What is dbt? The Data Build Tool, Defined

dbt (data build tool) is a SQL-based transformation framework that lets analytics engineers build, test, document, and orchestrate transformations against a cloud data warehouse using version-controlled SQL.

dbt (data build tool) is a SQL-based transformation framework that lets analytics engineers build, test, document, and orchestrate transformations against a cloud data warehouse using version-controlled SQL plus Jinja templating. It launched as an open-source project in 2016, was commercialized as dbt Cloud in 2018, and became the de facto transformation layer for modern cloud data stacks through the 2020s.

What dbt actually is

dbt Core and dbt Cloud, side by side

dbt launched as an open-source project in 2016 and was commercialized as dbt Cloud in 2018.

Source: Thinklytics data platform practice, 2026.

dbt is three things at once:

• A SQL templating engine that lets analytics engineers write modular, reusable, and testable transformation code in version control.
• A test framework that asserts properties of the data (uniqueness, not-null, referential integrity, custom assertions) at every build.
• An orchestration and documentation layer that produces a dependency graph (DAG) of all transformations and auto-generates documentation from the model definitions.

The shared idea is that the transformation layer should look like software engineering: tested, reviewed, versioned, deployed through CI/CD. Before dbt, the transformation layer was usually a wall of stored procedures in the warehouse with no test coverage and no documentation.

What people confuse it with

dbt is not the warehouse

• The warehouse. Storage and compute. Snowflake, BigQuery, Redshift, Databricks or Fabric holds the data and runs the queries.
• dbt. The T in ELT. Assumes data is already loaded, then organises, tests and orchestrates the transformations on top of it.

Mechanically dbt is SQL with Jinja templating. Operationally the test framework, documentation generation and DAG-based orchestration are the value.

Source: Thinklytics data platform practice, 2026.

• "dbt is a data warehouse." No. dbt runs SQL against a data warehouse (Snowflake, BigQuery, Redshift, Databricks, Fabric, etc.) and does no storage or compute itself.
• "dbt is an ETL tool." Closer to true that it is an ELT tool. dbt assumes the data has already been loaded into the warehouse and focuses on the T (transformation) step.
• "dbt is just SQL with Jinja." Mechanically true. Operationally, the test framework, documentation generation, and DAG-based orchestration are the value.

When dbt matters

Does dbt fit this team?

Four signals that dbt earns its place, and three that say the framework overhead is not worth it yet.

• The cloud warehouse is the analytical platform of record. Snowflake, BigQuery, Redshift, Databricks SQL or Fabric.
• More than one person writes transformation SQL. Several analytics engineers or analysts sharing a codebase.
• Certified metrics need consistent definitions in BI tools. The same metric should produce the same number in every downstream consumer.
• The transformation layer has outgrown ad-hoc views. Stored procedures and one-off views are no longer maintainable.
• The work belongs in pipelines, not SQL. Operational systems, log streams and ML feature stores are the wrong fit for a warehouse-side framework.
• A single analyst with low SQL volume. Not enough transformation work to justify the framework overhead.
• Fully Microsoft-stack-aligned organisation. Fabric's native transformation tools already cover the workload.

Before dbt, the transformation layer was usually a wall of stored procedures in the warehouse with no test coverage and no documentation. That is the baseline dbt is being compared against.

Source: Thinklytics data platform practice, 2026.

dbt matters when:

• The cloud data warehouse is the analytical platform of record (Snowflake, BigQuery, Redshift, Databricks SQL, or Fabric).
• The team has more than one analytics engineer or analyst writing transformation SQL.
• Certified metrics need to land in BI tools with consistent definitions.
• The transformation layer has grown past the point where ad-hoc views or stored procedures are maintainable.

When dbt does not help

dbt does not help when:

• The warehouse is the wrong primitive entirely (operational systems, log streams, ML feature stores) and the transformation work belongs in pipelines, not SQL.
• The team is a single analyst with no SQL volume to justify the framework overhead.
• The organization is fully Microsoft-stack-aligned and Fabric's native transformation tools cover the workload.

How Thinklytics works on dbt

We ship dbt as the transformation layer in nearly every cloud warehouse engagement, with a strong preference for dbt Cloud at scale. See dbt consulting for the engagement shape and dbt Cloud vs dbt Core decision framework for the build-vs-buy decision.

Frequently asked questions

What is dbt in one sentence?

dbt (data build tool) is a SQL-based transformation framework that lets analytics engineers build, test, document, and orchestrate transformations against a cloud data warehouse (Snowflake, BigQuery, Redshift, Databricks, Fabric) using version-controlled SQL plus Jinja templating.

What is the difference between dbt Core and dbt Cloud?

dbt Core is the open-source CLI tool. dbt Cloud is the commercial SaaS hosted by dbt Labs, with a managed scheduler, web IDE, semantic layer, CI/CD integration, and observability. Most production teams run dbt Cloud. See dbt Cloud vs dbt Core decision framework.

Is dbt a data warehouse?

No. dbt is a transformation tool that runs SQL against a data warehouse. The warehouse (Snowflake, BigQuery, Redshift, Databricks, Fabric) provides the storage and compute. dbt provides the framework for organizing, testing, and orchestrating the transformations.

What is analytics engineering?

Analytics engineering is the role and discipline that emerged with dbt. It sits between data engineering (pipelines, ingestion) and analytics (BI, semantic modeling, dashboards). The analytics engineer owns the transformation layer, modeling logic, and certified metrics that downstream analysts and BI consumers rely on.

How does dbt compare to Dataform?

Dataform is Google's alternative, acquired in 2020 and tightly integrated with BigQuery. dbt is cloud-agnostic and supports more warehouses. For BigQuery-only teams, Dataform is a credible alternative with no SaaS cost. For multi-warehouse or non-Google shops, dbt is the default.

What is the dbt Semantic Layer?

The dbt Semantic Layer (part of dbt Cloud) is a query interface that exposes certified metric definitions to BI tools (Tableau, Power BI, Looker, Hex, etc.) so the same metric definition produces the same number across every downstream consumer. It is dbt Labs' answer to LookML and the broader headless BI category.

How is dbt priced?

dbt Core is free (open-source). dbt Cloud has per-developer-seat pricing (around $100 per developer per month for Team, higher for Enterprise) plus consumption-based pricing for the semantic layer and dbt Mesh features. Most production teams land at $20,000 to $200,000 per year for dbt Cloud.

How does Thinklytics work on dbt?

We ship dbt as the transformation layer in nearly every cloud warehouse engagement, with a strong preference for dbt Cloud at scale. See dbt consulting for the engagement shape and dbt Cloud vs dbt Core decision framework.