video1.0<iframe src="https://www.loom.com/embed/c01032e2e58b40a19394404b8f78d5a8" frameborder="0" width="1720" height="1290" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12901720Loomhttps://www.loom.com12901720https://cdn.loom.com/sessions/thumbnails/c01032e2e58b40a19394404b8f78d5a8-fbb88442e6037d41.gif166.272A 3-minute runnable proof for the ClickHouse role. Same 50M-row events table in two schemas — naïve vs tuned. Both return the same answer, but the naïve one scans 50M rows / 1.8 GB while the tuned one reads 478K / 6 MB — ~280× less data, from ORDER BY + partitioning + a projection. Runs from one command (Docker only, synthetic data, no credentials), and a GitHub Actions job re-runs the benchmark on every push and fails if the tuning regresses. Repo: github.com/boheastill/clickhouse-dwh-tuning-demo3