The case for streaming-first data platforms

For two decades batch ETL was the default because storage was cheap and compute was expensive. That trade-off has flipped.

Modern warehouses (Snowflake, BigQuery, Databricks) all support streaming ingest at reasonable cost. The question is no longer can we, but should we.

Streaming-first unlocks real-time activation: the same pipeline powers dashboards, alerts and product features.

The key shift is in modeling: incremental dbt models, idempotent loads and clear contracts between producers and consumers.