dbt has become the default way analytics teams transform data in a warehouse, and dbt Learn is the official training. It is free, well constructed, and unusually good at teaching a way of working rather than a set of features. Start with what makes it different. Most analytics training teaches you to produce a result.
dbt Learn teaches you to produce a result that is version controlled, tested, documented, and reproducible by someone else. Models live in a git repository. Tests run automatically and fail the build when an assumption breaks. Documentation is generated from the code and the lineage graph is derived rather than drawn by hand.
These are software engineering practices applied to analytics, and analytics has historically been terrible at all of them. Anyone who has inherited a folder of untested SQL scripts with names like final_v3_actual.sql will understand why this matters. dbt Fundamentals is the entry course and it is about five hours. It covers project structure, sources, models, refs, materialisations, tests and documentation, working in an actual warehouse against an actual dataset.
The pacing is good, the exercises build, and by the end you have a working project you understand. It is one of the better free courses in the data space, vendor or otherwise. Beyond Fundamentals there are modules on Jinja and macros, on advanced testing, on deployment and orchestration, and on the semantic layer. Quality varies a little.
The Jinja and macros material is important because dbt's templating is where projects get powerful and also where they get unmaintainable, and it could go deeper on when not to reach for a macro. The prerequisite that catches people is SQL. dbt is a SQL tool. If you are not comfortable with joins, CTEs, window functions and aggregation, you will spend the course fighting the language rather than learning the framework.
Do a proper SQL course first. The training does not really warn you about this and it should. Warehouse setup is the practical friction. dbt needs somewhere to run, and while free tiers exist on BigQuery and Snowflake trials are available, getting credentials configured and a connection working is genuinely the hardest part of starting for an individual learner with no employer account.
Budget an evening for it and do not conclude you are bad at this when it takes longer than expected. The dbt Cloud alignment is worth noting. dbt Core is open source and dbt Cloud is the commercial product, and the training leans towards the Cloud experience. Everything conceptual transfers to Core, and some of the workflow specifics do not, and if your team runs Core with its own orchestration you will need to translate.
That is a normal vendor training compromise and it is handled more honestly here than in most. On the certification, $200 for the Analytics Engineering certification is a real cost and its value depends entirely on your market. In an organisation that runs dbt it is a reasonable signal. Outside that, it means little, and the skills matter more than the certificate in every case.
I would do the free courses and skip the exam unless someone is paying for it or a job posting names it. The wider point is why I rate this as highly as I do. Analytics engineering has emerged as a real role sitting between data engineering and analysis, and the practices in this training are what define it. Even if you end up somewhere that uses a different tool, the habits stay: put transformations in version control, test your assumptions explicitly, document as you go, and treat data work as engineering rather than as a series of one off queries.
My four point zero is for free, well made training that teaches transferable discipline rather than button positions, marked down for the SQL prerequisite going unstated, the setup friction, and the pull towards the commercial product.