Handling deviations during qualification

Qualification rarely runs exactly to plan. Sensors fail, doors are left open, a set point cannot be reached. Each event is a protocol deviation to be recorded at the time, assessed for impact on the validity of the test, and resolved — by repeating the test, by accepting the result with justification, or by declaring the requirement unmet.

Failure to meet acceptance criteria

• Confirm the data are valid (sensor calibration, placement, timing).
• Investigate cause: equipment, environment, protocol design.
• Options: remediate and repeat; restrict use (exclude a location); reject the system.
• Record the decision, the rationale and the approvals.

A qualification report with no deviations recorded is more suspicious than one with several well-handled ones.

Three kinds, handled differently

Lumping every departure into one deviation category makes the log unreadable. Separating them makes the pattern visible.

Investigating a failure without deciding the answer first

The temptation with a failed test is to look for a reason the data are wrong. Sometimes they are — a sensor that was never calibrated, one placed against a cooling coil, a clock offset that put readings in the wrong order. That is why data validity is checked first. But it is checked first in order to be ruled out, not in order to be found, and an investigation that stops the moment it has an excuse is not an investigation.

The order that holds up: confirm the data are valid; establish what physically happened; establish why; then decide. A position that genuinely ran warm ran warm, and the honest outcomes are to fix the cause and re-test, to restrict use of that location and say so in the report and on the shelf, or to conclude the system is not fit for the intended purpose.

What the report has to carry

• Every deviation, numbered, including the ones that turned out to have no impact.
• The assessment of impact on the individual test and on the overall conclusion — these are different questions, and a test that is individually fine can still leave a conclusion unsupported.
• The resolution and who approved it, at the same level as the protocol approval.
• Any restriction that follows the system into routine use, such as a location excluded from product storage, with the mechanism that enforces it.
• Open deviations, if any remain. A qualification can be approved with open items only where they are explicitly listed and cannot affect the conclusion.

A qualification report with no deviations recorded is more suspicious than one with several well-handled ones. Real studies of real rooms produce surprises; a perfectly clean record usually means the surprises were not written down.