How AQL Inspection Works for Bulk Plush Orders

A beautiful approval sample does not guarantee that several thousand finished plush toys will all look, feel and perform the same. During volume production, small variations can accumulate: one sewing line may create wider heads, one filling station may produce softer bodies, a later embroidery batch may shift an eye by several millimeters, or one packing shift may apply the wrong barcode. The risk is not that every unit will be wrong. The risk is that nobody knows how much variation exists until the shipment is already packed.

AQL inspection works by selecting a statistically defined number of units from a finished production lot, checking them against approved specifications, classifying nonconformities as critical, major or minor, and comparing the findings with predetermined acceptance and rejection numbers. For custom plush, the approved Golden Sample, final dimensions, materials, workmanship, accessories, labels, barcodes and packing specifications should form the physical and documentary inspection basis.

That distinction matters. AQL is not simply “looking at some pieces before shipment.” It creates a repeatable release decision.

Imagine opening twelve cartons from a 5,000-piece character plush run. Eleven look excellent. One has an eye attachment that separates under moderate force. Is that an isolated assembly error, or evidence of a weakness repeated across the lot? A mature inspection system is designed to uncover exactly that kind of uncertainty before the goods leave the production site.

What AQL Standard Should Be Used for Bulk Plush Orders?

For most finished plush orders, ISO 2859-1 or ANSI/ASQ Z1.4 attribute sampling with General Inspection Level II is a practical basis for final random inspection. A common commercial starting setup is zero acceptance for critical defects, AQL 2.5 for major defects and AQL 4.0 for minor defects. More sensitive character, baby, magnetic, weighted or electronic plush should normally use tighter limits or additional full-unit controls.

The sampling standard and the product quality requirement are not the same thing.

ISO 2859-1 or ANSI/ASQ Z1.4 determines how the lot is sampled and how acceptance sampling is structured. The plush specification determines what should be considered critical, major or minor and which characteristics cannot rely on random sampling alone.

As of 2026, ISO 2859-1:2026 is the current international edition for AQL-indexed lot-by-lot inspection by attributes. It replaced the 1999 edition and includes normal, tightened, reduced and skip-lot concepts. In the United States, ASQ identifies ANSI/ASQ Z1.4-2003 (R2018) as the current edition of its attribute-sampling standard, also using normal, tightened and reduced plans.

For plush production, a defensible inspection plan therefore needs four decisions before finished goods are sampled:

• Which sampling standard will be used?
• Which inspection level applies?
• Which AQL applies to each defect class?
• Which safety or functional characteristics require controls beyond AQL?

If those decisions are left until an inspector is already opening cartons, disagreements are almost inevitable.

ISO 2859-1 vs ANSI/ASQ Z1.4

Both systems are widely used for acceptance sampling by attributes.

ISO 2859-1:2026 is the international standard for sampling schemes indexed by Acceptance Quality Limit for lot-by-lot inspection. The 2026 edition provides single, double and multiple sampling approaches and switching rules that allow inspection severity to respond to historical performance.

ANSI/ASQ Z1.4 performs a closely related role in the United States. ASQ describes it as an acceptance-sampling system for percent nonconforming or nonconformities per 100 units, with normal, tightened and reduced plans.

For a custom plush program, the practical issue is less about proving that one system is universally “better” and more about ensuring that everyone uses the same named standard, edition, inspection level and defect limits.

A purchase specification stating only:

“AQL inspection required”

is incomplete.

A clearer instruction would be closer to:

ISO 2859-1, single sampling, normal inspection, General Level II, Critical Ac 0/Re 1, Major 2.5, Minor 4.0.

That gives the inspection party something operational rather than interpretive.

What AQL Level Is Standard for Plush Toys?

There is no single mandatory AQL value for every plush product.

For many standard retail plush programs, the following setup is a reasonable starting position:

Defect Class	Common Setting	Examples
Critical	Ac 0 / Re 1	needle fragment, dangerous magnet, leaking weighted material
Major	AQL 2.5	open seam, severe facial distortion, missing component
Minor	AQL 4.0	removable thread, slight pile disorder
Inspection Level	General II	normal final random inspection

More visually sensitive or higher-risk plush can use tighter limits.

