When you work as a data scientist, you get good at solving well-defined problems, model this, predict that, clean the data, ship the report. But when you try to build something real, something that people can actually use, you realize how far that world is from the one you’re used to. I’ve been thinking a lot about that gap lately.
In data science, you usually start with data. In product building, you start with people. That sounds obvious, but it’s a big shift. Most data projects begin with a dataset. Most products begin with a person, a frustration, a problem, a workflow that doesn’t quite work. When you’re building something for others, you can’t just ask “What can I model?” You have to ask “What’s worth solving?” And people rarely describe their problems as data tasks.
Then there’s the reality that models don’t matter if nobody uses them. You can have the cleanest model, the most elegant code, perfect accuracy, but if it never reaches someone’s hands, it’s just an experiment. In a job, there’s usually a handoff: “I built the model; now someone else will deploy it.” When you’re on your own, you are that someone else. You have to think about interfaces, usability, speed, even hosting costs. It’s humbling but also freeing.
Building products teaches you to think smaller, faster, and more focused. In data work, you tend to chase precision and completeness. In product work, you chase feedback. You don’t need the perfect model; you need a useful one that someone can try today. So you start trading accuracy for speed, complexity for simplicity, polish for learning. That’s where the real learning happens.
Bridging that gap means unlearning some habits. Waiting for the perfect dataset becomes starting with what you have. Thinking in models becomes thinking in workflows. Optimizing metrics becomes optimizing experience. Measuring success in precision becomes measuring in adoption. It’s a different game and it stretches how you think. I’m still at the beginning of this process. I wouldn’t call myself a “product person,” but I’m trying to become one, one small experiment at a time. If you’re a data scientist who’s ever felt this same tension, between analysis and action, between code and creation, then you’ll probably understand what I mean.
I’m starting to see that this gap isn’t just about tools or skills, it’s about mindset. The more I build, the more I realize how much there is to learn beyond the data. If you’ve been thinking about crossing that same line, maybe this is your reminder to just start small and see where it leads.
Thanks for reading and see you soon!