Overview
ISO/IEC 5259-2:2024 - "Artificial intelligence - Data quality for analytics and machine learning (ML) - Part 2: Data quality measures" defines a standardized data quality model, a catalog of data quality measures, and guidance for reporting data quality specifically in analytics and ML contexts. Applicable to all organization types, it supports consistent assessment and communication of data fitness for ML pipelines, analytics tasks, and data‑driven decision making.
Key topics
- Data quality model: structured components and how data quality fits into the analytics and ML data life cycle.
- Data quality characteristics: definitions and measures for inherent, system‑dependent, and additional characteristics such as:
- Inherent: Accuracy, Completeness, Consistency, Credibility, Currentness
- Inherent & system‑dependent: Accessibility, Compliance, Efficiency, Precision, Traceability, Understandability
- System‑dependent: Availability, Portability, Recoverability
- Additional: Auditability, Balance, Diversity, Effectiveness, Identifiability, Relevance, Representativeness, Similarity, Timeliness
- Measurement approach: definition of quality measure elements (property + measurement method) and guidance for designing measurement functions (informative annex).
- Implementation and reporting: practical guidance to implement a data quality model for a specific analytics or ML task and a reporting framework to communicate measure results.
- Supporting artifacts: UML model, characteristic overviews, and comparison with prior standards (e.g., ISO/IEC 25012).
Practical applications
- Establishing repeatable data quality assessments for training/validation data, feature stores, and production inference data.
- Defining measurable SLAs and quality gates for data ingestion, labeling, and pre‑processing in ML pipelines.
- Supporting model governance, auditing, and regulatory compliance by documenting measures like traceability, auditability, and representativeness.
- Informing dataset selection, augmentation, and bias mitigation via measures such as balance, diversity, and representativeness.
- Enabling cross‑team or cross‑organization data quality reporting to stakeholders, auditors, and regulators.
Who should use this standard
- Data scientists, ML engineers, and data engineers creating, validating, or deploying ML models
- Data stewards, data governance teams, and compliance officers
- AI assurance/audit teams and organizations seeking standardized data quality benchmarks
- Platform and MLOps teams implementing monitoring, data quality gates, and reporting dashboards
Related standards
Using ISO/IEC 5259-2:2024 helps organizations operationalize data quality for analytics and ML, aligning measurement, reporting, and governance across the data life cycle.