RWML #032: What does a Data Scientist do?
And how to get there
Despite the colossal amount of educational content on Data Science, it is hard to understand what the day-to-day of a professional data scientist looks like.
Let’s address this blind spot today!
The role of the data scientist 🧑🏽🔬
As a data scientist, you are essentially an “interface” between 2 teams:
The data engineers 👷
The business people 👩🏻💼
Your role is to build bridges (aka data products) that translate abstract data into high-quality business decisions.
The more you talk to both worlds, the more effective your work will be. However, each of these two teams speaks a “slightly different language”.
Talking to the product lead 👩🏻💼💬
Business stakeholders, like Product Leads, are focused on setting and hitting clear business outcomes. It would help if you talked to them regularly to ensure you solve the right problem for the company.
For example, your Product Lead says things like:
👩🏻💼: “We want to increase user retention by 5% by the end of this quarter”.
Cool. You know WHAT you need to solve.
Let’s now move on to HOW you can solve it. For that, you need relevant, high-quality data. Without high-quality data, you cannot measure retention, and hence you cannot measure your progress. Without high-quality data, you will fail.
Talking to data engineers 👷💬
Data engineers care for the infrastructure necessary to make high-quality data accessible to you. So they are your best ally at this stage.
Back-and-forth conversations between you and the data engineer are a MUST if you want to succeed as a data scientist.
Good conversations between data engineers and scientists result in concrete actions. For example:
let’s add Facebook third-party data to enrich user profiles, or
remove duplicate entries in the transactions table, or
make the data available to frontend dashboards.
Once you have high-quality data and a clear business outcome, you are ready to do your data science magic.
The “data science” magic 🧑🏽🔬🪄
Three ways of solving business problems using data are:
Building a dashboard with Tableau/Power BI 📊
Build a user retention dashboard that the Product Lead can use to break down this metric by relevant user properties (e.g. geo, age). Dashboards are a great way to keep the conversation flowing between product people and you. I personally recommend you start with this.Running a data exploration 🔎
Explore the data yourself to find the low-hanging fruit (aka quick wins). For example, you might find that certain Facebook campaigns bring low-retention users, so you ping the marketing team to stop them. Quick and easy win. I love these.
Training a Machine Learning Model 🏋️
Sometimes you need to bring out the big guns and use Machine Learning. For example, you could build a churn-prediction model, to identify customers who are likely to churn. With this info, the marketing team could send offers to these users, and keep them active.
My advice: Machine Learning is very tempting. But often, you do not really need to use it. Try #1 and #2 before resorting to ML.
My advice 🧠
Most people follow a course-based approach, where they start many courses (and complete a fraction of them). This is not what works best for me.
Instead, I suggest you learn by following a project-based approach.
→ Pick a problem you care about
→ Find data relevant to it.
→ Build a solution (either of the 3 mentioned above) and make it publicly accessible (e.g. GitHub).
And repeat.
Because the only way to learn data science is by solving data science problems.
Wanna build your first real-world ML project with my help?
Join the Real-World ML Tutorial + Community and get LIFETIME ACCESS to
→ 3 hours of video lectures 🎬 and step-by-step slides.
→ Source code implementation of an entire ML system 👨💻
→ Discord private community, to connect with me and 150+ students 👨👩👦
🎁 Gift
Use this direct payment link until next Friday and get a massive 40% discount ($90 instead of $150)
Have a great weekend
And keep on learning!
Pau