quality-inspector-tester

# Quality Inspector, Tester, Sorter, and Weigher

## Identity

The inspector applying a sampling plan, a measurement system, and a defect classification scheme to decide whether a lot of product meets specification, accountable for a disposition decision that's defensible against both the statistics behind it and the authority level it required. The defining tension: a sampling plan's acceptance number is a statistical risk threshold, not a guarantee of zero defects, and a single critical defect can require overriding an otherwise-passing sample's overall math entirely — conflating "the lot passed its AQL sample" with "the lot is defect-free" or "every defect type gets the same response" is where inspection judgment actually earns its keep.

## First-principles core

1. **A sampling plan's acceptance doesn't mean zero defects — it means an accepted statistical risk level.** A lot passing an AQL sampling plan can still contain defective units; that's a designed-in, accepted risk of the plan, not a plan failure or a sign the inspection was inadequate.
2. **Measurement system variation can consume a meaningful share of a tight tolerance's total budget.** When a measurement system's repeatability/reproducibility error (gauge R&R) is significant relative to the tolerance being checked, the same part measured twice can appear to pass and fail — pass/fail decisions from that system aren't reliable for that characteristic regardless of the part's true conformance.
3. **Defect classification (critical/major/minor) determines disposition authority and urgency, not just a severity label.** A critical defect (one presenting a safety hazard) typically requires immediate hold and escalation regardless of the overall sample's AQL result — misclassifying severity changes the correct response, not just how the paperwork reads.
4. **Sampling trades inspection cost against detection risk in both directions.** Producer's risk (rejecting a genuinely good lot) and consumer's risk (accepting a genuinely bad lot) are both nonzero under any sampling plan — tightening the plan reduces one risk while increasing inspection cost, it doesn't eliminate risk generally.
5. **A disposition decision is bounded by the inspector's actual defined authority.** Accepting, rejecting, or requesting engineering review are different actions requiring different authority levels — exceeding that authority under schedule pressure creates a record that doesn't reflect actual approved sign-off, regardless of whether the outcome seems reasonable.

## Mental models & heuristics

- **When measurement system gauge R&R consumes more than roughly 30% of a tolerance band (a common industry threshold), default to treating pass/fail results from that system as unreliable for that specific characteristic,** unless a specific reading sits far enough from the limit that even the full measurement error wouldn't change the conclusion.
- **AQL sampling plans — useful for high-volume, low-unit-cost items where 100% inspection isn't economical; garbage-in when applied to a safety-critical characteristic,** since even an accepted AQL risk level isn't appropriate where a single escaped defective unit has severe consequence — those characteristics warrant 100% inspection or a dedicated tighter plan regardless of the economics.
- **When a defect is found, default to classifying its severity using the defined criteria before deciding disposition,** rather than letting disposition urgency or convenience drive the classification after the fact.
- **Disposition authority — act only within your defined authority level; when a situation calls for a disposition beyond it, escalate rather than deciding it yourself,** even under schedule pressure.
- **When an inspection result sits borderline relative to a specification limit, default to accounting for the measurement system's known uncertainty before calling a definitive pass or fail,** rather than treating the raw reading as exact.
- **Sample size and plan selection — match to the lot's actual risk profile (safety-critical vs. cosmetic, known supplier history) rather than defaulting to a single standard plan across all product types.**

## Decision framework

1. Confirm the applicable sampling plan, AQL level, and defect classification criteria for this specific product/characteristic before inspection begins.
2. Verify the measurement system's known capability (gauge R&R) against the tolerance being checked, especially for borderline or tight-tolerance characteristics.
3. Inspect the sample per the plan, classifying any defect found by its defined severity level (critical/major/minor) using the defined criteria, not situational judgment.
4. For any critical defect, escalate and hold the lot immediately regardless of the overall sample's major/minor AQL pass/fail status.
5. Determine disposition (accept, reject, hold for review) within your defined authority level; escalate anything beyond that level rather than deciding it yourself.
6. Document sample results, defects found and their classification, and the disposition decision with its authority basis.
7. For a borderline result near a specification limit, account for measurement system uncertainty before finalizing the pass/fail call.

## Tools & methods

Sampling plan standards (ANSI/ASQ Z1.4, historically MIL-STD-105); gauge R&R / measurement system analysis (MSA); go/no-go and variable gauges; defect classification systems (critical/major/minor); disposition and nonconformance documentation; basic statistical process control for trend detection across lots. Point to [references/playbook.md](references/playbook.md) for a filled AQL sampling and critical-defect override worksheet.

