Data Quality Dimensions — Accuracy, Completeness, Freshness

Data quality is not a single property but a family of dimensions, each capturing a different way data can fail its consumers. A pipeline producing complete, accurate data two days late is unusable for a real-time dashboard. A pipeline producing fresh, timely data with 15% of rows missing key fields fails a different dimension. Understanding the six core dimensions — accuracy, completeness, freshness, consistency, uniqueness, and validity — enables data engineers to design targeted checks that catch the failure modes that matter for each specific dataset and consumer.

Data quality monitoring is a first-class pipeline responsibility, not an afterthought. The canonical data engineering anti-pattern is building a pipeline that produces data without any checks, discovering quality issues only when downstream dashboards show wrong numbers or a machine learning model produces nonsensical predictions, and then spending days forensically identifying when the problem started. Building quality checks directly into the pipeline — failing loudly and early when data does not meet expectations — is the engineering discipline that separates production-grade pipelines from prototypes.