When I started learning Data Science, I had a lot of preconceived ideas. Some of them helped me move forward, others left me quite confused. This path is fascinating, but it's also full of noise, myths, and oversimplifications that aren't always helpful.
Today I’m sharing 10 beliefs I had at the beginning, and what I learned as I moved forward. It's not a list of “mistakes”, but rather key learning moments. Because in the end, messing up is part of growing as a data scientist.
1. I believed I had to master everything before starting to practice
I was paralyzed by the feeling of not knowing enough. I thought I had to master Python, statistics, visualization,** machine learning**… all of it, before tackling real projects.
What I learned is that you learn way faster by doing than by waiting to “feel ready.” Practicing from day one—even with doubts—is what actually turns you into a professional.
2. I thought a good model was the most important thing
I focused on algorithms, accuracy, techniques… believing that was the core of my value.
What I learned is that the model is just a part of the puzzle. Truly understanding the problem, cleaning and transforming the data with intention, and communicating insights clearly is what really makes the difference.
3. I believed plots were just visual extras
I made charts because “you had to,” without really thinking about their purpose.
I learned that visualization is analysis. Being able to see patterns, anomalies, correlations, or errors in charts changes your understanding of the problem entirely.
4. I thought machine learning and AI were the same thing
I used both terms interchangeably. They seemed like synonyms.
What I learned is that machine learning is only a part of AI. There’s symbolic AI, logic-based systems, expert systems… and now LLMs. Understanding the difference gives you valuable perspective.
5. I believed a data scientist was just someone who could code and build models
I focused only on the technical side, thinking that was enough.
What I learned is that you also need good judgment, b*usiness context, **critical thinking, and the **ability to explain complexity with clarity*. A true data scientist connects data with decisions.
6. I thought complex models were always better
I was fascinated by powerful models like XGBoost or neural nets, thinking they would always beat the simple ones.
I learned that sometimes a well-thought-out regression or a decision tree can be more useful, more interpretable, and even more accurate—especially if your data doesn’t justify the complexity.
7. I believed more data = better model
I saw large datasets as an automatic advantage.
What I learned is that if your data is messy, biased, or irrelevant, more volume only amplifies the problems. Quality matters more than quantity.
8. I thought machine learning was just statistics
I believed that if I understood statistics, I was halfway there.
What I learned: you also need engineering, software skills, validation, pipelines, reproducibility… and a whole lot of things that go beyond theory.
9. I didn’t know what data leakage was
When I got unrealistically good results, I just celebrated them.
What I learned: if you train with information you shouldn’t have used, your model is useless. Knowing how to separate train, validation, and test isn’t a technical detail—it’s critical.
10. I thought cross-validation always worked
I used it by default without considering the type of data.
What I learned is that, for example, with time series data, cross-validation can lead to totally wrong conclusions. Choosing the right validation strategy for your problem is part of the craft.
The bottom line is that every one of these points helped me grow, even if they frustrated me at first. And I’m sure I’ll keep changing my mind about many things, because that’s part of deep learning too.
I’m not sharing this to give advice, but to show that even missteps—if you reflect on them—can bring you closer to your best self as a data professional.
Top comments (0)