Overview
ISO/IEC 5259-3:2024 defines requirements and provides guidance for establishing, implementing, maintaining and continually improving the quality of data used for analytics and machine learning (ML). It sets out a generic, organization‑agnostic framework for a data quality management process, including a reference life cycle and methods that organizations can tailor. The standard specifically does not prescribe detailed processes, methods or numeric metrics - instead it requires and guides the design of a quality management approach suitable to each organization’s context.
Key topics and requirements
- Data quality management objectives: establish a process to ensure data is fit for analytics and ML use and continually improved.
- Organizational aspects: foster a data quality culture; define roles, responsibilities and competence management.
- Management system integration: align data quality activities with existing management systems and resources.
- Documentation & auditing: require documentation, audits, assessments, confirmation reviews and measurable work products.
- Life‑cycle coverage: specify requirements across data life‑cycle stages including:
- Data motivation and conceptualization
- Data specification
- Data planning
- Data acquisition
- Data preprocessing
- Data augmentation
- Data provisioning
- Data decommissioning
- Horizontal processes: verification & validation, configuration management, change management, and risk management for data quality.
- Supply chain & tooling: guidance for managing data quality across supplier relationships and for the management of data processing tools and dependencies.
- Project‑specific management: tailoring, planning, coordination, data quality justification, decommissioning and project work products.
Practical applications
- Build or improve an organizational data quality management program for ML and analytics projects.
- Design data governance controls that ensure data fitness for model training, evaluation and deployment.
- Integrate data quality requirements into project plans, supplier contracts and tooling procurement.
- Structure audits, verification/validation activities and documentation to support reproducibility, compliance and risk mitigation.
- Tailor the reference life cycle to specific projects (e.g., model development, data pipelines, labeling efforts) without inventing fixed metrics.
Who should use this standard
- Data governance leaders, Chief Data Officers (CDOs) and CIOs
- Data engineers, ML engineers and data scientists
- Quality assurance, compliance and risk management teams
- Procurement and supply‑chain managers working with data suppliers
- Tooling and platform architects responsible for data pipelines and processing tools
Related standards
- Other parts of the ISO/IEC 5259 series and complementary ISO/IEC AI and data management standards (consult the ISO catalogue for specific titles). These provide complementary guidance on AI systems, ethics and technical controls.
Keywords: ISO/IEC 5259-3, data quality, machine learning, ML data governance, data quality management, analytics, data life cycle, AI data standards.