Overview
EN ISO/IEC 5259-1:2025 (ISO/IEC 5259-1:2024) - “Artificial intelligence - Data quality for analytics and machine learning (ML) - Part 1: Overview, terminology, and examples” provides a foundational, conceptual framework for understanding data quality in analytics and machine learning. Published by ISO/IEC and adopted by CEN, this Part 1 defines scope, core terminology, data-life-cycle concepts, and illustrative examples and use cases to help practitioners apply consistent data-quality thinking across AI projects.
Key topics
- Terminology and definitions: Standardized terms (aligned with ISO/IEC 22989 and ISO/IEC 23053) to reduce ambiguity in data quality, data originator, data life cycle and related concepts.
- Data quality concepts for analytics and ML: Fundamental considerations for how data characteristics affect model performance, reuse, sharing and downstream analytics.
- Data quality concept framework:
- Data quality management (processes to assess and improve data quality)
- Data quality governance (oversight, roles and responsibilities)
- Data provenance (lineage and origin tracking)
- Data life cycle for analytics and ML: Stages from conception to discontinuation, including cross-stage processes relevant to data preparation, labeling, evaluation and reuse.
- Examples and scenarios: Informative annex with use cases and usage scenarios to illustrate practical application of the concepts.
- Context within the ISO/IEC 5259 series: Part 1 links to other parts (5259-2: measures and reporting; 5259-3: management requirements; 5259-4: training/evaluation data guidance; 5259-5: governance framework; TR 5259-6: visualization).
Practical applications
- Establishing a common vocabulary for multidisciplinary AI teams (data scientists, engineers, analysts, auditors).
- Guiding data-quality planning across the data life cycle-from acquisition and labeling to model evaluation and data retirement.
- Informing data governance policies, provenance tracking and compliance checks for reuse and sharing.
- Supporting development of quality assurance processes for training and evaluation datasets used in supervised, unsupervised, semi‑supervised and reinforcement learning.
- Serving as a reference for tooling, reporting and data-quality visualization efforts in enterprise ML pipelines.
Who should use this standard
- Data scientists, ML engineers and data engineers
- Data governance and compliance officers
- AI system architects and project managers
- Regulators, auditors and procurement teams evaluating data‑quality practices
Related standards
- ISO/IEC 22989, ISO/IEC 23053 (AI concepts and frameworks)
- Other parts of ISO/IEC 5259 series (5259-2, 5259-3, 5259-4, 5259-5, TR 5259-6) for measures, requirements, governance, training data guidance and visualization.
This standard is essential for organizations aiming to formalize data quality practices for robust, trustworthy analytics and machine learning outcomes.