LEARN. PRACTICE. GROW.Concept and mechanism
Data quality depends on intended use. Completeness indicates whether required attributes are populated; accuracy requires them to represent reality appropriately. Consistency concerns compatible representations, such as country codes and units. Freshness matters when information can change. Duplication affects counts, relationships, and repeated decisions. Usage helps determine whether information serves intended processes. A file can be complete but wrong, recently exported but stale in substance. Measuring only nonempty fields encourages population without truth. A quality transformation preserves meaning: passing technical validation is insufficient. Criteria should identify attributes critical to the use case and how those attributes will be confirmed.
Guided application
When preparing customer data, start with identity and provenance. Different systems may assign different identifiers to the same entity. Similar names, however, may belong to distinct people. Define matching criteria, review exceptions, and preserve relationships before merging records. If operators classify the same request inconsistently, reconcile instructions and jointly review a sample. Converting labels to numbers does not resolve conflicting meaning. For emergency contacts, confirming the current person and role matters more than updating a technical record date. Recording criteria helps explain to a sponsor why a dataset still needs work despite an error-free import.
Example: all 1,200 contacts have a phone number, but 170 were not confirmed within the agreed period. Completeness is 100%; this does not prove all 1,200 numbers remain correct.
Common pitfalls
Merging by name; load date treated as freshness proof; quantity treated as quality; inconsistent labels hidden by shared coding.
Related topics: Choose the AI task and quality evidence · CRM capabilities, recommendations, and actions · Trust principles and real consequences
Accept data by meaning and use, with evidence of identity and quality.
Reference: Assess the Quality of Data · AI Associate historical Spring 24; retired 2026-02-02