In data science, like many other fields, you learn more by doing than by reading or merely studying the theoretical aspects of the field. Early in your data science career, you’ll spend a lot of time learning coding, mathematics, statistics, algorithms, visualization and business basics.

Although all these concepts and topics are extremely important, knowing a field’s theory doesn’t mean you’ll have a grasp on putting that theory into practice. Sometimes, as a beginner, you tend to make mistakes that are, in reality, easy to avoid because you lack experience or because you just weren’t taught to avoid those particular mistakes.

Once you start building more projects and working on real-world problems with different teams and data sets, you’ll develop intuition about how to approach any problem, plan specific steps to reach the solution and solve any problem that comes your way. So, although you will find your own way to avoid mistakes through trial and error, why wait? 

I’ve been where you are and one of the best decisions I made was talking with many data scientists about what they wish they’d known earlier in their career that would’ve helped them progress faster and better. What I heard over and over was: You learn better by doing. But how can you make the doing even more productive? 

In this article, we’ll walk through nine common mistakes often made by early career data scientists or data science students (and sometimes experts) that lead to false results or cause the project to take a much longer time to finish. 

9 Common Mistakes Data Scientists Need to Avoid

  1. Not having a plan
  2. Choosing the wrong visualizations
  3. Failing to consider bias in the data
  4. Neglecting to optimize the model for your data
  5. Focusing on accuracy over performance
  6. Ignoring that correlation doesn't equal causation
  7. Reusing implementations
  8. Choosing the wrong tools
  9. Forgetting the business


1. Not Having a Plan

Let’s start things off with the most commonly made mistake: launching into a project without having a plan of attack. Often, when we are given a data science problem, we need to answer why the data behaves the way it does and what story it’s telling us. To answer those questions, we need to be clear about our methodology. In other words: What are the questions we’re trying to answer and how will we go about answering them? Jumping into a problem without some kind of strategy or roadmap is a recipe for getting lost pretty quickly. 

More From Sara A. Metwalli4 Essential Skills Every Data Scientist Needs


2. Choosing the Wrong Visualizations

Choose your visualizations wisely. Visualizations are important in all stages of the project. For example, they’re critical in data exploration and help you either spot patterns or trends. On the other hand, bad visualizations can make you miss those trends completely. So, make sure you know what visualization tools are available, what graphs and charts you can use, and which one will best describe your data to help you understand it better.

Data Viz Tutorials on Built InForget Bokeh. Use Pygal to Create Data Visualizations Instead.


3. Failing to Consider Bias in the Data

In data science, there’s a famous saying: Your results are only as good as your data. Unfortunately, we don’t often don’t have a say in how or where the data is collected. That’s why, when we set up steps to solve a problem using a data set, we need to consider whether there’s inherent bias present in the data or whether it’s a good representation of the entire population. Doing so helps us avoid ending up with skewed models.

Related ReadingAn Introduction to Bias-Variance Tradeoff


4. Neglecting to Optimize the Model for Your Data

Your model has to be optimized for the data you have and follow the change in data over time. In machine learning, this falls under optimizing the values of your hyperparameters to reach peak performance. Optimizing your model is not just a one-time step; often, every time your data changes or a change occurs in the data itself, you will need to go back and modify your parameters to fit that change.

More From Built In ExpertsA Primer on Model Fitting


5. Focusing More on Accuracy Than Performance

This mistake is the one we all have fallen for at some point in our careers. Accuracy is important, but it is not the only factor of a good model. The accuracy of your solution depends on the algorithm you chose, the data you’re working with and the parameters you set. Changing any of these things will affect the accuracy of your results. So, focus more on interpreting your data; accuracy will follow. 

More on Accuracy and PrecisionWhen Accuracy Isn’t Enough, Use Precision and Recall to Evaluate Your Classification Model


6. Ignoring That Correlation Doesn’t Equal Causation

Correlation and causation are two very different things, but sometimes we tend to connect them, not just in data science projects but also in our personal lives. Correlation is a statistical technique that is used to refer to the existence of a relation between two variables or two factors. But, just because a relationship exists doesn’t mean that relationship is causal. So, test the data before jumping to conclusions.


7. Reusing Implementations

Here’s another common mistake: When we spend a lot of time working on a project, developing a methodology, and optimizing a model, we may assume that the model can be applied to similar problems with no alterations. Unfortunately, this is rarely the case. Each problem has its own variables and needs a custom-made solution. So, avoid reusing implementations for different problems.


8. Choosing the Wrong Tools

This is an easy mistake to make, even for the most seasoned of us. Today, there are what seem like an infinite number of tools that can help you with the different stages of implementing a data science project. But, because of the cornucopia of options at our fingertips,, we may choose the wrong tool or end up using too many tools. So, take some time in the planning stage to choose the best tools for the project. It will save you a lot of time and effort in the long run.

On That Note . . .4 Python Tools to Simplify Your Life


9. Forgetting the Business 

Data science is an interdisciplinary field; it covers a wide range of applications and scenarios. In the end though, all professional data science comes down to making informed business decisions. Always take time to understand how and why the data was collected and how the insights you find will be used by the business. Remember, making the wrong decisions in data science can cost millions.


The Takeaway

When I first started my journey toward becoming a data scientist, it took me a long time to grasp the field’s basics: mathematics, statistics, data visualization, communication, all alongside business fundamentals.

While I learned a lot by going through tutorials, online courses and textbooks, my real education came through my first year of actually building data science projects, working with other data scientists and exploring the different applications of data science. Through these interactions, I have learned to avoid many mistakes just to have a more efficient workflow. And let me be clear: I made a lot of these mistakes. Hopefully, now you won’t have to!   

Expert Contributors

Built In’s expert contributor network publishes thoughtful, solutions-oriented stories written by innovative tech professionals. It is the tech industry’s definitive destination for sharing compelling, first-person accounts of problem-solving on the road to innovation.

Learn More

Great Companies Need Great People. That's Where We Come In.

Recruit With Us