Key takeaway
A sampling result supports a particular lot decision under a particular rule. Preserve both the calculation and who was authorized to act on it.
A passed lot needs more than a status field
An inspection archive may contain thousands of accepted lots without retaining the rule behind those decisions. Ask whether the archive preserves lot boundaries, the plan version, selected units, findings, counting rules and authorized disposition. A final status alone cannot show whether the sample was complete, whether the correct defect category was counted or whether a supervisor approved a separate exception.
NIST describes acceptance sampling as a way to decide whether to accept or reject a lot. It explicitly distinguishes that purpose from estimating the quality of the lot. This matters when describing an archive: an accepted batch is not evidence that every unit was inspected or defect-free. The statistical guide, checked 8 October 2026, does not choose an appropriate plan for your product.
Reproduce a completed decision
This hypothetical incoming lot L80 contains 800 units. Its invented approved single-sampling plan S3 revision 2 selects 50 units and permits at most one unit with a counted minor defect. All 50 receive the specified inspection. Unit A17 has two minor findings; A43 has one. Because the local rule counts defective units rather than findings, the count is two, not three.
The completed record yields rejection under that rule: two defective units exceed the acceptance count of one. The inspection lead records observations, and the designated quality manager authorizes the lot disposition. The table illustrates evidence reconstruction; it does not recommend 50 units, this defect classification or this acceptance count for any real product. An actual plan requires its own risk, contractual and technical review.
| Evidence field | Completed illustrative value | Consequence |
|---|---|---|
| Lot and approved basis | L80; 800 units; plan S3 revision 2 | One identified population and rule |
| Selected and inspected units | 50 identified units; 50 inspection records | Planned sample completed |
| Counted findings | A17: two minor findings; A43: one | Two distinct defective units |
| Acceptance rule | Accept if defective-unit count is at most 1 | Count 2 exceeds 1; reject |
| Authority and disposition | Inspection lead observes; designated quality manager records rejection | Findings and authorized action remain distinct |
Store how the sample was selected
Retain the population from which units were selected, the selection procedure and the identities or traceable positions of selected units. Record substitutions and inaccessible units rather than quietly replacing them with convenient observations. If a historical sample was drawn only from the top of a pallet, describe that procedure accurately. Calling it random after the fact changes the meaning of the evidence.
Keep unit observations separate from the plan result. One unit can have several findings, and a plan may count a different measure from the raw number of defect codes. The illustrative review checks the two unit identifiers before applying the threshold. If the archive retains only an aggregate failure count, a reviewer may be unable to verify duplicate findings, classification changes or the completed sample size.
For a practical replay, compare the selected-unit register with the observations, group findings by inspected unit and apply the stated classification before counting. In L80, the reviewer can point to A17 and A43 rather than relying on a dashboard total. Retain the intermediate counted-unit list with the result. If a later classification correction changes whether a finding counts under S3, preserve both calculation versions and assign a new disposition review. A corrected count should not overwrite which evidence and rule supported the original authorized lot decision.
Do not replace plan risk with the observed count
NIST explains that single plans use a sample size and acceptance count, while other plans can require additional sampling and different decision rules. It also describes operating-characteristic curves and producer and consumer risks. A recorded zero or small count does not substitute for those plan characteristics. Preserve the plan basis separately from the count observed on this particular lot.
For an actual review, ask who approved the plan and defect definitions, which requirements it addressed and what changed between revisions. Locate the authority for a disposition exception rather than rewriting the original result. A conditional release can coexist with an inspection rejection when the records clearly explain the authorized decision. Whether that release was permissible is a question for the responsible quality and contract owners.
Use the decision record to narrow an evaluation
A prospective evaluator may want to study sampling decisions, individual defect observations or supplier-level trends. Those require different evidence and permissions. State whether the archive includes rejected and accepted lots, repeat inspections, plan changes and disposition exceptions. Do not present a filtered set of accepted lots as the entire inspection history or imply that a downstream model can replace required inspection authority.
The internal deliverable is a plan-to-lot reconciliation for one period, with unknown rules and missing unit records flagged. Inventory can capture these evidence categories; due diligence can establish which decision an evaluator actually needs. Use a fictional example to discuss the structure first. A later sample needs separate approval, including review of supplier terms, sensitive drawings, inspector details and information that could reveal confidential acceptance criteria.