AQL Inspection Guide — How We Use Acceptable Quality Limits to Make Sampling Decisions That Protect the Customer and the Production Line

Quality cannot be inspected into a product. But inspection can catch what process control missed — if the inspection is designed correctly.

AQL (Acceptable Quality Limit, formerly Acceptable Quality Level) is the cornerstone of statistical inspection for manufactured goods. It defines the maximum percentage of defective items that can be considered acceptable in a production batch — and it governs the sample size used to make the accept/reject decision.

When used correctly, AQL inspection is a powerful tool: it provides a statistically defensible basis for accepting or rejecting a production lot, it gives the production team clear quality targets, and it protects the customer from receiving batches with hidden defect concentrations. When used incorrectly — with the wrong AQL level, the wrong inspection severity, or the wrong criticality classification — AQL inspection can create a false sense of security or, conversely, trigger unnecessary rejections that disrupt the supply chain.

This article is written from the perspective of an LED therapy device OEM manufacturer — sharing how we define AQL levels for different defect categories, how we select the correct sampling plan, how we classify defects, and how we use AQL data to drive process improvement rather than just to pass or reject lots.

The Three Defect Categories — And Why the Distinction Matters

Before selecting an AQL level, the defect classification must be defined. In manufacturing, defects are classified into three categories based on their impact on the user:

Critical defects: Defects that can cause harm to the user or that violate mandatory safety or regulatory requirements. For LED therapy devices, critical defects include: battery cells that fail the overcharge protection test (fire hazard), silicone materials that show signs of skin irritation in material testing, missing or defective insulation on high-voltage components, and devices that fail hipot/ground continuity testing. A lot containing critical defects is not acceptable under any AQL level — even a single critical defect should trigger lot rejection.

Major defects: Defects that do not pose a safety hazard but that will prevent the product from functioning as intended or that are clearly unacceptable to the end user. For LED therapy devices, major defects include: LED output below the specified therapeutic irradiance threshold, a non-functional touch controller, LED wavelength outside the specified range (e.g., a red channel that reads as orange-red due to wavelength shift), a missing or defective silicone skin-contact pad, and cosmetic defects on visible surfaces that would be immediately noticeable to a consumer.

Minor defects: Defects that do not affect the product’s safety, functionality, or aesthetic appeal in any meaningful way — but that represent a deviation from the specification. For LED therapy devices, minor defects include: a minor cosmetic scratch on the rear housing that is not visible when the device is in use, a slight color variation in the silicone that falls within the approved color range but is slightly different from the reference sample, a small excess of adhesive visible at a seam that is not on a skin-contact surface, and a label that is slightly misaligned but legible and correct.

Why classification drives AQL: Each defect category has a different AQL level. Critical defects have an AQL of 0% — a lot with any critical defect is rejected. Major defects use a tighter AQL (typically 0.10% or 0.15%). Minor defects use a more permissive AQL (typically 0.65% or 1.0%). Using the same AQL for all defect categories — or inverting the relationship (tighter AQL for minor than for major) — is a misuse of the AQL system.

The AQL Sampling Table — How to Read ISO 2859-1

AQL inspection is governed by the ISO 2859-1 standard (Sampling Procedures for Inspection by Attributes). The standard provides tables that define the sample size and accept/reject numbers for any combination of batch size, inspection level, and AQL level.

The two key tables:

Table 1 — Sample size code letters: Defines the letter code (A, B, C, D, E, F, G, H, J, K…) that corresponds to the batch size and inspection level. For most beauty device production batches (51–500 units), the code letter is typically F, G, or H. For larger batches (501–3,200 units), the code letter is J or K.

Table 2 — Single sampling plans: Defines the sample size (n), the acceptance number (Ac), and the rejection number (Re) for each combination of code letter and AQL level. For example, for code letter G and AQL 0.10%: sample size = 32, Ac = 0, Re = 1. This means: randomly select 32 units from the lot, inspect them, and reject the lot if you find 1 or more defective units.

