes with weight and needs to be clear. Lesson: Amassing your learning units helps bolster confidence in your commitment to that future, but at some point, you have to say what it is, and build it.
If at this point as a founder you are confident that you can continually find ways to learn and articulate what comprises your startup’s first offering and why, then you are on your way. In our case, I felt confident about the first, but the second gave me cold sweats — not because defining our MVP was too hard, but because of what it meant.
Our first product had to be real and secure and prove our application of Deep Learning to the problem. The core technology was data-hungry, and that data was sensitive — if not taboo. Seeing it work live with real data in our early tests was eye opening. It offered us the “aha” moment, and the confidence to go against the advice of most of our startup research, and build something that could barely be described as minimum anything.
This meant our process more likely placed us in the Anti-lean portion of the startup spectrum, with longer timelines and more resources required. And we had to be ok with that. We had to figure out how to build it with all the financial constraints and traction challenges of a new company, while still needing validation of our initial assumptions.
So what were the options?
To build our beta, Sam and I worked up a target timeline and compiled options for either expanding our team or outsourcing a portion of the development to an agency. Given our timeline and location in the Midwest, assembling talent with the right skillset (including Deep Learning) for Option A was going to be really slow. And given the range of skills we would need to pull it off, Option B was going to really expensive, if we could even find the right agency. Luckily, we were also headed to TechCrunch Disrupt, where we serendipitously struck up a conversation with the representatives at the Udacity booth.
The team had a Udacity-branded autonomous vehicle, and I was most of the way through my Machine Learning Nanodegree program, so we were eager to say hello. We started talking with Colin Lernell about the impact the Nanodegree program was having for our startup trajectory and the challenges of being left with two less-than-ideal options for establishing a team to build our beta. Sure, there were ranges of solutions for either path — hiring one or two perfect-fit engineers into the core team versus a few contractors, or outsourcing to a high-profile agency versus an overseas outfit — but none of them had enough upside for cost or risk. Colin shared that he was actually working on a solution for just that problem, and it was going to be powered by graduates of Udacity’s Nanodegree programs.
Choosing Udacity Blitz
Learning is natural activity for Sam and I, and it’s key to the culture we want to establish at Astra. After all, it’s what our technology does. Deep Learning is on the frontier of innovation in technology and isn’t a typical competency for full-stack digital agencies yet. Across all of the options we vetted to develop our beta, none could compare to Udacity Blitz on mission alignment and technology acceleration. We had an opportunity to work with a team with front-end, back-end, full stack, and machine learning engineers. Even though it seemed like a natural fit from the beginning, we finished our due diligence by working up budgets and gathering quotes for our other options. Across multiple metrics — cost, risk, speed, and skillset — Udacity Blitz was our Goldilocks solution. It was just right.
We decided to partner with Colin, and the new Udacity Blitz team spun up to develop our beta. The process was a true collaboration supported by a contemporary stack of tools that foster productive interaction — Slack, Github, Kanbans — and our technical project managers were able to scale the team up dynamically. Strong talent was there from Udacity’s graduates to draw from if we needed a big push. And each of the team members really cared about what we were developing together. We had some tough challenges along the way, and they brought their technical skills and creativity to find good solutions.
The process of founding a company and creating your first product is hard. Remember that each startup is unique so plenty of this may not apply, but here is a recap of lessons we learned and that I think are crucial particularly when your problem is big:
- Continue to make progress and learn. This doesn’t have to be exclusively through your MVP.
- Decide on what that MVP will be and bring it to life in the most effective way possible.
- Find good partners, especially if your development plan requires a high degree of complexity. I recommend everyone in this situation to start their search with Udacity Blitz.
- Keep an eye out for Serendipity. If we hadn’t been actively learning through Udacity, and attended Disrupt, we might still be struggling with how to make it real.
As we approach the conclusion of our project, we have continued to learn as we create our first product, we have an established relationship with an innovative development partner, and we are launching our Early Adopter Program with a waitlist of eager beta testers. Whether our startup is Anti-lean or not, we definitively know our problem is big, our challenges for proving our technology are significant, and most importantly, that choosing Udacity Blitz as our development partner was our best option for making our first product real.
Want to test the beta we developed with Udacity Blitz? Join our Early Adopter Program.
Follow me to receive a notification about the next chapter in this series about our startup experience.