What does completeness mean?
Completeness asks whether all required values are present. A customer record without an email address is incomplete for a newsletter campaign, even if every other field is correct. It can be measured as the proportion of empty required fields, tracked per record and over time.
What does correctness mean?
Correctness is about whether a value accurately reflects reality. An outdated delivery address can be complete and still be wrong. Correctness is checked by comparing data against a reliable reference — such as an address validation service or a master system.
What does consistency mean?
Consistency checks whether the same information matches across systems and records. If a customer has a different name in the CRM than in the ERP and the online store, consistency has been violated. According to [Gartner]{slug:https://www.gartner.com/en/data-analytics/topics/data-quality), inconsistency is considered the hardest quality problem to resolve because it arises from separate silos. It can be measured through cross-system comparisons of defined key fields.
What does timeliness mean?
Timeliness indicates whether data is still valid at the point of use. Data begins to age from the moment it is captured — some of it within minutes, some of it never. It is assessed by comparing the age of a record against the maximum permissible age defined by the business: inventory levels need to be accurate in near real time, while a date of birth remains valid indefinitely.
What does uniqueness mean?
Uniqueness requires that every real-world entity is represented exactly once. Duplicates — the same customer appearing under multiple spellings — distort every count and analysis. They can be identified through matching rules and by tracking the proportion of merged records.
Are there additional dimensions?
Yes. Depending on the framework, two more are commonly added. Validity describes whether a value conforms to the permitted format and value range; a postal code containing letters is invalid. Integrity describes whether the relationships between records are correct; an order with no associated customer is an integrity violation. Which dimensions you adopt should depend on what your actual use cases require.
How are these dimensions measured?
Each dimension can be expressed as a metric — typically the proportion of records that meet a given criterion. This turns the statement "our data is poor" into a number that can be tracked and improved. How to assess the baseline is covered in the article on Evaluation; how to turn that into ongoing measurement is covered in the article on Monitoring.
Back to overview: Data Quality