Full Inspection vs. Sampling Inspection Decision — AQL Is Not the Only Answer

Summary: Should incoming and process inspections be full inspections or sampling inspections? Should AQL be set at 0.65 or 1.0? Should inspection level I or II be used? Many companies rely on experience or a customer's verbal request to determine the inspection plan, lacking a risk-oriented, data-driven, and supplier performance-linked systematic approach. This article, within the framework of GB/T 2828.1 / ISO 2859-1, provides a decision tree for inspection planning, a complete table example, normal/tightened/reduced transfer rules, and three industry-specific calculation cases to help quality engineers allocate inspection resources more effectively.

1. A Real Dilemma: Full Inspection Does Not Guarantee Delivery

A certain Tier 2 fastener factory has 6 IQC inspectors and over 800 types of incoming materials. The quality director set the tone for "full inspection of critical parts, sampling inspection of general parts," but the list of critical parts kept growing—every time R&D added a special characteristic, it was included in the list, which increased from 80 items to 230 items in six months. The inspection station had a queue of 4 hours, and production lines stopped due to material shortages; procurement complained about "quality department holding up goods."

The finance department did a calculation: for a certain M8 bolt, full inspection of thread pass/fail and appearance for a single batch of 30,000 pieces required 18 person-hours; however, the same supplier had a continuous 24 batches with PPM=120 and a process Cpk=1.6—the marginal benefit of full inspection was extremely low.

On the other hand, another plastic parts factory used AQL 4.0 sampling for "low-value auxiliary materials," resulting in a color difference in a certain batch of color masterbatches, leading to the rework of 6,000 pieces—the sampling plan did not match the defect consequences.

The root cause of the problems in both factories is the same: inspection planning has not answered four questions—(1) How severe are the failure consequences? (2) How reliable is the supplier and process? (3) Where is the economic threshold for full inspection vs. sampling inspection? (4) How can the plan be dynamically adjusted based on performance?

11.1.1 has defined the logic for the four inspection stations: IQC, IPQC, FQC, and OQC; 11.1.2 focuses on how to scientifically determine the sampling plan itself.

2. The Overall Framework for Inspection Planning: Grade First, Then Determine the Plan

2.1 Material and Characteristic Risk Grading

It is recommended to use at least a two-dimensional matrix:

Example Rules:

• A class + Supplier level II → tightened sampling or 100% automatic inspection of critical dimensions
• A class + Supplier level I for 6 consecutive months → can evaluate skipping batches or transferring to source inspection
• C class + Supplier level I → normal level II sampling is sufficient

2.2 Full Inspection vs. Sampling Inspection: Decision Tree (Text Version)

Start
 ├─ Is the defect safety/critical/regulatory-related? → Yes → 100% inspection or 100% automatic inspection + traceability
 ├─ No → Is the supplier newly certified or on probation? → Yes → tightened plan or temporary full inspection
 ├─ No → Is the process Cpk≥1.33 and SPC in control? → No → tighten/100% inspect until capability is proven
 ├─ Yes → Can inspection be automated at low cost? → Yes → 100% online inspection (not manual full inspection)
 └─ No → Sample according to AQL + transfer rules

Key Point: Full inspection is not "more responsible"; manual full inspection also has missed inspections, fatigue, and consistency issues—when 100% automatic inspection is possible, it is superior to manual sampling.

3. AQL, Inspection Levels, and Sampling Types — What the Standard Says

3.1 AQL Is Not "Permissible Defect Rate Target"

AQL (Acceptable Quality Limit) in GB/T 2828.1 is a sampling plan index parameter used to look up the sample size n and acceptance number Ac, rejection number Re—it does not mean telling the supplier "you can have 1% defects."

For supplier contracts, use PPM targets, Cpk requirements, zero-defect characteristics, etc. AQL is only used for sampling plan design.

3.2 Inspection Levels I / II / III

3.3 Normal, Tightened, and Reduced Inspection

The standard provides transfer rules (which must be written into the inspection work instruction):

Normal → Tightened (example conditions):

• Two consecutive batches rejected under tightened conditions; or
• Two out of five batches rejected under normal conditions (specific conditions depend on the adopted standard clauses)

Tightened → Suspension/Full Inspection: Five consecutive batches rejected under tightened conditions → supplier improvement, 100% inspection until five consecutive batches are accepted

Normal → Reduced: Ten consecutive batches accepted under normal inspection and production is stable → can be reduced (must be approved)

Common Mistakes in Companies: Always staying at "normal level II" and never tightening—resulting in poor suppliers using the same lenient plan.

