How To Build a Product When There Is No Problem

Problem

We were part of a startup team acquired by an enterprise and we were trying to understand how we could bring the technology that we had built in the machine learning space into the enterprise’s value chain. We had a mandate and were immensely motivated, but had to figure out what and how to build, how to test it, and validate it. Briefly, we were tasked to solve a blank problem within the boundaries of the existing IT infrastructure.

Actions taken

Considering the breadth of the problem space and the multitude of solutions, my first action was to define what product could be built and what were the technical pieces we could reuse from the existing infrastructure and avoid building things from scratch.

The challenge was to define-- together with a PM -- the narrow product boundaries in the vain of MVP and establish how we could accurately measure and roll it out. This is something small teams should pay particular attention to -- they would need to build something tangible that would bring value to the business. Having limited resources they should start small with a narrowly defined problem. The larger teams usually have some more padding to do more experimenting and meandering around.

As an engineering lead, I had to work with a PM and come up with an initial problem, then take the requirements they developed and also work with Engineering to see what we could use from the existing infrastructure. We did a quick survey of the enterprise’s architecture identifying pieces that we could use and that would help us build a hybrid model.

I still had to write the code, but more importantly, I had to intensively collaborate with the enterprise’s IT team and delivered clear guidelines for what we needed from them.

Lessons learned

• You need to define as narrow as possible boundaries of a new product. For a small team, it should be something you can deliver in a month or six weeks.
• You have to keep at bay the requirements from the business otherwise you will end up building overly complicated things that in the end may not bring any value.
• The hybrid model can work really well if you establish good collaboration with your peers outside of your startup. If you are not ready to negotiate or are quick to pull rank, no one would want to work with you and you will end up as a part of a losing proposition.
• You don’t need to build everything from scratch. Don’t chase the latest and greatest if there is an existing technology that is reliable.
• Build what is your core competency or things that are unique about your technical platform but everything else you should get from the open-source or through finding a vendor who you will pay to build it. A lot of engineers have nurtured the building mindset prior to becoming managers and their first impulse is to build software by themselves to solve a problem. The reality is that you can’t build all the software and why doing so when most of it already exists. Your job as a manager is to economize and see how you can reuse or repurpose the existing software.