Resources/Mechanical findings vs judgment

Guide

Mechanical findings vs judgment

Most of the pain in document review is not exotic science. A lot of what gets written up is checkable on the page — if anyone has the time. The rest is judgment. Mixing the two is how tools overclaim and reviewers burn out.

Two different jobs

Mechanical means a careful reader could point at text (or its absence) and agree something is inconsistent, incomplete, or circular — without deciding whether the product is safe or the investigation is scientifically sound.

Judgment means interpreting evidence, scope, risk, and adequacy in context: was this the right method, the right batch impact conclusion, the right severity? Two competent reviewers can disagree, and both can be acting in good faith.

Usually mechanical

  • Approval dated before the event it signs
  • Missing version / revision / effective date
  • Root cause that only restates the symptom
  • Undefined abbreviations used as if defined
  • Internal contradictions (limit A vs limit B in the same doc)
  • Empty signature block or blank required field the template expects
  • Cross-reference to a revision that does not match the attachment

Needs judgment

  • Whether the root cause is true, not merely non-circular
  • Whether impact assessment covered the right batches
  • Whether the CAPA is proportionate to risk
  • Whether data support a shelf-life or release conclusion
  • Whether a deviation should have been escalated
  • Whether a signature is authentic (not merely present)

Why the split matters under deadline

Inspection-readiness work is scarce attention. If a human spends the first pass hunting date order and missing identifiers, they have less left for the judgment calls an inspector will actually test in the room.

Mechanical triage is not a shortcut around science. It is a way to stop the avoidable stuff from consuming the budget of care.

What a triage tool should do

PASS on a criterion is not “document approved.” It only means no issue was found against that one rule on this pass. A full PASS list is still a work list of zeros — not a certificate.

Where people get burned

If someone skips their own review because a tool looked clean, and the tool was wrong, the tool was worse than useless.

In Dry Run every finding is per-criterion: PASS, FLAG, or REVIEW, with evidence and rationale. The summary is a count of work — not a grade. Signature-related rules are constrained so they cannot return PASS as “signature verified.”

Browse the criteria →  ·  Example: circular root causes →

See per-criterion findings on sample data — labeled simulated, no login.