Manufacturing Retest Authorization Evidence Rate

A failed test is repeated, passes, and the first result vanishes from the release packet. Manufacturing retest authorization evidence rate checks whether each retest had a valid reason, scope and permission before the new result was used.

Define what counts as a retest under the product and inspection plan, since a scheduled second sample is different from a repeat ordered because the first result failed. Distinguish retest from reinspection after approved rework, because Oracle describes reinspection of initially rejected work-in-process units after they are reworked and that sequence has its own trace.

Preserve the original result and raw evidence, so a later passing value does not silently overwrite a prior failure or create a false impression of first-pass quality. Identify the trigger, such as instrument fault, sample damage, procedure error, confirmed data import problem or approved rework, because unsupported dissatisfaction with a failing result is not a sound rationale.

Record who can authorize a repeat under the control plan, since the operator who sees the failure may not have authority to decide the retest scope alone. State whether approval was needed before a new sample or only before using its result, and link the authorization to unit or lot, test characteristic, method, sample plan and maximum repeats where relevant, because a general permission to try again is too broad.

When equipment failure is claimed, retain instrument logs and investigation, as a passing result from a different device does not itself prove the first device was wrong. If the sample was mishandled, preserve chain-of-custody and sampling evidence, otherwise it is difficult to distinguish a true invalid sample from convenient resampling.

Where a test is destructive, choose a new sample under the approved plan, because retesting the same unit may be impossible or may change its properties. If rework occurred, link the rework instruction, completion and new inspection plan, and for serial-tracked products ensure the new result attaches to the same serial or approved replacement, since results from a neighbouring unit cannot cure a failure.

For batch testing, document whether the new sample represents the same material population, because mixing or sorting after the first sample changes the population being judged. Record each attempt in sequence with date, analyst, method and outcome, as an average of results may be inappropriate unless the approved procedure explicitly permits it.

The US FDA has product-specific guidance for investigating out-of-specification results in pharmaceutical production, which is a regulatory example, not a universal rule for all manufacturing. Keep investigation and final disposition separate from authorization, because a retest can be properly authorized yet still fail or leave product on hold, and if an unauthorized repeat was done, retain the data and investigate the process since deleting the repeat would compound the record problem.

Define the denominator as retest events requiring authorization, not every routine replicate, and classify missing reason, wrong approver, late approval, excessive attempts and missing original data separately, as each points to different corrective action. For urgent production establish an available qualified approver rather than waiving review informally, audit selected retests backward from final disposition to all attempts, and show patterns by product, instrument and shift, remembering that the target is justified retesting and clear evidence, not a high pass rate after enough attempts.