Skip to main content
Bureau of AI
Category: Data Governance & Quality

ISO/IEC 5259 (Data Quality for Analytics and ML)

Also known as: ISO/IEC 5259 series, Data Quality for Analytics and Machine Learning
Simply put

ISO/IEC 5259 is a multi-part international standard series that provides tools and methods for assessing and improving the quality of data used in analytics and machine learning. It helps organizations describe what good data quality looks like, measure it, report on it, and govern the processes that maintain it. The series is issued by ISO and IEC as a voluntary standard rather than a law, so its use is not mandated unless an organization or regulator chooses to require it.

Formal definition

The ISO/IEC 5259 series is a set of jointly issued ISO and IEC documents addressing data quality in the context of analytics and machine learning. Based on the evidence, the foundational part (ISO/IEC 5259-1:2024) establishes the overall aims and framing of the series; ISO/IEC 5259-2:2024 specifies a data quality model, a set of measurable data quality characteristics or measures, and guidance on reporting data quality; and ISO/IEC 5259-5:2025 provides a governance framework to help organizations oversee and direct data quality for analytics and ML. Collectively the series is intended to provide methods to assess and improve the quality of data used in these contexts. As a voluntary consensus standard, it supports but does not by itself constitute regulatory compliance, and it should be distinguished from binding law and from other instruments. Note that the evidence provided does not enumerate every part of the series, the full content of parts 3 and 4, or the specific data quality characteristics defined; those details are out of scope of this entry and should not be inferred.

Why it matters

The quality of data used to train and operate machine learning systems shapes model behavior, and poor data quality is a recurring source of downstream model risk, from unreliable predictions to unfair or unstable outputs. ISO/IEC 5259 matters because it offers a shared, structured vocabulary and set of methods for describing what good data quality looks like, measuring it, and reporting on it. In an environment where organizations struggle to demonstrate the fitness of their data for a given analytics or ML use, a common reference point helps teams communicate consistently across data engineering, model development, validation, and oversight functions.

As a voluntary international standard issued by ISO and IEC, ISO/IEC 5259 does not by itself create legal obligations, and adopting it does not constitute regulatory compliance. Its significance is instrumental: an organization or a regulator may choose to reference or require it, and it can support—but not replace—the controls demanded by binding law or supervisory guidance. Professionals should treat it as a tool that can strengthen governance and risk practices rather than as a mandate that eliminates data-related risk.

Because the series separates the data quality model and measures (addressed in ISO/IEC 5259-2) from a governance framework for overseeing data quality (addressed in ISO/IEC 5259-5), it can help distinguish the technical work of measuring data quality from the organizational work of directing and accountable oversight. This distinction is useful for teams that must both improve data quality and demonstrate that responsible governance processes are in place, though the standard reduces rather than removes the risks associated with data used for analytics and ML.

Who it's relevant to

Data scientists and ML engineers
Those building and maintaining models can use the data quality model and measurable characteristics described in ISO/IEC 5259-2 to define, quantify, and report on the fitness of training and operational data. The standard provides a common vocabulary for describing data quality, though it does not by itself specify which characteristics apply to a particular use case in this entry's evidence.
Model risk and validation functions
Teams responsible for identifying, measuring, and controlling model risk may reference the series to assess whether data quality is adequately measured and reported. Because data quality is a distinct input to model performance and risk, the standard can support validation activities without substituting for an organization's broader model risk management practices.
AI governance and oversight roles
Those responsible for organizational accountability and oversight of AI systems may draw on the governance framework in ISO/IEC 5259-5 to help direct and oversee data quality processes. This supports governance structures but should be distinguished from the technical measurement of data quality addressed in other parts of the series.
Auditors and compliance professionals
Because ISO/IEC 5259 is a voluntary consensus standard rather than binding law, professionals evaluating an organization against it should treat conformance as evidence of practice, not as regulatory compliance. It can be used as a benchmark where an organization or a regulator has chosen to require or reference it.

Inside ISO/IEC 5259 (Data Quality for Analytics and ML)

Data Quality Framework for ML/Analytics
ISO/IEC 5259 is a multi-part standard series developed under ISO/IEC JTC 1/SC 42 (the subcommittee for artificial intelligence) that addresses data quality for analytics and machine learning. As a voluntary international standard rather than binding law, it provides a common structure for defining, measuring, and managing the quality of data used across the ML lifecycle.
Data Quality Measures and Characteristics
The series is commonly understood to define data quality dimensions or characteristics (such as completeness, accuracy, consistency, and other attributes) that can be applied to datasets used for training, testing, and operating ML systems. The intent is to give organizations a shared vocabulary for assessing whether data is fit for its intended analytical or ML purpose.
Data Quality Process and Governance Guidance
The standard series addresses processes for managing data quality across data lifecycle stages, which can support data governance activities. Note that data governance and broader AI governance are distinct from the data-focused scope of this standard, though the standard can inform them.
Multi-Part Structure
ISO/IEC 5259 is organized as a series of related parts covering different aspects of data quality for analytics and ML rather than as a single document. The precise number, titles, and publication status of individual parts should be verified against the current ISO/IEC catalogue, as they may have evolved.

