How Amplify optimized their incremental models with dbt on Snowflake

Like many other dbt users, Amplify has some very large data sets. (Their largest model needs to be updated every two hours and would cost $2.6 million to build annually if they fully refreshed it every time). Turning this into an incremental model was a natural choice, and helped a lot. However, they found that simply adding `materialized = ‘incremental` didn’t solve all of their problems.