Data quality and lineage

Trusted analytics requires measurable quality and traceable lineage. Quality dimensions: freshness (last successful load), completeness (null/unexpected rates), uniqueness (duplicate keys), validity (schema and business rules), consistency across sources.

Lineage answers: which upstream table/column produced this dashboard metric, and which downstream reports break if it changes. Tools: OpenLineage, dbt exposures, data catalogs, column-level tags in warehouse metadata.

					source.orders → staging.orders_clean → mart.revenue_daily → BI dashboard
				

On interviews: describe an incident where bad data reached executives and how freshness checks or lineage would have shortened detection; balance governance with team velocity.

Common pitfalls: dashboards without owner; no alerts on stale partitions; lineage only documented in wiki; quality checks that warn but never block; ignoring schema drift from CDC sources.

The trade-off is confidence in decisions versus tooling investment, process friction, and the social cost of naming data owners.

Checklist:

• List quality dimensions relevant to KPIs.
• Map lineage from source to dashboard.
• Automate freshness and volume anomalies.
• Assign owners for datasets and metrics.
• Tie breaking changes to consumer notification.