Data Quality Dimensions: A Clear Guide to the 6 DAMA Measures

Data quality dimensions are what turn “the data is bad” into something a team can measure and fix. DAMA UK’s 2013 working group settled on six primary dimensions for data quality assessment — completeness, uniqueness, timeliness, validity, accuracy and consistency — and they remain the reference set because they are narrow enough not to overlap and broad enough to cover what actually goes wrong.

This guide defines each one, gives the measure that goes with it, and covers the two dimensions that cannot be assessed the way the others can.

The six data quality dimensions

Validity and accuracy are not the same thing

This is the distinction that decides whether a data quality programme finds real problems. Validity asks whether a value obeys the rules; accuracy asks whether it is true. A date of birth of 01/01/1900 is perfectly valid — correct format, plausible range, passes every rule you wrote — and almost certainly wrong. A postcode that exists is valid; whether it is the customer’s postcode is accuracy.

Validity can be measured by machine against a rule set, cheaply and continuously. Accuracy generally cannot: it requires comparison against a trusted external source or a sampled verification exercise with a human in it. That difference in cost is why most dashboards quietly measure validity and label it accuracy.

Consistency is a cross-system measure

Consistency only means something once you have named the systems being compared. The same customer’s address in the CRM, the billing platform and the support tool either agrees or does not, and the disagreement rate is the measure. Within a single table, what people usually call consistency is validity with a conditional rule.

Using the data quality dimensions without building a bureaucracy

Measuring all six data quality dimensions across every attribute in the estate is how data quality programmes die. The version that survives is narrower:

1. Pick the critical data elements first. The handful of attributes that regulatory reporting, billing and customer contact actually depend on. Twenty attributes measured is worth more than two thousand profiled.
2. Choose the dimensions that matter per element. For a customer email address, validity and uniqueness carry the risk. For a price, accuracy and timeliness do. Not every dimension applies to every element.
3. Write the rule with the business, not for it. A validity rule reflects a business decision about what is acceptable; if the data owner did not agree it, the failures will be argued rather than fixed.
4. Set a threshold and an owner per rule. A measure with no threshold produces a number nobody acts on, and a threshold with no owner produces a breach nobody fixes.
5. Track the trend, report the exceptions. The useful report is which rules breached threshold this month and what is being done, not a wall of percentages.

Where data quality dimensions sit in a governance programme

Data quality is one knowledge area among several, and the dimensions are the measurement layer beneath it. They connect outward in three directions worth wiring deliberately:

• To stewardship. Every rule needs a data owner who can decide what “acceptable” means and a steward who chases the failures. Rules without those two roles become monitoring for its own sake.
• To the catalog. A rule attached to a defined element in a catalog can be found, reused and inherited. A rule embedded in a report cannot.
• To compliance obligations. Accuracy is a legal requirement in several privacy regimes, and a documented data quality measure is one of the cheaper ways to evidence it. Our guide to data governance and the DMBOK covers the wider structure.

Frequently asked questions

How many data quality dimensions are there?
Six primary dimensions in the DAMA UK reference set: completeness, uniqueness, timeliness, validity, accuracy and consistency. Longer lists exist, but they tend to subdivide these rather than add genuinely new ones.

What is the difference between validity and accuracy?
Validity means the value obeys the defined rules; accuracy means it is true of the real-world thing. A value can be perfectly valid and completely wrong.

Which dimension should we measure first?
Completeness and validity, because they can be measured automatically and usually surface the largest volume of fixable defects. Accuracy is the most valuable and the most expensive.

Do we need a tool?
Not to start. A rule set, a schedule and a spreadsheet of results will prove the value; tooling earns its place once the rule count grows past what a person can run.

Who owns data quality?
The data owner sets the acceptable threshold; a steward operates the measurement and chases remediation. Handing both to a central team is what makes quality “IT’s problem” and keeps it unfixed.

Where this leaves you

Use the six data quality dimensions as a vocabulary, not a checklist. Pick your critical data elements, choose the dimensions that carry the risk for each, agree the rules and thresholds with the people who own the data, and be honest about which of your measures are validity wearing accuracy’s name. Then report exceptions and trend rather than a grid of percentages — the point of measuring quality is that somebody fixes something.

References

• DAMA UK — The six primary dimensions for data quality assessment — the 2013 working group paper that defined the reference set.
• The Government Data Quality Framework — a public-sector application of the same dimensions.

More on data governance

• Data quality dimensions — you are here
• Data governance and the DMBOK
• Records of processing activities
• Data classification

Rule sets, stewardship roles and quality reporting templates are in the Data Governance Toolkit, or start with the free ISO templates.