Skip to main content

dbt Roadmap 2026

Transform data in the warehouse with dbt: models, tests, documentation, macros, incremental builds, deployment and the semantic layer.

4 stages · 8 units · Intermediate · about 45 hours · Last reviewed 19 November 2026

What is a dbt?

dbt, short for data build tool, is an open source framework for transforming data inside a warehouse using SQL. You write select statements as models, and dbt compiles them, runs them in dependency order, tests the results and generates documentation. Analytics engineers use it to keep transformations version controlled and reliable.

Step-by-step dbt roadmap

  1. 1

    Foundations

    A dbt project.

    • Basics

      dbt basics cover the structure of a dbt project and how a model becomes a warehouse object. A model is a SQL file with a select statement, configured in dbt_project.yml, and a materialisation such as view, table, incremental or ephemeral controls whether dbt builds it as a view, a table or something else.

    • Style

      dbt style is the convention for organising models into layers so a project stays readable as it grows. Staging models clean and rename one source each, intermediate models combine logic, and mart models expose business entities. Consistent naming, folder structure and CTE patterns make reviews and onboarding easier.

  2. 2

    Quality

    Trust the data.

    • Testing

      dbt testing is how you check that models behave as expected. Generic tests such as unique, not_null, accepted_values and relationships are declared in YAML, singular tests are custom SQL queries that should return no rows, and descriptions in YAML feed the documentation site built with dbt docs.

    • Reuse

      Reuse in dbt means removing repeated SQL with Jinja templating and macros. Jinja adds loops, conditionals and variables inside SQL files, and a macro is a reusable function defined in the macros folder. Packages such as dbt_utils provide ready-made macros, so teams share logic instead of copying it.

  3. 3

    Scale

    Efficient builds.

    • Performance

      dbt performance is about making builds faster and cheaper. Incremental models process only new or changed rows using the is_incremental macro and a unique_key, while state-based selection with the state method and defer runs only modified models and their dependents, which keeps CI and production builds short.

    • Architecture

      dbt architecture for large projects covers structuring many models and many teams. It includes folder and naming standards, model contracts and versions, access modifiers that mark models as public or private, groups, and dbt Mesh patterns that split a project into several connected ones with cross-project refs.

  4. 4

    Deploy and Serve

    Production.

    • Deployment

      dbt deployment is running dbt reliably outside your laptop. It covers scheduled production jobs in dbt Cloud, Airflow or another scheduler, separate dev and prod targets, and continuous integration that builds and tests only changed models on each pull request, so that bad code is caught before merging.

    • Semantic

      The dbt semantic layer defines business metrics once, in YAML, using semantic models, entities, dimensions and measures through MetricFlow. Tools and AI assistants then query governed definitions like revenue instead of rewriting SQL, which keeps numbers consistent across dashboards, notebooks and LLM-based question answering.

Frequently asked questions

Ready to practise for interviews?

Practise dbt questions with an AI mock interviewer and get feedback on your answers.

More roadmaps