Build Production-Ready dbt Transformation Pipelines

Defines dbt's medallion model-layer architecture - sources for raw data definitions, staging for 1:1 light-cleaned source models, intermediate for business-logic joins and aggregations, marts for final analytics tables - with its naming convention: stg_ prefix for staging like stg_stripe__payments, int_ for intermediate like int_payments_pivoted, and dim_ or fct_ for marts like dim_customers and fct_orders. Ships seven patterns: source definitions (a _stripe__sources.yml declaring raw source tables), staging models, intermediate models, mart models building dimension and fact tables, testing and documentation (a _core__models.yml applying unique and not_null tests, accepted_values value lists, dbt_utils.expression_is_true for numeric bounds, dbt_utils.recency for freshness, and relationships tests validating foreign keys against ref()), macros for DRY code (a cents_to_dollars conversion macro, a generate_schema_name macro customizing target schema naming, and a limit_data_in_dev macro that restricts a model to recent rows only when running against the dev target), and incremental strategies: delete+insert as the default for most warehouses, merge for late-arriving data with explicit merge_update_columns, and insert_overwrite for partition-based loads, all gated by an is_incremental() check in the model's WHERE clause. Documents the dbt_project.yml and folder structure a project starts from, and the core CLI commands grouped by development, testing, documentation, and debugging workflows.

Use it when structuring dbt models across layers, writing source definitions or staging, intermediate, or mart models, adding tests and documentation, writing reusable macros, or choosing an incremental materialization strategy. Skip it for anything unrelated to dbt transformation patterns or outside this scope.

Input is a transformation need - a new source, a staging or mart model, a test, a macro, or an incremental-load strategy; output is the matching dbt SQL model, YAML test or documentation block, or Jinja macro, following the medallion layer and naming conventions above.

Built on dbt (dbt_project.yml, the ref() and source() functions, Jinja macros, target.name and target.schema for environment-aware behavior) with dbt_utils for extended test macros such as expression_is_true and recency. Targets any dbt-supported warehouse, with incremental strategies noted as varying by warehouse - delete+insert as the general default, merge and insert_overwrite as warehouse-dependent options.

Analytics engineers building or restructuring a dbt project who need concrete layer, testing, macro, and incremental-model patterns rather than an abstract data-modeling overview. The layer-naming convention (stg_ for staging, int_ for intermediate, dim_/fct_ for marts) is part of the same review; it also covers data-quality testing and incremental models for large datasets.