How to Become a Data Scientist: Skills, Path and Reality

One of the questions I hear most often from students and from professionals thinking about a career change is: “Where do I start if I want to become a data scientist?” It sounds simple, but the answer is not a list of courses. Data science sits where statistics, programming, domain knowledge and communication meet, and whichever of those you are missing shows up very quickly once you start the job. In this post I will go through the skills you need, a sensible order to learn them in, portfolio projects that actually help, and an honest picture of what junior roles look like.

What does a data scientist actually do?#

The U.S. Bureau of Labor Statistics (BLS) describes the job as using analytical tools and techniques to extract meaningful insights from data. The day-to-day duties it lists include collecting and cleaning data, building and testing machine learning models, visualising findings and making recommendations to decision-makers.

Even the order of that list is instructive. Many people picture the “building models” part; in practice a large share of the time goes into understanding where data comes from, sorting out missing or wrong records and pinning down the right question. As someone who works on data quality, I can say this with some confidence: what sets a good data scientist apart is rarely knowing the most complex algorithm. It is the habit of asking what the data represents, and what it does not.

Core skills#

The BLS list of important qualities includes analytical skills, computer skills, communication, logical thinking, maths and problem solving. In practice they group into four areas:

People are sometimes surprised that SQL ranks so high. But in your first job, the early weeks will very likely be spent learning to pull the right data from the right tables in the right way. A simple example:

SELECT city,
       COUNT(*)    AS order_count,
       AVG(amount) AS average_amount
FROM orders
WHERE order_date >= '2026-01-01'
GROUP BY city
ORDER BY order_count DESC;

Writing that query is easy. The real skill is looking at the result and asking: “Were cancelled orders counted? Does the same customer appear twice?”

A sensible learning order#

Everyone’s path is different, but this order works for many people:

1. Basic statistics and thinking with data. Mean versus median, sample size, correlation versus causation.
2. One programming language. Going deep in Python or R is worth more than skimming both.
3. SQL. Practise on a realistic dataset with several related tables.
4. Data cleaning and exploratory analysis. Missing values, outliers, inconsistent categories.
5. Core machine learning. Linear and logistic regression, decision trees and, above all, evaluating models properly.
6. Presenting and storytelling. Getting a finding across with one chart and a few sentences.

On education, the BLS says that in the US this occupation typically requires at least a bachelor’s degree in mathematics, statistics, computer science or a related field, and that some employers prefer a master’s or doctorate. That is a US description, and a degree is not the only door: people from physics, economics, engineering or the social sciences do move into the field after strengthening their statistics and building a solid portfolio.

Portfolio: what should it show?#

A good portfolio project shows how you think. A few suggestions:

• Start with your own question. Instead of the datasets everyone uses, find open data on something you care about: local government, transport, air quality, sports.
• Go end to end. Getting the data, cleaning it, analysing it, drawing conclusions and stating limitations. Two finished projects beat five half-finished ones.
• Write down the limitations. Saying “this data does not cover these groups, so the result does not generalise to them” is a sign of maturity.
• Make it readable. A tidy repository, a short README and one paragraph written for a non-technical reader.
• Do not obsess over accuracy. Compare your model with a simple baseline, such as always predicting the most common class. That comparison says far more than a single impressive number.

The reality of a first job#

I am including this section because the gap between expectation and reality discourages a lot of people early on.

• Job titles are messy. What one organisation calls “data scientist” another calls “data analyst” or “machine learning engineer”. Read the duties, not the title.
• The entry door is often analytics. Starting with reporting, SQL and dashboards and moving towards modelling over time is a common route.
• There is competition. The field is popular, and entry-level postings often attract many applicants. Showing interest and knowledge in a specific sector, such as health, finance, retail or energy, can help you stand out.
• Demand is visible but local. The BLS projects US employment of data scientists to grow 35% between 2025 and 2035. That is a projection for one country and should not be read as a forecast for Turkey or any other market.

I am deliberately not quoting salary figures. They vary hugely by country, city, sector and experience, and numbers without a clear source and date tend to mislead.

Frequently asked questions#

Sources

• U.S. Bureau of Labor Statistics – Occupational Outlook Handbook: Data Scientists ↗
• U.S. Bureau of Labor Statistics – Career Outlook: Data occupations (2023) ↗
• Google for Developers – Machine Learning Crash Course ↗
• American Statistical Association – Statement on Statistical Significance and P-Values (2016) ↗