Data quality

Data quality measures whether data is fit for its use. It is usually assessed along six dimensions:

• Accuracy: it reflects reality (the address is correct).
• Completeness: no required values are missing.
• Consistency: the same data matches across systems.
• Timeliness: it is up to date when needed.
• Validity: it follows formats and rules (an ID number with the right check digit).
• Uniqueness: there are no duplicate records.

Quality is not achieved with a one-off clean-up: it is defined with rules and thresholds approved by the Data Owner, measured periodically and fixed at the source. It is especially critical for AI, because a model trained on poor data learns errors; that is why Article 10 of the AI Act requires relevant, representative and, as far as possible, error-free data.

Read more: The 6 data quality dimensions with examples and our complete data quality guide.