Quality Assurance & Compliance Flashcards
7 cards from real CAP practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Quality Assurance & Compliance flashcards as text
An analytics model used for credit decisions must comply with fair lending laws. Which testing approach directly addresses this compliance need?
Answer: Disparate impact analysis across protected classes
Disparate impact analysis tests whether model outcomes disproportionately disadvantage protected groups, a core fair lending requirement.
What is the main purpose of establishing acceptance criteria with stakeholders before building an analytics solution?
Answer: To define measurable standards the solution must meet to be considered successful
Pre-agreed acceptance criteria provide objective, measurable standards for judging whether the deliverable meets requirements.
A model card or fact sheet accompanying a deployed model typically serves what quality assurance function?
Answer: Communicating intended use, performance, and limitations to downstream users
Model cards document intended use cases, performance characteristics, and known limitations so users apply the model appropriately.
An analyst discovers a coding error that materially changed results already delivered to a client. According to professional conduct standards, the analyst should:
Answer: Promptly disclose the error and provide corrected results
Professional integrity requires prompt disclosure and correction of material errors affecting delivered work.
Which control best mitigates the risk of unauthorized changes to a production analytics model?
Answer: Change management with approvals and separation of development and production environments
Formal change management and environment separation ensure only reviewed, approved changes reach production.
In data governance, what does 'data lineage' describe?
Answer: The documented origin, movements, and transformations of data through systems
Data lineage traces where data came from and how it was transformed, supporting auditability and trust.
A model performs well overall but fails badly on a small, high-stakes customer segment. From a quality assurance standpoint, what should the team do?
Answer: Evaluate and report performance by segment and address the high-risk gap before deployment
Segment-level evaluation reveals localized failures that aggregate metrics hide, especially where errors are costly.