PSLE Science Reality Lab Vol No.239 | “AQL = 1.0” — Does Every Accepted Lot Have at Most 1% Defects?

Stable internal ID: PSLE-SCI-REALITY-0239

Wait, what? A supplier’s inspection sheet says AQL = 1.0. A shipment is inspected, the lot is accepted, and someone announces: “Good. That proves no more than 1% of the items in this accepted lot are defective.”

The statement sounds tidy. There is a percentage-like number, a quality label and an ACCEPT decision. It is tempting to fuse all three into one simple conclusion. But an acceptance-sampling system does not work that way.

This Reality Lab is about a very specific scientific habit: do not confuse a design point in a sampling plan with a direct measurement of every item in a lot. That habit transfers straight into PSLE Science because the current 2026 assessment objectives include interpreting and analysing information, evaluating observations, information and methods, and communicating explanations and reasoning. The 2023 Primary Science syllabus also encourages healthy scepticism and careful attention to assumptions, uncertainty and evidence.

Quick Answer

No. In current ISO acceptance-sampling terminology, AQL means Acceptance Quality Limit. An AQL-indexed sampling scheme helps determine how many units to inspect and what acceptance or rejection rule to use across a continuing series of lots. An AQL value is not a guarantee that every accepted lot contains no more than exactly that percentage of defective units.

Acceptance sampling deliberately makes a decision about a whole lot from a sample rather than examining every unit. Because the decision is based on a sample, there is always a chance that a lot with more defects than the AQL can still be accepted, and a chance that a good lot can be rejected. The sampling plan controls those risks; it does not make them vanish.

AQL tells you how a sampling scheme is indexed. It does not tell you the exact defect percentage of one accepted lot.

The Owned Learner Job

This article owns one narrow real-world evidence-transfer job: how to evaluate a manufacturing, inspection or quality-control statement that cites an AQL without treating the AQL value as a directly measured maximum defect percentage for each accepted lot.

It does not replace the existing PSLE Science owners for sampling, representative evidence, probability, fair testing, measurement, variables or how far a conclusion can travel. It also does not teach an industrial quality-control course. The goal is to help a Primary 5/6 learner read one common communication object correctly.

First, a Terminology Repair

You may see AQL expanded online as Acceptable Quality Level. That wording appears in older materials and informal explanations. The current ISO 2859-1:2026 title and terminology use Acceptance Quality Limit. The important evidence habit is not the label itself. It is understanding that AQL is part of an acceptance-sampling scheme, not a direct census of defects in every accepted lot.

Rebuild the Evidence Object

Here is an original teaching example. It is deliberately invented and is not a table copied from ISO, a factory manual or a commercial inspection company.

Teaching-plan feature	Invented value
Lot size	5,000 plastic clips
Sample size	80 clips selected according to the inspection procedure
Acceptance rule	Accept if 2 or fewer sampled clips are nonconforming
Reject rule	Reject if 3 or more sampled clips are nonconforming
Label on the quality plan	AQL = 1.0

Suppose the sample contains one nonconforming clip. Under this invented rule, the lot is accepted.

What was actually observed? One defect in a sample of 80. What was not directly observed? The other 4,920 clips. Therefore the accepted decision cannot magically tell us the exact number of defects in the whole lot.

The inspection result is a decision under a sampling rule. It is not a complete count.

Four Quantities That Must Stay Separate

Quantity	What it means	What it does not automatically mean
AQL	A quality level used to index an acceptance-sampling scheme	The exact defect rate of this one lot
Sample defect fraction	Defects found divided by units inspected in the sample	The exact whole-lot defect fraction
Whole-lot defect fraction	The actual fraction of nonconforming units in the entire lot	Usually known exactly from a sample alone
Probability of acceptance	How likely a given sampling plan is to accept a lot at a particular quality level	A statement that acceptance proves the lot has that quality level

Many bad interpretations happen because two of these quantities share the same percentage sign. Science does not care that the symbols look similar. It cares whether the quantities answer the same question.

Observed, Claimed and Inferred

Using the invented plan above:

Observed: one nonconforming unit was found among 80 inspected units.
Rule applied: the teaching plan accepts lots when the sample has no more than two nonconforming units.
Decision: this lot is accepted under that plan.
Reasonable inference: the sample result was sufficiently good to meet this plan’s acceptance rule.
Unsupported upgrade: the lot is proven to contain at most 1% defects.
Even stronger unsupported upgrade: every item in the lot is good.

Why an Accepted Lot Can Still Contain More Than the AQL

Imagine a huge jar containing mostly white beads and some black beads. You cannot inspect every bead, so you draw a sample. Even if the true jar contains more black beads than you would prefer, a particular sample might happen to contain only a few. A different sample from the same jar might contain more.

