Who this is for: quality managers and lab directors in CLIA-certified and FDA-regulated laboratories. Every observation below comes from published 483 data; every prevention is a system change, not a reminder to try harder.
1. Investigations that don't investigate (21 CFR 211.192 / 820.100)
The most-cited failure in FDA inspections, every year: a deviation record whose root cause restates the problem — "operator error," "equipment malfunction" — with no analysis behind it, and a corrective action that is really just a correction.
Prevent it structurally: make root cause analysis a required, evidence-attached phase that must be signed before actions can open, and gate closure on effectiveness verification after a real interval. If your system allows a CAPA to close the day its action completes, it is documenting the citation for the inspector.
2. The obsolete SOP at the bench
An inspector picks up the procedure taped near the analyzer. It's revision B; the effective version is D. One glance, one finding — and a data point about the whole document system.
Prevent it structurally: controlled distribution with obsolete-version lockout, and watermarked printouts that identify their version and print date. A sweep-the-benches policy is a patch; lockout is a fix.
3. Training records that can't survive a by-name sample
Inspectors sample people, not spreadsheets: one tech, one procedure, evidence of training on the current revision. Reconciled-quarterly matrices fail this test on the revision date every time.
Prevent it structurally: revision-triggered assignments — the new SOP version assigns training before work proceeds under it — and a matrix computed live from the records rather than maintained by hand.
4. QC failures with results already released
The finding with patient impact: a failed control discovered at weekly review, four days after the results shipped. CLIA 493.1256 requires control procedures that detect immediate error — "we review on Fridays" is an admission, not a procedure.
Prevent it structurally: acceptance criteria evaluated at the moment of entry, with affected results held automatically until a signed corrective decision.
5. Equipment used while out of calibration
Part 820.72 asks two things labs on paper systems can't answer: was anything measured with this instrument after it drifted, and what did you do about it? The impact assessment done by hand takes weeks and proves incompleteness.
Prevent it structurally: calibration schedules that escalate before the due date and quarantine at it — an overdue instrument that cannot accept results is a finding that cannot occur.
6. Audit trails nobody reviews
FDA's data integrity guidance expects routine audit trail review. Most systems write trails nobody opens; the trail itself then records the batch-signing and correction-cluster patterns the inspector will chart for you.
Prevent it structurally: exception-based review — the system surfaces off-pattern activity daily, and the disposition of each flag becomes the review evidence the guidance asks for.
7. Supplier controls that stop at the certificate
820.50 requires evaluation proportionate to risk and ongoing monitoring. The classic finding: a PO issued to a supplier whose qualification lapsed a year ago, discovered by comparing two documents you handed over yourself.
Prevent it structurally: an approved supplier list your receiving workflow actually checks, with expiry escalation on certificates and scorecards built from receiving and deviation data.
The pattern behind the patterns
None of these are knowledge failures — every lab knows the requirements. They are system failures: paper and spreadsheets can record compliance but cannot enforce it. The labs that stopped collecting 483s didn't train harder; they moved enforcement into the system, so the failure mode became structurally impossible.