Propel is an Analytics Platform as a Service that connects to Snowflake to serve data via low-latency APIs to power customer-facing applications. Propel's architecture uses an event-driven system, where different services emit events whenever there is a state change. The company was experiencing high Snowflake credit consumption despite processing more and more data. To optimize costs, they ran several experiments, including transforming JSON during ingestion, removing the GROUP BY PROPEL_EVENT_ID operation in dbt models, using the dbt merge incremental strategy without duplicates, deduplicating only new data, and limiting the dbt merge by time. These optimizations resulted in a 20x reduction in Snowflake costs, and the company is now planning to migrate their entire pipeline to Dynamic Tables.