That variability is not a flaw that sampling experts forgot to notice. It is the central reason sampling plans are designed with acceptance probabilities and risks. AQL-indexed schemes are built so that lots produced around an intended process-quality level are accepted with high probability under the specified system, while switching rules and other parts of the scheme respond when quality worsens. They do not make the sample identical to the whole lot.

Worked Case 1: Zero Defects in the Sample

An inspector checks 50 items and finds zero defects. The lot contains 10,000 items. A headline says, “Inspection Proves the Lot Is Defect-Free.”

What is wrong?

Zero defects were found in the inspected sample. That is evidence in favour of good lot quality, especially if the sampling method is sound. But 9,950 units were not individually observed. A defect could exist among them. The conclusion must not travel farther than the evidence permits.

This is why “no defect detected in the sample” and “no defect exists in the lot” are different scientific statements.

Worked Case 2: AQL = 1.0 and a 2% Lot

Suppose a lot really contains 2% defective units. Does AQL = 1.0 force the sampling plan to reject it every time?

No. A sample is a chance selection from the lot. Some samples will contain more defects; some fewer. A lot poorer than the AQL may still be accepted on a particular occasion. The probability should generally become less favourable as quality worsens, but acceptance is not a perfect detector of the hidden whole-lot percentage.

The correct question is not, “Can a poor lot ever pass?” It is, “What acceptance probability does this sampling scheme give at different quality levels, and what risks does the scheme control?”

Worked Case 3: Same AQL, Different Sampling Plans

Two factories both print “AQL 1.0” on a quality document. Factory A samples 20 items under one rule. Factory B samples 200 under a different rule. Can you conclude that the two plans give exactly the same protection?

No. The AQL label is not the whole sampling plan. Sample size, acceptance number, rejection number, inspection level, switching rules and the statistical design all matter. Current ISO 2859-1:2026 explicitly describes AQL-indexed single, double and multiple sampling schemes and switching between normal, tightened, reduced or skip-lot inspection in suitable continuing production contexts.

A label can be the front door to a method. It is not a substitute for the method.

Worked Case 4: One Accepted Lot Versus a Continuing Process

AQL systems are often designed for a continuing series of lots, not as a magic statement about one isolated shipment. If a supplier repeatedly produces poor lots, switching rules can tighten inspection or interrupt the sampling arrangement. This means the history of the process matters.

So an accepted lot is one decision inside a wider control system. It should not be advertised as proof that the process is perfect, the lot contains exactly the AQL defect percentage, or every item was tested.

The Representation Check

When an infographic, inspection certificate or product-comparison page shows “AQL 1.0,” inspect the representation before interpreting the number.

Does the label say AQL but omit the sampling plan?
Is the sample size shown?
Is the acceptance/rejection rule shown?
Was the sample selected according to an established procedure?
Is the document describing a continuing series of lots or one isolated lot?
Does “1.0” refer to nonconforming items, nonconformities per 100 units, or another defined basis?
Is the page reporting a plan setting, or reporting measured results from this particular sample?

One number without its denominator, decision rule and sampling context is an incomplete scientific communication object.

The Denominator Check

AQL numbers are easy to misread because people mentally attach “percent of this lot” even when that is not what has been measured. Always ask, “One percent of what population, under what definition, over what sampling system?”

Compare these three sentences:

“One of the 80 sampled items was nonconforming.”
“The sample nonconforming fraction was 1/80.”
“Exactly 1% of all 5,000 items in the lot are defective.”

The first two come from the sample. The third requires knowledge about the whole lot that the sample alone does not directly provide.

The Selection Check

A clever sampling rule cannot rescue a bad sample. Suppose the inspector always chooses boxes from the top of a pallet because they are easiest to reach, but defects are more common in boxes damaged near the bottom. The sample may follow the correct counting rule yet still fail to represent the lot fairly.

This is where the Reality Lab routes back to the broader PSLE Science sampling owners: the decision is only as meaningful as the evidence entering the decision.

Producer Risk and Consumer Risk

Acceptance sampling has two obvious kinds of mistake:

A good lot can be rejected. That is costly to the producer.
A poor lot can be accepted. That is risky to the consumer.

These possibilities are why a serious sampling plan is built around probabilities, not certainty. AQL is part of the design of that risk system. NIST guidance has long emphasised that AQL alone does not describe consumer protection for an individual lot; you need the operating characteristics of the actual plan.

For a Primary 5/6 learner, the useful transfer is simple: every sample-based decision has a chance of disagreeing with the hidden whole population. Good methods make those chances visible and controlled rather than pretending they do not exist.

What Evidence Strengthens an AQL-Based Claim?

The exact sampling standard and current edition are identified.
The sample size and acceptance/rejection criteria are stated.
The sampled units were selected according to the intended procedure.
The lot is clearly defined.
The definition of “nonconforming” is stated before inspection rather than invented afterward.
The inspection records show consistent application of the plan.
The process history and any switching rules are available where relevant.
Independent process-quality information supports the same general picture.

What Evidence Weakens the Claim?