4. Calculation Case 1: IQC Count Sampling (Normal Inspection)

Conditions:

• Material: plastic housing, batch size N = 2000 pieces
• Inspection: surface scratches (count type)
• Plan: GB/T 2828.1, general inspection level II, AQL = 1.0

Table Lookup Results (illustrative, refer to the current effective standard table):

• Sample size code → sample n = 125
• Ac = 3, Re = 4 (i.e., ≤3 defects in 125 pieces accepted, ≥4 defects rejected)

On-Site Execution:

1. Randomly sample 125 pieces (must define sampling method: random, multi-stack, multi-box)
2. Record the number of defects d
3. d ≤ 3 → batch accepted; d ≥ 4 → batch rejected, initiate CAR (9.2.3)

If Rejected: issue an 8D to the supplier; for this batch: isolate, 100% re-inspect, or return—the handling after sampling rejection must be defined in the procedure, not "reject and then concede."

5. Calculation Case 2: Triggering Tightened Inspection

Continuing from the previous example, the supplier's continuous performance:

According to the transfer rules → enter tightened inspection (larger n, smaller Ac under the same AQL). At this point, procurement and quality jointly meet with the supplier, and IQC freezes the reduction of this material until improvement is verified.

Management Significance: The AQL plan is a dynamic agreement, not a static table.

6. Strategies Beyond Standard AQL

6.1 C=0 Sampling Plan (Zero Acceptance)

High-consequence characteristics (safety parts, critical fit dimensions) often use C=0 (Ac=0)—the sample size is usually larger than the AQL-based Ac≥1 plan, but there is no gray area after rejection.

Applicable to: regulatory parts, customer CSR designated, internal Class A characteristics.

6.2 Skip-Lot

Prerequisite: Supplier level I, 12 months PPM target met, no major non-conformities in second-party audits.

Example Rules:

• Inspect 1 out of every 5 batches (Skip-Lot 1/5)
• Any batch rejected → immediately resume batch-by-batch inspection and cancel skip-lot eligibility for 3 months

Must have customer written approval or contractual authorization.

6.3 Source Inspection Transfer

Supplier's outgoing inspection data is credible (laboratory accreditation, SPC online, regular second-party audits) → the OEM reduces the IQC sample size and strengthens:

• Data review (Cpk per batch, inspection records)
• Surprise inspections
• Periodic destructive verification

Linked with 9.1.2 second-party audits and 6.3 SPC.

6.4 100% Automatic Inspection

Go/no-go gauges, visual inspection, weight check—when the cycle allows and the equipment GR&R is qualified, 100% automatic inspection is often more cost-effective and reliable than manual sampling (see 12.2.3).

7. What Should the Inspection Planning Document Include

One document per material family, suggested sections:

1. Scope and Referenced Standards (GB/T 2828.1 edition, CSR)
2. Risk Grading (A/B/C judgment)
3. Inspection Items and Characteristic Types (count/measurement)
4. Sampling Plan (full inspection/sampling, AQL, level, Ac/Re or C=0)
5. Transfer Rules (triggers for normal/tightened/reduced)
6. Handling After Rejection (isolation, re-inspection, return, concession must follow escalation path 3.4.3)
7. MSA Requirements (6.2.1, measurement characteristics)
8. Record and Traceability Fields
9. Correspondence with CP/PFMEA Detection Controls

Audit Focus: Is the plan being followed on-site? Are rejected batches closed-looped? Is tightening triggered?

8. Interface with PFMEA and Control Plan

PFMEA detection controls (D) should be reflected in inspection planning:

The "sample/frequency" column in the CP must be consistent with the IQC document—the most common non-conformity in audits: CP states full inspection, IQC performs sampling.

9. Common Misconceptions and Countermeasures

10. Conclusion

Full inspection vs. sampling inspection is not a matter of stance but a function of risk, cost, capability, and supplier performance.

AQL is the language of sampling plans, not a substitute for quality targets. By forming a closed loop with grading, table lookup, transfer, and data feedback, inspection can transform from a "bottleneck department" into the early warning system for supply chain risks.

Next steps recommended: select the top 20 incoming materials, write an inspection planning document for each, and use 3 months of data to verify whether tightening/reduction is triggered according to the rules—this is more valuable than another debate on "full inspection or sampling."

Knowledge Number: 11.1.2

Version: v20260528

Author: Quality Excellence Think Tank