Common questions

Answers to the questions practitioners most commonly ask about ISO/IEC 5259 (Data Quality for Analytics and ML).

Is ISO/IEC 5259 a legally binding regulation that organizations must comply with?
No. ISO/IEC 5259 is a voluntary international standard developed by ISO and IEC, not binding law issued by a governmental authority. Adopting it is generally a matter of organizational choice or contractual requirement rather than statutory obligation. It may be referenced in procurement terms, internal policy, or as supporting evidence of good practice, but conformity is not a legal mandate in itself. Where a jurisdiction's law or regulator references or incorporates a standard, obligations flow from that legal instrument, not from the standard's existence alone. Treat any claim that the standard is universally required with caution.
Does following ISO/IEC 5259 guarantee that a dataset is free of bias or that model outputs will be fair?
No. Data quality frameworks address dimensions of data suitability and fitness for analytics and machine learning purposes; they do not, on their own, ensure that data or resulting models are unbiased or that outcomes are fair. Bias and fairness are distinct concepts that involve considerations beyond data quality, including problem framing, chosen fairness definitions, deployment context, and downstream impact. Improving data quality can reduce certain data-related risks, but it does not eliminate bias or fairness concerns, and it should not be presented as doing so.
How does ISO/IEC 5259 relate to an organization's broader AI governance and model risk management activities?
As commonly positioned, ISO/IEC 5259 focuses on data quality for analytics and machine learning, which is one input into both AI governance and model risk management rather than a substitute for either. Data quality work can support model risk management, where data-related risk is a recognized source of model risk, and can inform governance policies that assign accountability for data. Organizations typically use it alongside, not in place of, their governance structures and model risk controls. The distinction between governance (organizational oversight and accountability) and model risk management (identifying, measuring, monitoring, and controlling model risk) should be preserved when integrating any data quality standard.
What role can a data quality framework like ISO/IEC 5259 play in model validation and verification?
Data quality considerations can support both validation and verification activities, which remain distinct. Verification typically concerns whether a model or process was built correctly against specifications, while validation typically concerns whether it is appropriate for its intended use. Assessing data quality dimensions can provide evidence relevant to input data assumptions examined during these activities. However, a data quality standard does not by itself constitute a validation or verification methodology, and it does not replace the independent review functions many frameworks expect. Its outputs are typically one source of evidence among several.
How might responsibility for data quality be assigned across lines of defense?
In organizations that use a three-lines model, data quality responsibilities can be distributed rather than owned by a single function. In many frameworks, the first line owns and operates data processes, the second line provides oversight and challenge, and the third line provides independent assurance. A data quality standard can help define common terminology and dimensions used across these roles, but it does not prescribe a specific organizational structure. How responsibilities are allocated depends on the organization's own governance design and should be documented in policy rather than assumed from the standard.
What practical limitations should teams keep in mind when applying ISO/IEC 5259?
Teams should recognize that a data quality standard addresses data suitability and does not cover the full range of AI-related risks, such as model performance degradation over time, security, or deployment context. Its terminology and dimensions may need to be interpreted for specific sectors, since data quality expectations can vary between regulated environments such as banking and general enterprise use. It also does not eliminate risk; it provides structure for identifying and managing certain data-related concerns. Organizations should confirm the standard's current scope and edition directly, and avoid assuming it satisfies obligations arising from separate laws, guidance, or contractual terms.

Common misconceptions

ISO/IEC 5259 is a legally binding regulation that organizations must comply with.
It is a voluntary international standard issued through ISO/IEC, not law. Adoption is typically elective unless a contract, sector regulator, or internal policy specifically requires conformance. It does not carry the force of instruments such as the EU AI Act, nor is it interchangeable with regulatory guidance.
Following ISO/IEC 5259 constitutes a complete AI governance or model risk management program.
The standard is scoped to data quality for analytics and ML. It does not by itself establish organizational accountability structures (AI governance) or the identification, measurement, monitoring, and control of model risk (model risk management). Data quality is one input among many; strong data quality reduces but does not eliminate model risk.
Meeting the standard's data quality characteristics guarantees a fair, accurate, or unbiased model.
Data quality addresses attributes of the data itself and does not guarantee model outcomes. Concerns such as bias, fairness, model performance, and model risk are distinct dimensions that require their own assessment. Good data quality supports but does not ensure these properties.

Best practices

Verify the current publication status, structure, and part numbering of ISO/IEC 5259 against the official ISO/IEC catalogue before citing specific parts, since the series is evolving and details may change.
Use the standard's data quality vocabulary to establish a shared language across data, modeling, and compliance teams, while keeping data quality distinct from separate governance and model risk concerns.
Treat data quality assessment as one input into model risk management rather than a substitute for validation, monitoring, or independent review.
Document how selected data quality characteristics map to your organization's intended use case, since fitness for purpose depends on context and no single set of measures is universally sufficient.
Where conformance to the standard is claimed, define the scope precisely and avoid implying that it covers fairness, bias, or overall model performance, which require separate evaluation.
Integrate data quality processes into existing data governance and lifecycle controls rather than treating the standard as a standalone or one-time exercise.