dbt Fundamentals

dbt (data build tool) is a transformation framework that lets analytics engineers write modular SQL models, test their outputs, and document lineage. It runs inside the data warehouse, turning raw data into trusted, version-controlled datasets through a software engineering workflow.

itData engineering and analytics

Don't Panic: dbt Fundamentals

dbt is the bit of the data stack that turns already-loaded data into models people can inspect, test, and reuse. It is not an extractor with a very stern expression, and it is not a dashboard wearing a SQL hat. It occupies the awkward but valuable middle ground: after data arrives in a platform, before somebody asks a report a question with consequences.

The unit of work is a model, usually a SQL select statement in a file. dbt compiles that statement and asks the data platform to run it. A materialization decides what remains afterward, such as a view that reruns its query or a table that stores a rebuilt result. The platform still supplies the SQL dialect, permissions, compute, and bill. dbt is not secretly a warehouse in a trench coat.

The memorable part is the DAG, the directed acyclic graph of dependencies. ref says that one project resource needs another. source says that a model depends on data an outside loading process delivered. Those references let dbt build in order, select related work, and draw lineage. A project with hardcoded relations can still run, much as a bridge can remain standing while the inspection paperwork is on fire, but dbt cannot see the missing connection.

Tests and descriptions are not decorative safety cones. A data test looks for records that violate an assertion, such as a missing key or an unexpected duplicate. No failing rows means the assertion passed. That proves only the assumption you wrote down, which is useful but not a licence to stop thinking about what a customer, order, or revenue row actually means.

Start with the Introduction when you need the whole arrangement. Use Slides for the flow from source to staging model to data product. Keep the Cheatsheet nearby when choosing a command, materialization, or first diagnostic. The Practice Reference and exercise turn the graph into a small working project. The Reference tab is where the current engine and platform details live, because software has an alarming habit of continuing to change after diagrams are printed.

Don't Panic: dbt Fundamentals

Relevant careers