dbt review

dbt is a data transformation platform that organizes SQL queries into models, then manages how those models are built and connected. It sits in the transformation layer of a data stack: data lands in a warehouse, and dbt helps teams define and run the SQL transformations that produce analytics-ready models.

The workflow centers on modular, tested and version-controlled SQL. dbt adds dependency management, interactive lineage across projects, column-level tracing and tools for refactoring models and columns. Its SQL awareness and local validation are intended to catch issues before warehouse execution, while multi-dialect compilation is designed to keep transformation logic portable as a data platform changes.

The platform includes a browser-based IDE and supports local development, with an IDE extension for named editors. The homepage says dbt never stores your data. It does not describe dbt as a general-purpose ingestion service or database; its role is to manage and execute the transformation logic in a data workflow.

The primary language named for building transformations is SQL. Teams organize work as models and use dbt’s SQL comprehension, local validation, lineage and refactoring features to maintain those projects. The homepage also points developers to a quickstart guide and names a browser-based IDE for working in the cloud.

For local work, dbt offers an IDE extension for Cursor, Claude Code, Windsurf and VS Code. Those integrations give developers an editor-based workflow, but they are not a substitute for knowing how automation is exposed. The homepage does not name a REST API, SDK, CLI or supported SDK languages. Confirm API, SDK and CLI support in the vendor’s documentation before designing scripts or deployment tooling around them.

The site also names integrations including Snowflake, Databricks, Tableau and Fivetran. Treat those as ecosystem integrations, not proof of a particular API contract or SDK. Check the quickstart and platform documentation for warehouse compatibility, project setup and the specific workflow your team intends to automate.

The free Developer plan includes one developer seat, one project, a browser-based IDE and 3,000 successful models per month. That is a useful starting point, but teams should check how their model runs fit the monthly quota before relying on it for production. The plan does not include a published allowance for queried metrics.

Starter costs $100 per user per month and includes five developer seats, one project, 15,000 successful models and 5,000 queried metrics per month. The pricing page lists no annual price or discount. Enterprise and Enterprise+ use custom pricing; each includes 100,000 successful models and 20,000 queried metrics, with 30 projects on Enterprise and unlimited projects on Enterprise+.

Add-ons can change the bill. dbt State costs $0.094 per Daily Active Target Table, with custom upfront pricing available. dbt Wizard uses token-based pricing that varies by AI model; Enterprise and Enterprise+ include $100 and $200 in monthly credits respectively, and extra usage is a consumption add-on. Ask how quota overages are handled before scaling.

The vendor’s homepage describes dbt as interoperable and says it uses open standards, but it does not identify a software licence or explain whether the current platform can be self-hosted. Ask the vendor which components are available as source, what licence applies and whether your deployment can run in your own environment before treating self-hosting as an option.

The homepage says dbt never stores your data. That is useful context, but it does not explain project export formats or how to retrieve platform metadata, lineage and configuration. Confirm those details, along with what remains accessible if a subscription ends.

SQL models and multi-dialect compilation may help keep transformation logic portable, but dbt-specific project structure, tests, metadata and workflows can still require migration work. Test an export and rebuild path with representative projects before making dbt a core dependency.

Plan selection sets important monthly allowances: Developer includes 3,000 successful models; Starter includes 15,000 and 5,000 queried metrics; Enterprise tiers include 100,000 and 20,000 respectively. Check how those quotas are counted against your run patterns, and ask what happens when a project exceeds them. The pricing page does not list overage rates.

The homepage promotes local validation, lineage and state-aware builds through dbt State. It says dbt State builds only models that need updating, but gives no reliability or savings guarantee for an individual deployment. Confirm security controls, compliance coverage, support response terms and production service commitments with the vendor before depending on the platform for critical pipelines.

Reddit discussions show both appreciation for lineage and documentation and uncertainty about dbt’s role. One commenter calls those features useful; another says the tooling can feel over-engineered. A thread also describes SQL files in Git with an external orchestrator as a workable approach. Evaluate the workflow against your existing orchestration and warehouse setup.