Code letter F: Batch size 151–500 units, sample size 20, Ac = 0, Re = 1 at AQL 0.10% (major), Ac = 1, Re = 2 at AQL 1.0% (minor) Code letter G: Batch size 501–1,200 units, sample size 32, Ac = 0, Re = 1 at AQL 0.10% (major), Ac = 1, Re = 2 at AQL 1.0% (minor) Code letter H: Batch size 1,201–3,200 units, sample size 50, Ac = 0, Re = 1 at AQL 0.10% (major), Ac = 1, Re = 2 at AQL 1.0% (minor) Code letter J: Batch size 3,201–10,000 units, sample size 80, Ac = 0, Re = 1 at AQL 0.10% (major), Ac = 2, Re = 3 at AQL 1.0% (minor)

The inspection level: The inspection level (I, II, or III) determines the relationship between the batch size and the sample size code letter. Level II is standard. Level I is reduced inspection (used for suppliers with a strong quality history). Level III is tightened inspection (used for new suppliers or suppliers with a history of marginal quality).

AQL Levels We Use at RainbowDO — And Why

Defect Category	AQL Level	Basis
Critical	0% (zero tolerance)	Any single critical defect triggers lot rejection
Major	0.10%	Standard for medical/health devices; statistically 1 defective in 1,000
Minor	0.65%	More permissive for cosmetic/minor deviations

Why 0.10% for major defects: For a consumer health device like an LED therapy mask, a 0.10% AQL for major defects means accepting lots where the defect rate is at or below 1 in 1,000. At the standard sample size of 32 units (code letter G), this translates to Ac = 0, Re = 1 — meaning: reject if we find even 1 major defect in the sample. This is an exacting standard that requires a defect rate of well below 1% in the production lot to reliably pass.

The producer’s risk and the consumer’s risk: No sampling plan eliminates all risk. There are two types of error:

Producer’s risk (α): The probability of rejecting a lot that is actually at or below the AQL. This is typically set at 5%. In plain terms: a good lot has a 5% chance of being rejected.
Consumer’s risk (β): The probability of accepting a lot that has a defect rate worse than the AQL. At AQL 0.10% and code letter G, the Lot Tolerance Percent Defective (LTPD) is approximately 6–7% — meaning a lot with a 6–7% defect rate has a 10% chance of being accepted. This is the limitation of AQL sampling: it cannot catch low-level concentrations of defects reliably unless the defect rate is extremely high.

The implication: AQL inspection is not a substitute for process quality. A lot with a 1% defect rate has a high probability of passing an AQL 0.10% inspection (sample size 32) — because the sample of 32 is too small to reliably detect 1 defective item among 32. AQL inspection catches concentrated defect clusters (e.g., an entire shift producing out-of-spec units), not low-level uniform defect rates.

The Inspection Process at RainbowDO — From Sampling to Lot Decision

Step 1: Define the inspection lot

An inspection lot is a group of units produced under consistent conditions: the same production line, the same shift, the same material batch, and the same production date. A lot should not mix units from different shifts or different material batches — mixing obscures the root cause if the lot fails.

Step 2: Determine the sample size

Using ISO 2859-1 Table 1, identify the sample size code letter based on the batch size and the inspection level (Level II is standard). Then use Table 2 to find the acceptance and rejection numbers for each AQL level.

Step 3: Random sampling

Samples must be selected randomly from the lot — not hand-picked to include good or bad units. Random sampling is essential for the statistical validity of the AQL decision.

Step 4: Inspect each unit

Inspect each sampled unit against the complete inspection checklist — which includes critical, major, and minor defects separately. Record the number of defects found in each category.

Step 5: Make the accept/reject decision

Compare the number of defects found to the acceptance criteria:

Critical defects: Any critical defect → reject the lot immediately
Major defects: Ac = 0 at AQL 0.10% → reject if ≥ 1 major defect found
Minor defects: Ac = 1 at AQL 0.65% → reject if ≥ 2 minor defects found

Step 6: Document and disposition

Record the inspection results, the lot number, the production date, the inspector, and the accept/reject decision. If the lot is rejected, initiate the non-conforming material process (sorting, rework, or scrap as appropriate).

Step 7: Use the data

AQL inspection data is only as valuable as the action taken on it. Track defect types, defect rates by production line and shift, and AQL pass rates over time. Rising defect rates trigger process review before they trigger increased lot rejections.

Switching Rules — Tightened and Reduced Inspection

ISO 2859-1 includes switching rules that allow the inspection severity to be adjusted based on the supplier’s quality history:

Normal → Reduced inspection: When 10 consecutive lots have been accepted under normal inspection (at the specified AQL), the inspection level can be switched to reduced inspection (smaller sample size). This rewards consistent quality with lower inspection cost.