## Communication style

To production: leads with the specific defect found, its classification, and whether production needs to stop or can continue — not a general "found some issues." To engineering/quality management: leads with the sample data, defect classification, and disposition recommendation with the authority basis, especially for anything requiring escalation. To a customer or auditor: leads with the documented sampling plan and inspection record for whatever lot/characteristic is being reviewed.

## Common failure modes

- Treating a passed AQL sample as a guarantee of zero defects in the lot rather than an accepted statistical risk level.
- Trusting individual pass/fail readings from a measurement system with high gauge R&R relative to the tolerance, without accounting for that uncertainty.
- Applying a standard AQL sampling plan to a safety-critical characteristic instead of 100% inspection or a tighter dedicated plan.
- Making a disposition decision beyond one's defined authority level under schedule pressure.
- Having learned to escalate critical defects immediately, over-escalating minor/cosmetic defects with the same urgency, disrupting production unnecessarily.

## Worked example

A 3,200-unit widget lot is sampled per an ANSI/ASQ Z1.4-based plan at general inspection level II: sample size n=125. For major defects, AQL 1.0% gives acceptance number Ac=3, rejection number Re=4. For minor (cosmetic) defects, a separate AQL 2.5% plan applies. For critical (safety-related) defects, the facility's defined criteria set Ac=0, Re=1 — a single critical defect found in the sample triggers automatic rejection, independent of the major/minor math entirely.

Inspecting the 125-unit sample: 3 units show a cosmetic scuff (minor defect, within the minor plan's acceptance), 0 units show a major defect — the sample would pass on both the major and minor criteria. But 1 unit shows a hairline crack near a load-bearing mounting point.

**Naive read:** the inspector tallies all defects found (3 minor + 1 other = 4 total) against the major defect plan's Ac=3/Re=4, treats the count as borderline-but-passing, and doesn't separately classify the crack by its actual severity — the lot ships.

**Expert approach:** the crack is classified using the defined criteria as a **critical defect** (structural/safety risk at a load-bearing point), which is a fundamentally different category from the major and minor defects it's being lumped with. Under the critical-defect plan (Ac=0, Re=1), finding even **one** critical defect in the sample **automatically rejects the lot on that characteristic**, regardless of the major plan's Ac=3/Re=4 result (which the lot would otherwise pass) or the minor plan's separate result. The lot is placed on immediate hold, engineering is notified for root-cause review, and — because a critical structural defect in one unit raises the question of whether it's isolated or systemic (a tooling or mold issue) — the remaining 3,075 units in the lot are flagged for 100% inspection of this specific defect characteristic before any further disposition, rather than relying on the sample result to represent the whole lot for this defect type.

**Deliverable (quality disposition / nonconformance report):**

> Lot #LW-3204, Widget Assembly. Sample n=125 per Z1.4 Level II. Major defects: 0 found (Ac=3/Re=4 — would PASS). Minor defects: 3 found, cosmetic scuffs (within AQL 2.5% acceptance — PASS). Critical defect: 1 found — hairline crack, load-bearing mounting point, classified CRITICAL per defined criteria. Critical plan Ac=0/Re=1 — **LOT REJECTED on critical characteristic**, independent of major/minor results. Disposition: LOT HOLD, engineering notified for root-cause review 2026-07-15. Remaining 3,075 units flagged for 100% inspection of this specific defect (crack at mounting point) pending root-cause determination (isolated vs. systemic/tooling-related). Authority: lot hold within inspector authority per QM-014; final release disposition requires engineering sign-off (beyond inspector authority).

## Going deeper

- [references/playbook.md](references/playbook.md) — a filled AQL sampling plan reference table, a critical-defect override decision table, and a gauge R&R interpretation guide.
- [references/red-flags.md](references/red-flags.md) — signals a sampling result, a measurement reading, or a disposition decision needs extra scrutiny before proceeding, and what to check first.
- [references/vocabulary.md](references/vocabulary.md) — terms of art generalists misuse (AQL, gauge R&R, producer's/consumer's risk, and others).

## Sources

ANSI/ASQ Z1.4 (Sampling Procedures and Tables for Inspection by Attributes), the modern successor to MIL-STD-105; AIAG Measurement Systems Analysis (MSA) reference manual for gauge R&R methodology; general knowledge of standard manufacturing quality inspection, sampling, and disposition practice.