Risk-Stratified Acceptance Sampling Plan¶
Sampling protocol — instantiates Conformance Control and Corrective Feedback
Sets inspection intensity by defect risk and criticality, then accepts or rejects each lot on a predeclared sample rather than checking every unit.
When inspecting every unit is destructive, slow, or simply uneconomic, you inspect a sample and let it stand in for the lot. Risk-Stratified Acceptance Sampling Plan is the design that decides how much to look at and how hard, deliberately spending more inspection on the characteristics whose escape would hurt and less on the ones that wouldn't. Its defining move is that sample size and the accept/reject number are not fixed by habit but derived from risk — defect severity, how detectable the defect is, and how capable the supplying process has proven to be — so a safety-critical dimension gets tight scrutiny while a cosmetic one rides on a small sample. It is a plan for lot-level evidence, not for judging individual units; its output is "accept this lot / reject this lot," and it hands the actual pass/fail comparison and the release decision to other mechanisms.
Example¶
A manufacturer receives incoming lots of ≈10,000 fasteners a week from a supplier. Inspecting all of them is out of the question, and most attributes (finish, marking) rarely fail. But a few fasteners are structural, and a soft one escaping into an assembly is expensive and dangerous. The plan stratifies: cosmetic attributes get a small attribute sample under a loose acceptance quality level; the structural hardness and thread characteristics — the critical-to-quality set — get a larger sample and a near-zero accept number. The team anchors the numbers on a published attribute-sampling standard[n1] rather than inventing them, and adds a switching rule: after a lot fails, inspection tightens automatically; after a run of clean lots, it can relax.
The result is a written plan: "draw n=200 from each incoming lot, accept on ≤1 nonconforming for cosmetic attributes and 0 for structural; tighten to n=315 after any rejected lot." That plan is what lets the receiving dock turn a mountain of parts into a defensible accept-or-hold decision in an hour — without pretending a small sample tells them anything about the rare defect it was never sized to catch.
How it works¶
- Stratify by criticality. Split characteristics into severity classes and assign each class its own inspection intensity — the critical set drives the tight plans, the trivial set the loose ones.
- Size the sample from risk, not convenience. Choose sample size and acceptance number so the plan's producer's and consumer's risk match the escape cost — larger samples and lower accept numbers where a miss is expensive.
- Balance the ledger. Trade appraisal cost and throughput against expected escape cost; the plan is the point where inspecting more stops being worth it for that characteristic.
- Switch on evidence. Tighten after failures, relax after sustained clean history, so the plan tracks the supplier's actual performance instead of a static assumption.
Tuning parameters¶
- Acceptance quality level (per class) — the defect rate the plan is built to pass. Lower it for critical characteristics; the trade is more inspection and more good lots occasionally rejected.
- Sample size vs. lot size — bigger samples cut consumer's risk but cost time and parts; the curve flattens, so doubling the sample rarely halves the risk.
- Switching-rule aggressiveness — how fast the plan tightens after a bad lot and relaxes after good ones. Aggressive switching reacts to drift but thrashes on noisy suppliers.
- Attribute vs. variable sampling — pass/fail counts are simple but weak; measuring a variable and using its distribution extracts more information from a smaller sample, at more measurement effort.
- Stratification granularity — how many severity classes. Finer classes target effort better but add plan complexity and room for misclassification.
When it helps, and when it misleads¶
Its strength is making high-volume or destructive inspection feasible at all, and making the cost/coverage trade explicit and risk-proportionate rather than "inspect whatever is convenient." Tightened-and-reduced switching lets a plan follow a supplier's real trajectory instead of freezing one assumption.
Its central deception is that a sample sized for the average defect says almost nothing about a rare, clustered, or critical one — the operating-characteristic curve that governs the plan buys down producer's and consumer's risk only for the defect rate it was designed around.[n2] Convenient or predictable sampling (top of the pallet, first off the line) quietly destroys representativeness, and a plan is often run backwards — sample size chosen to hit a throughput target, then justified after the fact. The disciplines that keep it honest are a defined sampling frame with randomization or stratification, and the rule that an intolerable critical escape is never bought down by ordinary sampling — it gets 100% or automated inspection instead.
