Data quality

Data quality describes how accurate, complete, consistent, timely and fit for purpose a dataset is. It is usually assessed against explicit dimensions, accuracy, completeness, consistency, uniqueness, timeliness and validity, rather than treated as one vague notion of "good data."

In practice, data quality is enforced through tests and rules applied at specific points: constraints in a database, assertions in a transformation tool like dbt, or dedicated data quality platforms that check for nulls, duplicates, referential integrity and business-rule violations. This is a narrower, more testable concept than data observability, which watches for anomalies and drift continuously; quality checks encode specific, known expectations, while observability catches unknown or unexpected failures. Data quality issues frequently trace back to poor master data management, where the same entity is represented inconsistently across systems.

Data quality matters because every downstream decision, a dashboard, a forecast, a machine learning model, inherits the quality of its inputs; no amount of good data modeling or data governance compensates for unreliable source data. Common pitfalls include checking quality only at the end of a pipeline instead of at the point of entry, defining rules without input from the business users who understand what "correct" means, and fixing symptoms in reports rather than root causes upstream, which a data catalog can help trace.