DBT

dbt, short for data build tool, is an open-source command line tool used to transform data inside a data warehouse or data lakehouse. It supports the modern ELT workflow by allowing analysts and engineers to write modular SQL transformations that are executed within the target warehouse, rather than extracting data to an external processing layer. dbt emphasizes data quality, version control, and governance in analytics workflows.

Core concepts in dbt revolve around models, tests, and documentation. A model is a SQL file that defines a transformation; when run, dbt compiles these models into executable SQL and builds a directed acyclic graph of dependencies. Materializations determine how results are stored, with options such as view, table, incremental, and ephemeral (the latter creates no physical object). Projects are configured in a dbt_project.yml file and connect to warehouses through profiles, typically stored in profiles.yml.

dbt supports a range of features to manage the analytics lifecycle. Seeding allows loading static data from CSVs, and snapshots enable tracking slowly changing dimensions. Built-in and custom tests enforce data quality and integrity. Documentation can be auto-generated from code, exporting a data dictionary and lineage information. dbt packages enable sharing of reusable code, macros, and configurations.

Ecosystem and usage vary by deployment. The core project (dbt Core) is open source, while dbt Cloud provides a hosted service with scheduling and collaboration features. dbt integrates with major warehouses such as Snowflake, BigQuery, Redshift, Databricks, and Postgres, among others. The tool is widely adopted by analytics engineers to enforce repeatable, testable, and auditable data transformations.