Which database should I learn first?

Any mainstream relational one. PostgreSQL is a good default because it is free, widely used and strict enough to teach good habits. What transfers is the modelling and the query thinking, not the dialect.

Is SQL worth learning if I am not a developer?

It is worth more, if anything. Analysts, marketers and operations people who can answer their own questions stop waiting in a queue, and that independence is visible in a way few other skills are.

How long until I am useful?

A few weeks to write queries that answer real questions, a few months to write them efficiently against tables you did not design, and rather longer before you can design tables someone else will thank you for.

Do I still need this if AI writes the query?

Yes, because you have to know whether the query it wrote answers the question you asked. Generated SQL is plausible by construction, and a join that silently drops rows produces a confident wrong number.

What are window functions and why do interviews ask about them?

Calculations across a set of rows related to the current one, without collapsing the rows the way grouping does: running totals, rankings, the previous row's value. Interviews ask because they separate people who have written reports from people who have only filtered tables, and because every modern engine, PostgreSQL included, supports them.

Databases & SQL Courses

Model, query, and tune the data layer - from your first SELECT to production-grade database design

SQL was designed in the nineteen-seventies, has survived every technology that was going to replace it, and is still the thing a hiring manager tests when they want to know whether you can work with data. That longevity makes this the best-value section in the directory: almost nothing you learn here will be obsolete in five years, which is not true of any other section.

Where the difficulty sits

Selecting rows is easy and everybody stops there. The parts that take work, and the parts interviews probe, are joins that do what you meant when rows do not match one to one, aggregation with grouping you can reason about, window functions, and the moment you have to explain why a query that ran in a second now takes four minutes. That last one is where a database course separates from a SQL tutorial: indexes, query plans and the cost of a badly shaped table are database knowledge, not syntax.

Relational first, then the alternatives

Engine	Why learn on it	Where you will meet it
PostgreSQL	Free, strict, feature-complete; teaches good habits and punishes sloppy ones	Most new applications, most analytics stacks, Getting Digital's own backend
MySQL and MariaDB	Free, forgiving, everywhere on the web	WordPress, older web applications, shared hosting
SQLite	A single file, no server; the fastest way to practise	Mobile apps, desktop software, embedded systems, tests
SQL Server	Microsoft's engine, with its own dialect	Corporate environments already on Microsoft; the Power BI courses lean on it
Document, key-value and graph stores	After the relational model, by contrast with it	Systems that outgrew one table shape, and usually run a relational store alongside

Learn a relational database properly before exploring document stores, key-value stores or graph databases. The relational model is the shared vocabulary of the field, the alternatives are usually explained by contrast with it, and most systems that adopted an alternative in a hurry now run both. Which relational engine matters far less than the depth you reach in one: the dialect differences are a weekend, the modelling instincts are years, and PostgreSQL is the default for a learner who has no employer choosing for them.

Design is the part courses skip

Normalisation until you understand it, then denormalisation when you can justify it. Both are decisions, not doctrines.
Keys and constraints as the database's own defence against bad data. Applications forget; constraints do not.
Transactions and what isolation levels promise, which is where surprising bugs are born.
Migrations: changing a schema that already has data in it, without downtime, is the skill nobody teaches and everyone needs.

Data Lake

A data lake is a store that keeps data in its original form, files of any shape in cheap object storage, loaded before anyone has decided what questions will be asked of it, so that the structure is applied when the data is read rather than when it is written.

The field, explained: Data & Analytics

This category belongs to the field Programming and Software Development, which maps the topics, concepts and certifications behind these courses.

Frequently asked

Which database should I learn first?: Any mainstream relational one. PostgreSQL is a good default because it is free, widely used and strict enough to teach good habits. What transfers is the modelling and the query thinking, not the dialect.
Is SQL worth learning if I am not a developer?: It is worth more, if anything. Analysts, marketers and operations people who can answer their own questions stop waiting in a queue, and that independence is visible in a way few other skills are.
How long until I am useful?: A few weeks to write queries that answer real questions, a few months to write them efficiently against tables you did not design, and rather longer before you can design tables someone else will thank you for.
Do I still need this if AI writes the query?: Yes, because you have to know whether the query it wrote answers the question you asked. Generated SQL is plausible by construction, and a join that silently drops rows produces a confident wrong number.
What are window functions and why do interviews ask about them?: Calculations across a set of rows related to the current one, without collapsing the rows the way grouping does: running totals, rankings, the previous row's value. Interviews ask because they separate people who have written reports from people who have only filtered tables, and because every modern engine, PostgreSQL included, supports them.

Why Learn Databases & SQL?

SQL as a career skill: querying and joining with confidence, designing schemas that survive real data, and administering PostgreSQL, MySQL, MongoDB, and the systems behind almost every application. Useful to developers, analysts, and anyone whose job touches data.

Last reviewed 26 September 2026 · Getting Digital