A data quality dimension is an angle from which data or a data set is judged: its accuracy, its completeness, the time it takes to be made available. Data quality dimensions are a classification convention. The UK government’s data quality framework (“The Government Data Quality Framework”, published on 3 December 2020 on gov.uk) takes up the wording of the DAMA association: “measurable features or characteristics of data”.
In the public texts we have read, these dimensions do not form a single list. The UK framework sets out six, the European regulation on statistics sets out seven criteria, and the Basel Committee devotes four principles to risk data aggregation capabilities. These lists differ because they do not judge the same object. We read them here from the original sources, consulted on 12 November 2026, before proposing a way of choosing one.
Data quality dimensions according to the texts consulted
| Text | Object judged | List, in the terms of the text |
|---|---|---|
| “The Government Data Quality Framework”, UK government, 3 December 2020 | Government input data | Six “core data quality dimensions”: completeness, uniqueness, consistency, timeliness, validity, accuracy |
| Regulation (EC) No 223/2009 on European statistics, Article 12, paragraph 1 | European statistics | Seven “quality criteria”: relevance, accuracy, timeliness, punctuality, accessibility and clarity, comparability, coherence |
| “Principles for effective risk data aggregation and risk reporting”, known as BCBS 239, Basel Committee, January 2013 | A bank’s risk data aggregation capabilities | Principles 3 to 6: accuracy and integrity, completeness, timeliness, adaptability |
The terms in the first row are quoted in the framework’s own English. The UK framework attributes these six dimensions to DAMA UK. We cite them from the framework alone, without having read the DAMA UK document it refers to: the association’s resources page is reserved for its members (dama-uk.org, consulted on 12 November 2026).
Regulation (EC) No 223/2009 is read in its consolidated version of 26 December 2024. The quality assurance framework of the European statistical system (“Quality Assurance Framework of the European Statistical System”, version 2.0, 2019), which its introduction presents as non-binding, takes up these notions in its principles 11 to 15, devoted to statistical outputs.
The Basel Committee document does not use the word “dimension”. Our BCBS 239 entry sets out the headings of its principles 3 to 6 as accuracy and integrity, completeness, timeliness, adaptability.
This vocabulary can also be read in two EU regulations. For the training, validation and testing data sets of high-risk AI systems, Article 10, paragraph 3, of the AI Act, Regulation (EU) 2024/1689 read in its consolidated version of 27 July 2026, states: “relevant, sufficiently representative, and to the best extent possible, free of errors and complete in view of the intended purpose”. Article 5, paragraph 1, point (d), of the GDPR, Regulation (EU) 2016/679, requires personal data to be “accurate and, where necessary, kept up to date”.
Why the lists do not overlap
The first reason is the object judged. The UK framework says it concerns input data first rather than analytical outputs: it judges records, and uniqueness there means the absence of duplication between them. The statistical regulation judges a published output. Relevance, which it defines as “the degree to which statistics meet current and potential needs of the users”, or comparability between areas and periods, has no meaning for an isolated record, and uniqueness does not appear in it. The Basel Committee judges a capability: adaptability, that is the ability to respond to ad hoc requests, describes the aggregation arrangements more than a piece of data.
The second reason is vocabulary: the same word changes in content depending on the text.
- Accuracy is, for the UK framework, the degree to which data matches reality; for the statistical regulation, “the closeness of estimates to the unknown true values”.
- The UK framework has only one word for time lag, timeliness. The regulation has two: timeliness, the lag between the event and the availability of the information, and punctuality, the lag between publication and the target date.
- Consistency refers, in the UK framework, to values that do not contradict one another for the same entity; coherence, in the regulation, to “the adequacy of the data to be reliably combined in different ways and for various uses”.
- Completeness covers, in the UK framework, all expected records and their essential values; for the Basel Committee, the capture and aggregation of all material risk data across the banking group.
The framework says so itself: its list is not prescriptive (“not a prescriptive list”), a seventh dimension can be added to it, and other organisations define their dimensions slightly differently.
There remain two paid standards, ISO 8000 and ISO/IEC 25012: we have not consulted them, and nothing of their content is used here.
Choosing a list for your organisation
What follows is a proposed method, separate from the texts read.
The choice starts from the objects to be judged and their uses: the records in master data, an exposed data set, a published indicator, a training set. A list designed for records is not enough for an output delivered on a fixed date.
We then suggest starting from a public list and supplementing it as needed. The six dimensions of the UK framework suit records; our page on data quality adopts six dimensions with the same names, timeliness becoming freshness, and shows how a rule is attached to them. A use can add one: representativeness for a training set, punctuality for a dated delivery. An organisation subject to a text has an interest in adopting its wording.
Each dimension adopted is given a one-sentence definition, with its boundary. The UK framework gives an example: a date of birth entered in American format remains valid and becomes inaccurate once read in British format. The definition says where validity ends and accuracy begins. It goes into the business glossary, whose upkeep and arbitration body are described in our page on data governance.
Finally, each quality rule is attached to a single dimension. A dimension without a rule is withdrawn; a rule without a dimension signals a definition to be reworked.
The limits of a grid of dimensions
A dimension classifies: it carries no rule, threshold or measure. Calculation belongs to metrics, presented in our article on data quality metrics.
Dimensions can also pull in opposite directions. The UK framework speaks of “trade-offs” and cites the gross domestic product publication model introduced in 2018 by the UK statistics office: monthly estimates, at the cost of a trade-off between timeliness and accuracy. Classifying a defect therefore does not say whether it should be corrected: that judgement depends on use, which the same framework sums up as “fitness for purpose”.
Finally, the same defect can fall under two dimensions. Two records for the same supplier breach uniqueness and, if their addresses differ, consistency. The classification convention is decided once and written down with the list.
The list adopted is put to the test in a working session. The open anomalies in a domain are classified one by one with the data owner. Those that fit no dimension, or that fit several, point to the definitions to be rewritten.