What does incremental materialization allow in dbt?

Incremental materialization in dbt is designed to improve efficiency by allowing transformations to be applied only to new or changed data since the last run. This approach minimizes the amount of data processed during each run, reducing compute time and resource usage. When using incremental models, dbt maintains the existing data in the table and only adds new records or updates existing ones based on defined criteria. This capability is particularly useful for large datasets where processing all data every time would be resource-intensive and time-consuming.

The emphasis on only handling new or changed data is what distinguishes incremental materialization from full-refresh strategies, where all data is reloaded. This selective approach is crucial for workflows where data is continuously being updated or appended, allowing for a more efficient and responsive data transformation process.