Getting Started with dbt and Snowflake: A Complete ELT Data Transformation Guide

Last updated: January 2026

Introduction

dbt with Snowflake lets you write SQL SELECT statements that automatically materialize as tables, views, or incremental models in your warehouse — with built-in testing, documentation, and lineage tracking. Setup takes 15 minutes: install dbt-snowflake, configure profiles.yml with your Snowflake credentials, and run dbt init. This guide covers the complete workflow from first model to production deployment with CI/CD. that integrates seamlessly with Snowflake.

This dbt Snowflake guide walks you through setting up dbt with Snowflake using Partner Connect, creating your first transformation models, defining data quality tests, and running transformations in dbt Cloud. By the end, you will understand the ELT architecture pattern and how dbt fits into your Snowflake data pipeline.

What is ELT Architecture?

ELT stands for Extract, Load, Transform. Unlike traditional ETL (Extract, Transform, Load), ELT loads raw data into the data warehouse first and then transforms it using the warehouse's processing power. For a comparison of how to manage compute workloads for ETL vs analytics in Snowflake, see our dedicated guide. This is the modern approach adopted by most cloud-native data platforms.

• Extract: Pull data from source systems such as APIs, databases, flat files, and SaaS applications.
• Load: Ingest raw data into Snowflake staging tables without modification. Tools like Fivetran, Airbyte, or Snowflake's COPY INTO command handle this step. You can load data from Azure Cloud, Google Cloud Platform, or AWS S3.
• Transform: Use dbt to clean, model, and aggregate data inside Snowflake. This is where business logic lives, and dbt manages it with version-controlled SQL.

This approach leverages Snowflake's elastic compute to handle transformations at scale, eliminating the need for separate ETL servers. You pay only for the compute time used during transformation, and Snowflake automatically scales resources up or down based on workload demands.

Why Use dbt with Snowflake?

dbt has become the standard transformation layer in the modern data stack for several compelling reasons:

• SQL-based transformations - no Python or Spark required. Analysts who know SQL can build and maintain production data models.
• Version-controlled data models using Git - every change to your transformation logic is tracked, reviewed, and auditable.
• Automated data quality testing with schema.yml - define tests like unique, not_null, accepted_values, and relationships to catch data issues before they reach dashboards.
• Built-in documentation generation - dbt automatically generates a searchable documentation site from your model descriptions and column definitions.
• Incremental model support for large datasets - process only new or changed rows instead of rebuilding entire tables, reducing compute cost and execution time.
• Native integration via Snowflake Partner Connect - connect dbt Cloud to Snowflake in minutes with automatic warehouse, database, and role provisioning.

Step 1: Connect dbt to Snowflake via Partner Connect

Snowflake Partner Connect provides a streamlined way to set up dbt Cloud with your Snowflake account. It automatically provisions all necessary resources so you can start building models immediately.

How to Set Up Partner Connect

1. Log into your Snowflake account as ACCOUNTADMIN (required for Partner Connect).
2. Navigate to Admin → Partner Connect in the left sidebar.
3. Find and select dbt from the list of available partners.
4. Click Connect to begin the automatic provisioning process.

Snowflake automatically creates the following resources:

Step 2: Register for dbt Cloud

After the Partner Connect setup completes, Snowflake redirects you to dbt Cloud to finish account creation.

1. You are redirected to the dbt Cloud signup page with your Snowflake connection details pre-filled.
2. Provide your email address, account name, and create a password.
3. Accept the dbt Cloud Terms of Service.
4. Your dbt Cloud trial account is activated automatically.
5. The Snowflake connection is pre-configured - no manual connection setup required.

The dbt Cloud Developer plan is free for individual use and includes everything you need to get started: a web-based IDE, job scheduling, and documentation hosting.

Step 3: Navigate dbt Cloud IDE

Once your account is set up, navigate to the dbt Cloud IDE (called dbt Studio) where you will write and test your transformation models.

1. Open dbt Cloud and navigate to Studio from the top menu.
2. dbt Studio provides an online IDE for writing SQL transformation models, previewing results, and running commands.
3. Familiarize yourself with the key files and folders in your project:

Step 4: Create Your First dbt Model

A dbt model is a SQL SELECT statement saved as a .sql file in the models/ directory. When you run dbt, each model creates a corresponding table or view in Snowflake. Models use the ref() function to reference other models, which allows dbt to automatically determine the correct execution order.

