Skip to main content

Data quality

How good data is for the job it needs to do, whether it is accurate, complete, up to date, and not misleading.

The Simple Version

How good data is for the job it needs to do, whether it is accurate, complete, up to date, and not misleading.

Detailed Explanation

Data quality is formally defined and structured in the ISO/IEC 5259 series, which provides terminology, measurement frameworks, management requirements, process models, and governance approaches for data quality in AI contexts. High data quality is prerequisite for trustworthy AI: the EU AI Act requires providers to implement data governance and management practices that address quality across training, validation, and test datasets, including examination for biases. Quality is not an absolute, it is always relative to fitness for a specific intended purpose.

Key Characteristics

  • Multi-dimensional: accuracy, completeness, consistency, timeliness, representativeness, and more
  • Relative to intended use, not an absolute property of a dataset
  • Measurable through quantitative data quality measures and dimensions
  • Foundational requirement for EU AI Act high-risk system compliance

Why It Matters

Poor data quality in AI training datasets propagates errors, biases, and compliance gaps into production models, making data quality investment one of the highest-leverage activities in responsible AI programmes.

Real-World Analogy

Like the quality of ingredients in cooking, a recipe with poor ingredients will produce a poor dish however skilled the chef; similarly, poor training data produces unreliable AI outputs regardless of model sophistication.

Common Misconceptions

  • Data quality is purely a technical concern for data engineers, quality requirements must be defined by business stakeholders who understand the intended use.
  • More data automatically means better quality, volume is one dimension; accuracy, representativeness, and consistency are equally important.

Related Terms

Related Articles

Sources & Further Reading