How to Load Data Into Redshift With dbt

How dbt loads data into Redshift

dbt (data build tool) is not an ingestion layer - it transforms data that is already in the warehouse. You write a .sql file in models/, run dbt run, and dbt executes your SELECT and materializes the result as a table (or view) in the database. Redshift and Postgres share the same SQL dialect for everything dbt touches, so dbt-postgres and dbt-redshift are interchangeable at the model level.

The model file

A dbt model is just a SELECT. Put it in models/orders.sql:

SELECT
  order_id,
  customer_id,
  total_cents / 100.0 AS total_dollars
FROM raw_orders
WHERE status = 'completed'

That is the complete model. dbt reads the file, wraps it in a CREATE TABLE AS (because materialized = 'table' is set in dbt_project.yml), and runs it against the warehouse.

Running the model

cd /workspace
dbt run

dbt prints each model name and whether it succeeded:

Running with dbt=1.x.x
Found 1 model, 0 tests
Concurrency: 1 threads

1 of 1 START table model public.orders ........................ [RUN]
1 of 1 OK created table model public.orders .................. [SELECT 3 in 0.12s]

Finished running 1 table model in 0.21s.
Completed successfully.

To inspect the result directly:

psql -h 127.0.0.1 -U postgres -d app -c "SELECT * FROM orders;"

Project wiring

Two config files connect the model to the warehouse:

• dbt_project.yml - declares the project name, model paths, and default materialization (table vs view).
• profiles.yml - holds the connection details (host, port, database, user, password). Point DBT_PROFILES_DIR at the directory containing this file:

DBT_PROFILES_DIR=/workspace dbt run

Staging, intermediate, and marts

In production dbt projects, you layer models:

• staging models (stg_orders.sql) do light cleanup on raw tables - fix types, rename columns, cast nulls.
• intermediate models join and enrich staging data.
• mart models (orders.sql) are the final, BI-ready tables.

Models reference each other with ref() so dbt builds the full DAG and runs them in dependency order:

-- models/orders.sql (mart referencing a staging model)
SELECT
  order_id,
  customer_id,
  total_cents / 100.0 AS total_dollars
FROM {{ ref('stg_orders') }}
WHERE status = 'completed'

Add dbt test to catch not_null and unique violations on every run.