A customer wants an auditable, scalable way to prove their models were properly evaluated before release. What is the best recommendation?
Select an answer to reveal the explanation.
Short Explanation
You need a factsheet and evidence record for each model to show its purpose, data, evaluations, and decisions. That gives you auditable proof that survives audits, unlike memory, personal notes, or blanket statements.
Full Explanation
Factsheets and evidence records create a structured, per-model artefact that is retrievable and auditable, which is what scalable proof of evaluation requires. Relying on the development team's collective memory is incorrect because institutional memory is neither reliable nor auditable as teams and staff change. Keeping evaluation results only in personal notes of whichever data scientist ran them is incorrect because personal notes are not accessible to auditors or to the organisation once that individual moves on. Publishing a single company-wide statement asserting that all models are evaluated responsibly is incorrect because a general statement provides no model-specific evidence at all. For the exam, recommend factsheets and evidence records as the auditable mechanism, not informal alternatives.