“AQL 1.0” is shown without any sampling-plan details.
The sample was chosen conveniently rather than representatively.
The acceptance rule changed after the results were seen.
The product page implies every accepted lot has an exact measured defect rate equal to or below the AQL.
The lot identity is unclear or multiple production runs were mixed.
Inspection covers only one defect type but the advertisement generalises to “perfect quality.”
Rejected units are quietly replaced and the original result is not reported.

How Far Can the Conclusion Travel?

Suppose an AQL-indexed plan accepts a lot. A careful conclusion could be:

The inspected sample met the stated acceptance rule, so the lot was accepted under that sampling plan.

Do not automatically upgrade that to:

exactly 1% or fewer of the lot is defective;
every unit was inspected;
the lot is defect-free;
the process always produces at the AQL;
the product is safe for every use;
the quality system is scientifically perfect.

The first statement describes the evidence and decision. The later statements add claims that need additional evidence.

Tempting but Invalid Reasoning

“AQL 1.0 means exactly 1% defects.” No. It is a sampling-scheme index, not a census result.
“The lot passed, so its true defect percentage must be below the AQL.” Not guaranteed. Sampling decisions have probabilities of error.
“Zero sampled defects means zero lot defects.” Not unless every unit was inspected or other complete evidence establishes that.
“A larger sample is automatically unbiased.” A large convenience sample can still miss the wrong part of the lot.
“The same AQL means the same protection everywhere.” You need the full sampling plan and context.
“Acceptance sampling proves product safety.” A sampling decision only supports the defined inspection job. Safety claims require their own authoritative evidence.

PSLE-Style Transfer Case

A school has 2,000 identical batteries from one shipment. A pupil randomly tests 40 and finds that all 40 work. The pupil concludes, “All 2,000 batteries work because the sample passed.”

What should the pupil say instead?

A stronger answer is: The tested sample provides evidence that the shipment may have a high proportion of working batteries, but the untested batteries were not directly checked. The conclusion should be limited to what the sample supports, and the reliability of the inference depends on how the sample was selected and how many units were tested.

A Second Transfer: The Marble Bag

A bag contains 1,000 marbles. You draw 50 at random and all are blue. Does that prove every marble in the bag is blue? No. The evidence makes “mostly blue” more plausible, but one red marble could still be hidden among the 950 unobserved marbles.

Now imagine a formal rule: “Accept the bag as meeting the colour requirement if at most one red marble appears in the sample.” That rule creates a decision. It does not transform the sample into a complete count of the bag.

Practice Set

Practice 1

An inspection certificate says AQL 0.65 and “PASS.” What is the first question you should ask before interpreting 0.65?

Explained answer: Ask what sampling standard and plan were used, including the sample size and acceptance/rejection rule. The number alone is not the measured defect percentage of the lot.

Practice 2

A sample of 100 contains two defective units. May you say the whole 10,000-unit lot contains exactly 200 defective units?

Explained answer: No. Two per hundred is the observed sample fraction. The whole-lot count is unknown unless all units are examined. The sample can be used to make a statistical decision or estimate under a method, but it is not a complete census.

Practice 3

Two inspectors use the same AQL but different sample sizes and acceptance rules. Can you assume their chance of accepting a poor lot is identical?

Explained answer: No. The operating behaviour depends on the complete sampling plan, not the AQL label alone.

Practice 4

An inspector chooses the easiest-to-reach cartons and follows the counting rule perfectly. What evidence problem remains?

Explained answer: Selection bias. The sampling rule cannot compensate for a sample that systematically misses parts of the lot.

Delayed Independent Return

Tomorrow, without reopening this article, answer these two questions:

What did the inspector actually observe?
What did the inspection plan decide?

If you can keep observation and decision separate, you have learned the durable idea. A sample gives evidence. A rule turns evidence into a decision. Neither step magically reveals every unobserved unit.

Route to Existing Canonical PSLE Science Owners

Parent and Tutor Teaching Guide

Teach this with physical objects before introducing industrial terminology. Put 100 counters in an opaque bag, with an unknown number of red counters. Ask the learner to draw ten without looking. After observing the sample, ask for two separate sentences: one beginning “I observed…” and one beginning “I infer…”

Then add a simple acceptance rule such as “accept if no more than one red counter appears.” The most important question is: Did the rule change what we observed, or did it only change the decision we make from the observation?

Only after that distinction is stable should you show the learner an AQL label. This prevents the acronym from becoming a magic keyword. The scientific skill is not memorising what AQL stands for. It is understanding the evidence architecture behind the claim.

Authoritative Sources

Quiet Return

The next time a certificate says AQL = 1.0, resist the urge to read it as “this lot is proven to have at most 1% defects.”

Ask instead: What was sampled, what rule was used, what was actually observed, and what decision does this sampling plan support?

That is a much stronger scientific habit than trusting a neat number because it arrived beside the word PASS.

Small Group Tutorials