# Data quality dimensions explained: completeness, accuracy, timeliness and more

> Data quality dimensions are measurable properties such as completeness, accuracy and timeliness that show whether data is fit for a specific purpose.

Published: 2026-11-03 · Dataplaz editorial team · https://dataplaz.lu/en/data-quality-dimensions-explained/

## Key takeaways
- The most widely used set of data quality dimensions comes from DAMA UK: completeness, uniqueness, consistency, timeliness, validity and accuracy.
- The UK Government Data Quality Framework, published on 3 December 2020, adopts these six dimensions and stresses that the list is not prescriptive.
- ISO/IEC 25012 defines a broader model of 15 data quality characteristics, split into inherent and system-dependent quality.
- The ECB's May 2024 guide on risk data aggregation expects banks to measure at least accuracy and integrity, completeness and timeliness, with tolerance levels.
- The GDPR requires personal data to be accurate and kept up to date (Article 5(1)(d)), and the AI Act requires training data for high-risk AI to be, as far as possible, free of errors and complete (Article 10(3)).

Data quality dimensions are measurable properties, such as completeness, accuracy and timeliness, used to decide whether data is fit for a specific purpose. The most widely used set, defined by DAMA UK and adopted by the [UK Government Data Quality Framework](https://www.gov.uk/government/publications/the-government-data-quality-framework/the-government-data-quality-framework) on 3 December 2020, has six dimensions. For Luxembourg organisations they are not just theory: the ECB's May 2024 guide for banks, the GDPR and the EU AI Act all name specific dimensions.

## Why "good data" is not a measure

Ask three departments whether the customer data is good and you will get three answers. Finance means the totals reconcile. Sales means the phone numbers work. Compliance means the beneficial owners are documented.

All three are right. That is the problem.

Imagine this: your organisation launches a data quality programme, buys a profiling tool and produces a dashboard of 400 rules. Six months later nobody looks at it, because no one agreed which rules matter for which decision. Dimensions solve this by giving everyone the same vocabulary, so that "the data is bad" turns into "the onboarding date is 12% incomplete for clients acquired before 2020".

## The six dimensions

| Dimension | Question it answers | Example check |
|---|---|---|
| Completeness | Are the required records and values present? | Share of client records with a country of tax residence |
| Uniqueness | Is each entity recorded only once? | Number of suppliers with two or more active records |
| Consistency | Do values agree across records and systems? | Same legal entity identifier in the CRM and the accounting system |
| Timeliness | Is the data up to date and available when needed? | Days between a change of address and its update in the system |
| Validity | Is the value in the expected format and range? | IBAN structure, date formats, allowed product codes |
| Accuracy | Does the data match reality? | Sample of records checked against source documents |

The UK framework defines accuracy as "the degree to which data matches reality" and validity as "the degree to which the data is in the range and format expected". The difference matters. A perfectly valid postcode can still belong to the wrong customer. **Validity is cheap to automate; accuracy usually needs a reference outside the system.**

The framework also warns that dimensions compete. Publishing data faster can mean publishing it less complete, and those trade-offs should be explained to users rather than hidden.

## ISO/IEC 25012: the longer list

If you need a formal reference, [ISO/IEC 25012](https://iso25000.com/en/iso-25000-standards/iso-25012) describes a data quality model with 15 characteristics. It separates inherent quality, which concerns the data itself (accuracy, completeness, consistency, credibility, currentness), from system-dependent quality, which depends on the systems holding the data, such as accessibility.

Use it as a checklist, not as scripture. Few organisations need all 15 to make better decisions.

## Where Luxembourg rules ask for dimensions

| Rule | Who it concerns | Dimensions named |
|---|---|---|
| [ECB guide on risk data aggregation](https://www.bankingsupervision.europa.eu/ecb/pub/pdf/ssm.supervisory_guides240503_riskreporting.en.pdf) (May 2024) | Significant banks under direct ECB supervision | At least accuracy and integrity, completeness and timeliness, with tolerance levels |
| [GDPR, Article 5(1)(d)](https://gdpr-info.eu/art-5-gdpr/) | Every controller of personal data | Accuracy, kept up to date where necessary |
| [AI Act, Article 10(3)](https://artificialintelligenceact.eu/article/10/) | Providers of high-risk AI systems | Relevant, sufficiently representative, as far as possible free of errors and complete |

The ECB guide is the most concrete. It expects data quality indicators covering at least accuracy and integrity, completeness and timeliness, "including tolerance levels and robust correction processes", reported periodically to the management body. It is the operational side of [BCBS 239](/en/bcbs-239-risk-data-aggregation-luxembourg-banks/).

The GDPR adds a twist: accuracy is judged "having regard to the purposes for which they are processed". In other words, data quality is always fit for a purpose, never in the abstract.

## The human layer

Dimensions are measured by tools, but chosen by people. The pattern that works is simple.

First, pick the data elements that drive a decision or a report: a client's risk rating, a fund's net asset value, a product's master record. Then let the business owner of each element choose which dimensions matter and set the tolerance. IT builds the checks; it should not decide that 95% completeness is acceptable.

A threshold without an owner is just a number on a dashboard.

Finally, connect each failing check to its cause. [Data lineage](/en/data-lineage/) shows where a value came from, which is usually where the fix belongs. For a deeper treatment of how quality connects with [master data](/en/master-data-management-mid-sized-companies/), there is [a practitioner's book-length guide to master data and data quality](https://harro.lu/#books).

The good news: you do not need a programme to start. Three dimensions on ten critical data elements, measured monthly, beat four hundred rules nobody reads.

## What to do now

1. List the ten data elements behind your most important report or regulatory submission.
2. For each, name one business owner and let them choose two or three relevant dimensions.
3. Agree a tolerance level per dimension and what happens when it is breached.
4. Automate validity, completeness and uniqueness checks first; plan sample-based accuracy checks against source documents.
5. Report the results to management every quarter, with the trade-offs you accepted.

## Questions & answers
**What are the six dimensions of data quality?**
In the DAMA UK set used by the UK Government Data Quality Framework: completeness, uniqueness, consistency, timeliness, validity and accuracy.

**What is the difference between validity and accuracy?**
Validity checks whether a value has the expected format and range, for example a correctly structured IBAN. Accuracy checks whether the value matches reality, for example whether it is the customer's actual account.

**Do Luxembourg regulations require specific data quality dimensions?**
Some do. The ECB guide for significant banks names accuracy and integrity, completeness and timeliness; the GDPR names accuracy for personal data; the AI Act names error-free and complete training data for high-risk systems.

**Should every dataset be measured on every dimension?**
No. The UK framework says organisations should prioritise the dimensions that align with user and business needs, and may decide some are not relevant.

## Sources
1. [The Government Data Quality Framework (3 December 2020)](https://www.gov.uk/government/publications/the-government-data-quality-framework/the-government-data-quality-framework) · UK Government Data Quality Hub
2. [ISO/IEC 25012 data quality model](https://iso25000.com/en/iso-25000-standards/iso-25012) · iso25000.com
3. [Guide on effective risk data aggregation and risk reporting (May 2024)](https://www.bankingsupervision.europa.eu/ecb/pub/pdf/ssm.supervisory_guides240503_riskreporting.en.pdf) · ECB Banking Supervision
4. [Regulation (EU) 2016/679 (GDPR), Article 5](https://gdpr-info.eu/art-5-gdpr/) · gdpr-info.eu
5. [Regulation (EU) 2024/1689 (AI Act), Article 10](https://artificialintelligenceact.eu/article/10/) · artificialintelligenceact.eu
