Kahan Data Solutions on YouTube · Open on YouTube
Analytics engineering, data quality and governance · Proficient · Lesson 67 of 99 · 11 min video, then about 2 hours of practice
Incremental models, snapshots and warehouse cost
After this you can convert a slow full-refresh model to an incremental one, track history with snapshots, and estimate and reduce what a model costs to run.
Why it matters
In cloud warehouses, analyst SQL costs money every time it runs. Incremental models process only new or changed rows; snapshots record how a dimension changed over time (slowly changing dimension type 2). Learn when each is worth the extra complexity, and how to find your most expensive models. The trap is premature optimisation: a full refresh that runs in a minute is simpler and safer than an incremental model with subtle bugs.
How to use this resource
A practical tutorial on building incremental models in dbt, including the is_incremental() pattern.