Example Model: models/customer_orders.sql

SELECT
    c.customer_id,
    c.customer_name,
    COUNT(o.order_id) AS total_orders,
    SUM(o.order_amount) AS total_revenue,
    MAX(o.order_date) AS last_order_date
FROM {{ ref('raw_customers') }} c
LEFT JOIN {{ ref('raw_orders') }} o
    ON c.customer_id = o.customer_id
GROUP BY c.customer_id, c.customer_name

When you run dbt run, this SELECT statement executes against Snowflake and creates a table or view named customer_orders in the PC_DBT_DB database. The {{ ref('raw_customers') }} and {{ ref('raw_orders') }} references tell dbt to resolve the fully qualified table names and ensure those models run first.

Understanding Model Materializations

dbt supports several materialization strategies that control how your model is stored in Snowflake:

Step 5: Define Schema and Tests in schema.yml

The schema.yml file defines model metadata and data quality tests. These tests run automatically when you execute dbt test and help catch data issues before they propagate to downstream reports and dashboards.

Example schema.yml

version: 2

models:
 - name: customer_orders
    description: "Aggregated customer order summary"
    columns:
 - name: customer_id
        description: "Unique customer identifier"
        tests:
 - unique
 - not_null
 - name: total_orders
        description: "Total number of orders placed"
        tests:
 - not_null
 - name: total_revenue
        description: "Total revenue from all orders"

Available Built-in Tests

You can also write custom tests using SQL and Jinja. Custom tests are stored in the tests/ directory and return rows that fail the test condition.

Step 6: Compile and Run dbt Models

With your models and tests defined, you are ready to compile and execute them against Snowflake. dbt provides three core commands for this workflow:

dbt compile    # Validates SQL and resolves ref() references
dbt run        # Executes models and creates tables/views in Snowflake
dbt test       # Runs data quality tests defined in schema.yml

What Each Command Does

• dbt compile: Parses all SQL files, resolves ref() and source() references, and generates compiled SQL. This step validates your project without executing anything against Snowflake. Use this to catch syntax errors early.
• dbt run: Compiles each model, determines the correct execution order based on ref() dependencies, and runs each SELECT statement against Snowflake. Tables and views are created or replaced in PC_DBT_DB.
• dbt test: Executes all tests defined in schema.yml files. Each test runs a query against Snowflake and reports pass or fail results. Failed tests return the rows that violated the test condition.

After running dbt run, log into Snowflake and verify that your tables were created in the PC_DBT_DB database under the appropriate schema.

Best Practices for dbt with Snowflake

Follow these best practices to build maintainable, scalable, and reliable dbt projects on Snowflake:

• Use ref() to reference other models instead of hardcoding table names. This ensures dbt builds models in the correct dependency order and makes your project portable across environments.
• Organize models into staging, intermediate, and mart layers:
  • Staging models - clean and rename raw source data (one model per source table).
  • Intermediate models - apply business logic, joins, and calculations.
  • Mart models - final tables optimized for reporting and analytics consumption.
• Use incremental models for large tables to avoid full refreshes. Configure the is_incremental() macro to filter for only new or changed rows.
• Add tests to every model to catch data quality issues early. At minimum, test primary keys for uniqueness and not_null.
• Use dbt docs generate to create automatic documentation. This produces a searchable website with model lineage graphs, column descriptions, and test definitions.
• Use Snowflake's warehouse sizing strategically. Start with X-Small for development and scale up for production runs with large datasets.
• Leverage dbt's source freshness checks to monitor when raw data was last loaded, ensuring your transformations run on up-to-date data.

Conclusion

dbt transforms Snowflake from a data warehouse into a complete analytics engineering platform. By combining Snowflake's elastic compute with dbt's SQL-based transformation framework, you get a powerful, testable, and version-controlled data pipeline that scales with your organization's needs.

In this guide, you learned about the ELT architecture pattern and how it differs from traditional ETL, how to connect dbt to Snowflake using Partner Connect, how to create SQL models and define data quality tests in schema.yml, and best practices for organizing and running your dbt project. With these foundations in place, you are ready to build production-grade data transformations on Snowflake using dbt.

Frequently Asked Questions