Data completeness is the degree to which all records and values required for a defined purpose are present in a dataset. Teams measure it against explicit requirements, not against the assumption that every field must always contain a value. A dataset may therefore be complete for one task and incomplete for another when the tasks require different attributes, populations, time periods, or levels of detail.
For example, a customer table may contain every required account identifier but lack the consent status needed for a specific analysis. A simple populated-field percentage can identify missing values, but it cannot determine whether the available records represent the expected population or whether each value is accurate. For AI work, completeness checks should name the required fields and records, the intended use, the measurement scope, and the observation time. This makes missing context visible and allows teams to compare results without treating completeness as a complete readiness decision.