A pipeline with a quality gate.
The job reads synthetic order rows, blocks invalid or duplicate records, then writes a typed, compressed Parquet dataset for analytics.
No silent row drops: if the input violates the contract, the workflow fails before deployment.
Build result
This report is generated by the same build that publishes the site.
—curated orders
—synthetic sales value
—invalid rows / duplicate IDs
Pipeline contract
CSV → validate required fields and unique order_id
→ normalize typed columns and derive revenue
→ write Zstandard-compressed Parquet
→ run OLAP model and deploy static report