What the data cannot answer comes back as a question.

Before any record moves, the definition of the audience is agreed. The same record arriving twice is collapsed, and which copy won is recorded. On a repeat build, only what actually changed is reported.

Source appraisal

Establishing where each field can come from, how current it is, and what it is permitted to be used for.

Brief interrogation

Reading a requirement back as filters, and asking about the parts that cannot be expressed that way.

Spelling normalisation

Settling variant forms of the same organisation name so counting does not double.

Start with a single segment and judge the rest on it.