Configuring an incremental model in dbt involves several key steps. First, the model must be explicitly defined as incremental by setting materialized='incremental' in its configuration block (C). The core logic for an incremental run is to process only new or changed data. This is achieved by filtering the source data, typically by adding a WHERE clause to the model's SQL that leverages the isincremental() macro (B). To handle updates to existing records and prevent duplicates, a uniquekey can be specified. This key enables dbt to match records between the source and destination, performing an update/merge operation instead of a simple append (D).

A. partitionby is a data warehouse-specific configuration for performance optimization that physically organizes data. It is not a primary method for defining the incremental logic of which records to process.

1. dbt Labs Documentation

"Incremental models":

Reference for C: Under the "How do I use incremental models?" section

the first step listed is: "To use incremental models

you need to set the materialized configuration for a model to incremental."

Reference for B: The same section explains the second step: "You also need to tell dbt how to filter the rows for an incremental run... using the isincremental() macro." This macro is used within a WHERE clause.

Reference for D: The section "How do I use uniquekey?" states: "The optional uniquekey parameter enables updating existing rows instead of just appending new rows. If a uniquekey is specified

dbt will use a merge statement to build the model."

2. dbt Labs Documentation

"About materializations":

Reference for C: This page defines the incremental materialization: "Models can be configured as incremental... dbt will insert new records and update existing records in the table." This confirms materialized='incremental' is the foundational configuration.

3. dbt Labs Documentation

"BigQuery configs" (as an example of adapter-specific configs):

Reference for A: The partitionby configuration is listed as a specific configuration for warehouses like BigQuery to improve query performance by partitioning tables. It is described as a physical layout property

distinct from the logical filtering (WHERE clause) or update strategy (uniquekey) of the incremental model itself.