If you want one skill that opens the most doors in data work, it is SQL. It appears in analyst, marketing, operations, finance, product and engineering job postings, it is stable across decades, and you can reach a useful level in a few weeks of steady practice.
What SQL actually is
SQL is the language for asking questions of data stored in tables. “Show me customers who ordered twice last month, grouped by region” is one query. Almost every company keeps its important data somewhere a SQL query can reach.
You do not need to install anything to start; several free sites give you a database in the browser.
Learn it in this order
Week 1: reading data
- `SELECT` and `FROM`: choosing columns and tables
- `WHERE`: filtering rows
- `ORDER BY` and `LIMIT`: sorting and sampling
- `NULL`: why empty values behave strangely
Week 2: grouping and combining
- `GROUP BY` with `COUNT`, `SUM`, `AVG`
- `HAVING` versus `WHERE`
- `JOIN`: inner and left joins, and why a join can multiply your rows
Week 3: shaping answers
- `CASE` statements for categories
- Dates: truncating to months, comparing periods
- Subqueries and common table expressions (`WITH`)
- Window functions: running totals, rankings, period-over-period change
Week 4: doing it for real
- Take a messy public dataset and answer one genuine question
- Check your results against a second method
- Write up what you found in plain language
That sequence covers most of what analyst job postings ask for.
Where to practice free
- Kaggle Learn has short SQL courses with exercises in the browser.
- SQLBolt teaches the basics interactively from zero.
- freeCodeCamp covers relational databases alongside wider data work.
- Public datasets from your city or national statistics office give you something real to query.
Our guide to free ways to learn online lists more, and building a learning habit helps you keep the sessions going.
The part beginners underestimate
Writing a query that runs is easy. Writing one that is correct is the skill. Before trusting any result:
- Check row counts before and after a join.
- Look for duplicates created by the join.
- Ask whether `NULL` values are being silently excluded.
- Sanity-check totals against a known number.
- Compare to a simpler query you trust.
Employers hire for this habit. An analyst who catches their own errors is worth far more than one who is fast.
Using AI while learning
AI assistants write SQL well, which makes them useful and dangerous while you are learning. Use them to explain errors and suggest approaches, then write the query yourself and verify the output. If you cannot explain what a query does line by line, you do not yet know it, and an interview will reveal that quickly. Our guide on AI skills in interviews covers how to talk about this honestly.
Showing it to employers
Put one or two finished analyses in a portfolio: the question, your queries, what you found and what you would do next. Our guide on portfolio projects covers the format, and our comparison of data analyst and data scientist roles helps you decide which direction to aim for.
Frequently asked questions
How long until I can use SQL at work?
Most people can write useful queries after two to three weeks of regular practice, and become genuinely independent within a few months of real use.
Which database should I learn?
PostgreSQL is a good default and widely used. The core language transfers everywhere; only the edges differ between systems.
Do I need Python too?
Not to start. SQL plus spreadsheets is enough for many analyst roles. Add Python when you need automation, statistics or larger-scale work, as our guide to AI skills employers want describes.
