Data-Warehouse

Cleanse and standardize the raw data stacked in Bronze. Fix types, unify column names, remove duplicates. We define this process as SQL models in dbt.

There are two ways to load data into Bronze. Overwrite everything, or bring only what changed. Which one you choose completely changes the complexity of your pipeline.

Bronze, Silver, Gold. What changes when you load data into separate layers. We build it hands-on with DuckDB and dbt.

Super-sub types clarify business classifications at the logical model level. When converting to a physical model, three options emerge — and in a DW, that choice reshapes the entire dimension design.

Same Crow’s Foot, different meaning. A single dashed line means different things in different tools. If you want models to serve as a shared language on your project, start by aligning on notation.

Even when the ERDs look similar, the design philosophies are completely different. OLTP is about transactional integrity; DW is about analytical access paths. That difference creates unfamiliar things like Unknown records and point-in-time data.

How DW modeling considerations have shifted with Synapse, BigQuery, and Redshift. Kimball, Data Vault, One Big Table — practical criteria for choosing the right approach.