01. What is Incremental Materialization
An incremental model is a dbt model that does not rebuild the full table every time.
Instead, it processes only the new or changed rows and adds them to the existing target table.
In our earlier RAW → SILVER → GOLD demo, the silver_orders model is a good example:
models/silver/silver_orders.sql
That model joins:
stg_ordersstg_customers
For a small dataset, silver_orders can be a normal table. For a large dataset, it becomes a strong candidate for incremental materialization.
Why incremental materialization helps
Incremental materialization is ideal when a model:
- processes a large amount of historical data once
- then only loads new or changed rows later
- avoids rebuilding the full target table on every run
This saves warehouse compute and improves run performance.
Simple beginner idea
Think of incremental materialization like this:
| Run | What dbt processes |
|---|---|
| First run | All completed orders |
| Next run | Only newly arrived completed orders |
| Future runs | Only rows allowed by the incremental filter |
On the first run, dbt creates the full silver_orders table. On later runs, it uses your SQL filter to decide which rows should be processed.