ISO 8000 - Data Quality 8:5.2: 5.2 Semantic quality

What does 8:5.2 require?

Semantic quality is about each identifiable unit of data corresponding, uniquely and without ambiguity, to the real entity it stands for, as that entity is seen through a conceptual model: in short, how far the data match what they describe. The correspondence is shown by checking six criteria, each illustrated with an employee register: complete mapping (every entity in the domain is represented; broken when an employee is missing); consistent mapping (each entity is represented by no more than one unit, or by several units that agree, or by several whose disagreements are cleared up within an acceptable time; broken when one employee has two numbers); meaningful mapping (every unit stands for at least one real entity; broken by a number that belongs to nobody); unambiguous mapping (every unit stands for no more than one entity; broken when two employees share a number); correct mapping of entities (every unit points to the right entity; broken when two employees' numbers are swapped); and correct mapping of properties (every attribute value matches the real entity's property; broken by a wrong address). A semantic verification meets this part only when there exist: a conceptual model, set down in writing, of the universe of discourse (the part of the world in scope), relationships included; the data shown in a form that lets them be read through that model; the verification methods prescribed; how the rules drawn from the criteria are turned into measurements; for every semantic rule, how compliance is measured and what failing it means; a list of entities or types in the model that are not checked; how departures are recorded and reported; and the counts of checks run and of occurrences passing, per rule. Because checking every unit against the real world is rarely practical, sampling and statistics can give a probabilistic picture, and an already verified electronic record (a trusted surrogate) can be checked against instead.