Normal → Tightened inspection: When 2 out of 5 consecutive lots have been rejected under normal inspection, the inspection level must be switched to tightened inspection (larger sample size, tighter acceptance criteria). This is a mandatory rule — the decision is not discretionary.

Tightened → Normal: When 5 consecutive lots have been accepted under tightened inspection, the inspection level returns to normal.

Reduced → Normal: If a lot is rejected under reduced inspection, the inspection level returns to normal.

These switching rules are designed to maintain the producer’s incentive for consistent quality — and to catch deteriorating quality before it becomes a serious problem.

Common AQL Mistakes — And How to Avoid Them

Mistake 1: Using the same AQL for all defect categories Using AQL 1.0% for critical, major, and minor defects alike is a common error. Critical defects must have zero tolerance (AQL = 0%). Using a single AQL level across categories creates a safety risk.

Mistake 2: Confusing AQL with the defect rate AQL is not the defect rate — it is the maximum defect rate that a good lot is allowed to have. A lot with a 0.10% defect rate should pass AQL 0.10% inspection most of the time. A lot with a 0.5% defect rate will fail AQL 0.10% most of the time — but it will also likely pass a single sample of 32 units, because the probability of finding a defective unit in a sample of 32 from a 0.5% defective lot is only about 15%.

Mistake 3: Treating AQL inspection as the only quality control AQL inspection is a statistical tool for lot acceptance — it is not a substitute for process control, in-process inspection, or statistical process control (SPC). A production line that relies on AQL inspection to catch defects is accepting that some defective units will reach the customer before the defect is detected.

Mistake 4: Using AQL 10% as a substitute for “no inspection” Some teams use AQL 10% because they want to accept almost everything. AQL 10% means accepting lots with up to 10% defective units — which is not quality control, it is quality acceptance. AQL 10% is appropriate only for certain non-critical, non-functional cosmetic defects — and even then, only when the defect category and AQL level are explicitly agreed upon with the customer.

AQL in Practice — Common Questions

Q1: We inspected a lot with 500 units, found 0 major defects in the sample of 32, and accepted it. But the customer returned 5 units with major defects from that lot. What went wrong?

The AQL plan does not guarantee that the lot is defect-free — it only tests whether the defect rate is below the AQL at a defined statistical confidence level. With a sample size of 32, the probability of detecting a defective unit in the sample is approximately 3% per defective unit in the lot. A lot with a 0.5% defect rate (2.5 defective units in 500) has only about a 7% chance of being detected in a sample of 32. The solution is not to blame the AQL plan — it is to improve the production process so that the defect rate is consistently below the AQL, and to implement in-process inspection and SPC to catch defects before they reach final inspection.

Q2: Our supplier says their AQL is 0.10% for major defects. But they keep sending lots with visible defects. How is this possible?

There are two possibilities: (1) The supplier’s internal AQL uses a different sampling plan — possibly a smaller sample size or a different acceptance number (Ac = 1 instead of Ac = 0). Verify the exact sampling plan by reviewing ISO 2859-1 Table 2. (2) The defects are being classified differently — what you classify as a major defect, the supplier may be classifying as minor. Establish a joint defect classification standard before production begins, and reference physical defect samples (golden samples) to eliminate ambiguity.

Q3: When should we use tightened inspection instead of normal inspection?

Switch to tightened inspection when: (1) Two out of five consecutive lots have been rejected at normal inspection — this is a mandatory ISO 2859-1 requirement; (2) You receive a new or modified product from a supplier with no quality history; (3) The supplier has changed materials, components, or manufacturing process; (4) Your incoming quality data shows a trend of increasing defect rates even if individual lots are still passing AQL. Tightened inspection increases the sample size (improving detection probability) and lowers the acceptance number — making it harder for marginal lots to pass.

This article is written from the perspective of an LED therapy OEM manufacturer that uses AQL inspection as one component of a comprehensive quality system — including in-process inspection, statistical process control, supplier qualification, and continuous improvement. The AQL sampling system and switching rules referenced are defined in ISO 2859-1 (Sampling Procedures for Inspection by Attributes). Specific AQL levels, defect classifications, and inspection protocols should be defined in the product’s Quality Inspection Plan, tailored to the device’s safety classification and customer requirements.