DBT Tutorial

What is DBT?

DBT (Data Build Tool) is an open-source analytics engineering tool that enables data analysts and engineers to transform data in the warehouse more effectively. With DBT, you can write modular SQL, version control it with Git, test it, and deploy it like code.

It allows the T in ELT (Extract, Load, Transform). You extract and load data using tools like Fivetran or Airbyte, and transform it in your warehouse using DBT.

Why Use DBT?

DBT supports major warehouses like Snowflake, BigQuery, Redshift, Databricks, Postgres.

DBT Architecture Overview

DBT operates in the data transformation layer. Here’s how a typical ELT flow works with DBT:

DBT sits between your warehouse and BI tools. It helps to:

• Clean and transform raw data

• Apply business logic

• Create unified models for analytics

DBT Core vs DBT Cloud

Use DBT Core for local/dev environments, and DBT Cloud for team collaboration and production deployment.

Installing DBT Core

You can install DBT Core using pip:

Then initialize a project:

This creates:

• /models: where your SQL models live

• /dbt_project.yml: your project settings

• /profiles.yml: connection settings (in ~/.dbt/)

• Also Read: DBT Interview Questions

DBT Project Structure

• models/: Core of your transformation logic

• snapshots/: For slowly changing dimensions

• seeds/: CSV files loaded into the warehouse

• macros/: Jinja functions and reusable logic

DBT Model Basics

What is a Model?

A model is a SQL file (e.g., stg_orders.sql) that transforms data. Example:

Run it with:

The output table will be created in your warehouse.

Ref and DAG

Use ref() to define dependencies between models:

This tells DBT to build the DAG (Directed Acyclic Graph) automatically.

Materializations

In dbt_project.yml:

DBT Testing

You can test your models using:

Schema Tests (YAML)

Custom Tests

Create a SQL file in /tests:

Run tests with:

Jinja and Macros in DBT

DBT uses Jinja templating in SQL. Common uses:

• {{ ref('model') }} – model dependency

• {{ config(...) }} – model-level configs

• {{ var('env') }} – use of variables

• {{ macros.custom_macro(...) }} – custom logic

Macro Example

Use in model:

DBT Documentation

Generate documentation:

This gives a UI with:

• Model lineage (DAG)

• Descriptions from YAML

• Test coverage

• Column-level metadata

DBT Snapshots

Used for slowly changing dimensions (SCDs).

Example:

Run:

DBT Sources

Define raw data as sources:

Use in model:

DBT Seeds

Seeds are static CSVs loaded into your warehouse.

Example:

data/countries.csv:

Load with:

Access in models:

DBT Incremental Models

Use is_incremental() in SQL:

Saves time on large datasets.

DBT Deployment & CI/CD

In DBT Cloud, set up:

• Jobs: scheduled runs

• Environments: dev, staging, prod

• CI/CD: auto-test on pull request

Use GitHub/GitLab integration for CI:

• Run dbt run on merge

• Validate with dbt test

Popular DBT Packages

• dbt_utils: helper macros (must-have)

• dbt_expectations: test library

• audit_helper: audit trail macros

• codegen: auto-generate model skeletons

Install in packages.yml:

DBT Real-Time Project Flow

1. Load data from Fivetran into Snowflake

2. Use DBT to create:

   • Staging models (stg_)

   • Fact/dim models (dim_, fact_)

3. Add tests, snapshots, docs

4. Schedule in DBT Cloud

5. Connect output to Looker or Tableau

DBT Best Practices

• Use clear naming (stg_, fct_, dim_)

• Always use ref()

• Document everything in schema.yml

• Break logic into layers: staging → intermediate → marts

• Add tests early

• Use is_incremental() for large datasets

• Avoid SELECT * in production

DBT Job Roles and Use Cases

DBT Interview Questions Preview

1. What is the role of DBT in the modern data stack?

2. Difference between ref() and source()?

3. Explain incremental model in DBT.

4. How do you implement tests in DBT?

5. What is the use of the is_incremental() macro?

For more, visit our DBT Interview Questions blog.

Final Thoughts

DBT has transformed how data teams build and manage transformations—bringing the best of software engineering into data. From modular SQL to automated testing and documentation, DBT is a must-learn tool for modern data professionals