How it implements the components¶
measurement_and_sampling_plan— its core output: the sampling frame, sample size, draw method, and accept/reject numbers per characteristic.critical_to_quality_characteristic_map— the stratification step classifies characteristics by escape consequence and detectability, driving where inspection intensity goes.quality_cost_and_throughput_balance— the sample size is the balance point between appraisal cost/throughput and expected escape cost.
It does not execute the unit-level pass/fail comparison (that's Automated Conformance Check) or hold the accepted lot until authorized release (that's Release Hold and Signoff); nor does it establish whether the gauge is trustworthy — Measurement-System Capability Analysis does that.
Related¶
- Instantiates: Conformance Control and Corrective Feedback — this plan supplies the evidence-gathering design the conformance gate acts on.
- Consumes: Measurement-System Capability Analysis — a plan is only as trustworthy as the gauge behind each measured sample.
- Compare: the plain Acceptance Sampling Plan sizes one sample to the lot's overall risk; this plan stratifies that decision per characteristic, giving each criticality class its own sample size and accept number.
- Sibling mechanisms: Automated Conformance Check · Control Chart and Trigger Rule · Release Hold and Signoff · First-Article and Setup Approval · Nonconformance Report and Review Board · Measurement-System Capability Analysis
Editorial Notes¶
Form Classification¶
Form family: Experiment, Test & Rehearsal
Rationale: Risk Stratified Acceptance Sampling Plan operates by actively draws and inspects risk-sized samples to generate acceptance evidence by criticality class. That concrete deployed or enacted form is Experiment, Test & Rehearsal under the frozen taxonomy.
Nearest alternative: Protocol, Workflow & Routine — Although Protocol, Workflow & Routine can support this mechanism, the frozen evidence makes its operative form the act that actively draws and inspects risk-sized samples to generate acceptance evidence by criticality class; the alternative is therefore secondary rather than defining.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Single lineage
Present-day reach: Specialized
Rationale: Acceptance sampling by defect risk and lot criticality is canonical statistical quality control.
Related originating lineages:
- Data Science & Analytics — Data science, analytics, and operational monitoring supplies a parallel or contributing lineage for the mechanism's defining operation: sets inspection intensity by defect risk and criticality, then accepts or rejects each lot on a predeclared sample rather than checking every unit.
- Engineering & Design — Manufacturing assurance materially applies the sampling plans.
- Mathematics — Mathematical modeling, proof, and abstract-structure practice supplies a parallel or contributing lineage for the mechanism's defining operation: sets inspection intensity by defect risk and criticality, then accepts or rejects each lot on a predeclared sample rather than checking every unit.
- Operations Research — Inspection-resource optimization contributes risk-tiered intensity.
Review resolution: Both blind reviewers agree that statistics_experimental_design is the primary historical origin. Explicit reconciliation of alternate origin disagreement starts from reviewer_a’s mechanism-specific evidence: Acceptance sampling by defect risk and lot criticality is canonical statistical quality control. Reviewer A proposed alternates=engineering_design, operations_research, origin_mode=single_lineage, domain_reach=specialized, and encyclopedia_synthesis=false; reviewer B proposed alternates=data_science, engineering_design, mathematics, origin_mode=single_lineage, domain_reach=specialized, and encyclopedia_synthesis=false. The final record retains every independently supported alternate from either review (engineering_design, operations_research, data_science, mathematics) without an arbitrary cap, selects origin_mode=single_lineage to represent the combined lineage evidence, and keeps domain_reach=specialized and encyclopedia_synthesis=false from the more mechanism-specific assessment. Present-day transfer is recorded as reach and is not treated as proof of historical origin.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
Acceptance sampling is a release decision aid, not a process-improvement tool — it tells you whether to take this lot, not why the supplier's process drifts. When the same lots keep failing, the answer is not a tighter sampling plan but upstream corrective action; sampling that only sorts, forever, is the classic inspection-as-sorting trap the archetype warns against.
[n1] Published attribute acceptance-sampling systems (e.g. ANSI/ASQ Z1.4, historically MIL-STD-105) provide standard sample-size and accept-number tables indexed by lot size and acceptance quality level, with normal/tightened/reduced switching rules — the reason a real plan cites a standard rather than hand-picking numbers. ↩
[n2] The operating-characteristic curve plots probability of accepting a lot against its true defect rate; it is the honest picture of a sampling plan's discrimination, and it is why a plan sized for a 1% defect rate gives almost no protection against a rare critical defect. ↩