By comparing three sources: what the policy states, what the system record shows, and what the people doing the work report. Each one alone is misleading; the divergence between them comes with the reason attached.
What are the three sources, exactly?
The document — policy, procedure, flowchart. The record — what the system actually stored about each case. The conversation — the why, which neither of the other two holds. See how they converge in the diagram in why the layer is a graph.
Why isn’t the policy enough on its own?
Because it describes the intent at the moment it was written, not the operation today. An out-of-date process on paper keeps looking correct — until it’s compared with what the system and the people say.
Why isn’t the system record enough on its own either?
Because it shows what was typed in, not what happened before it was typed in. A deadline can sit within target in the system and still have required a workaround the record doesn’t capture.
Why isn’t the conversation enough on its own either?
Because individual memory is selective and varies with who’s telling it. Without checking against document and record, an isolated account becomes opinion — and cross-checking is exactly what separates the two.
The document looks obsolete. The record looks like efficiency. The conversation looks like anecdote. None of the three, in isolation, would support a decision — together, they show a process nobody had designed that way.