Quality Assurance and Quality Control Systems
This domain tests how quality is planned, measured, and corrected: QA versus QC, sampling plans, calibration, statistical process control, root-cause analysis, and document control. Expect questions that distinguish a system failure from a product failure.
Core rules
- QA is process-oriented and preventive, covering systems such as GMPs, HACCP, ISO 22000, SQF, and BRCGS. QC is product-oriented and inspection-based, testing materials and finished goods against specifications.
- Acceptance sampling uses an AQL and an operating characteristic curve. AQL is not a guarantee of zero defects; it defines the quality level with a defined acceptance probability. Tightened, normal, and reduced inspection respond to supplier history.
- Calibration compares an instrument to a traceable standard; verification confirms that the instrument or process meets requirements. Measurement uncertainty must be considered when judging a result against a limit.
- SPC uses control charts: X-bar and R charts for variables, p or c charts for attributes. Western Electric rules flag trends, runs, and points beyond limits. A process can be in control but not capable; capability indices compare spread to specification.
- CAPA closes the loop: contain, investigate root cause with tools such as 5 Whys or fishbone, correct, verify effectiveness, and update documents.
Worked example
An X-bar chart shows seven consecutive points above the centerline, though all are inside the control limits. This violates a run rule and signals an assignable cause such as a new ingredient lot or a drifted filler. The correct action is to investigate the process, not to adjust the specification. If the points remain inside limits but capability is poor, the process spread relative to tolerances is the issue.
Traps that cost marks
Confusing accuracy with precision. Treating calibration as a one-time activity rather than scheduled with traceable standards. Assuming AQL zero means zero defects. Adjusting specifications to make out-of-control data appear acceptable.