For example:

Plush Type	Critical	Major	Minor	Possible Inspection Level
Standard sewn plush	Ac 0/Re 1	2.5	4.0	II
Retail character plush	Ac 0/Re 1	2.5	4.0	II
Premium licensed character	Ac 0/Re 1	1.5	2.5	II or III
Baby plush	Ac 0/Re 1	1.0–1.5	2.5	III
Magnetic plush	Ac 0/Re 1	1.0–1.5	2.5	III + full magnet checks
Electronic plush	Ac 0/Re 1	1.0–1.5	2.5	III + full function checks

These are quality-planning examples, not universal legal values.

Heyzizi’s documented outgoing criteria follow the same risk-based logic: regular commercial and retail programs use critical 0, major 2.5 and minor 4.0, while premium IP and higher-safety products use tighter major limits.

The most important decision is not whether 2.5/4.0 is familiar. It is whether those limits are appropriate for the actual plush.

Why Is General Inspection Level II Common?

General Inspection Level II is widely used because it provides a middle level of sampling coverage.

The three general inspection levels can be understood as:

Level	Relative Sample Size	Best Fit
General I	smaller	lower-risk or established stable production
General II	normal	routine finished-goods inspection
General III	larger	first runs, sensitive or higher-risk programs

Inspection level and AQL perform different jobs.

Inspection level determines how many units are examined.

AQL determines the acceptance limit applied to the sample.

Changing from Level II to Level III does not automatically change Major AQL 2.5 into Major AQL 1.5. Instead, it normally increases the amount of evidence collected from the lot.

That is useful when a new plush has:

• several sewing operations;
• complicated clothing;
• multiple embroidered areas;
• new filling specifications;
• magnets;
• electronics;
• weighted inserts;
• many SKUs;
• a history of inconsistent production.

Heyzizi’s documented QC procedure uses General Level II as its normal outgoing basis but allows first runs and baby, electronic, magnetic or weighted items to move to General Level III or tighter inspection.

How Does Inspection Level Change Sample Size?

The lot quantity does not directly become the inspection quantity.

The sampling table first converts the lot size and inspection level into a sample-size code letter. That code then determines how many finished units are inspected.

Under General Inspection Level II, frequently encountered examples are:

Finished Lot	Code Letter	Sample Quantity
281–500	H	50
501–1,200	J	80
1,201–3,200	K	125
3,201–10,000	L	200
10,001–35,000	M	315
35,001–150,000	N	500

This stepped structure explains why a 20,000-piece order does not require inspecting twenty times as many units as a 1,000-piece order.

For example:

1,000-piece lot → 80 sampled units

5,000-piece lot → 200 sampled units

20,000-piece lot → 315 sampled units

The sampling system is designed to provide statistical lot evidence without requiring every piece to undergo the same final random inspection.

However, a correct numerical sample can still be poorly selected.

If 200 inspected pieces from a 5,000-piece lot all come from the same ten cartons, the quantity may be correct while the randomization is weak.

The sample should represent the physical shipment.

How Should Random Samples Be Selected?

The inspector should not simply inspect the most convenient cartons.

For a large plush order, selection should be spread across:

• different carton locations;
• different pallet positions;
• early and later packing;
• different sewing lines where identifiable;
• different embroidery lots where relevant;
• different colors;
• different SKUs.

This matters because plush variation often appears in clusters.

One filling station may produce softer bodies for half a shift. One embroidery setup may create slightly wider eyes for 600 pieces. One packing team may apply the wrong hangtag to several cartons.

If the random sample misses the affected production section, the AQL result becomes less informative.

The lot definition therefore matters almost as much as the sample quantity.

A “lot” should represent merchandise produced under reasonably consistent conditions. If two portions of an order use different materials, different structures or substantially different processes, treating them as one homogeneous lot may hide useful information.

What Are Ac and Re Numbers?

After the sample size is determined, the selected AQL gives an Ac and Re number.

Ac = acceptance number

Re = rejection number

For commonly used single-sampling tables, an example at Major AQL 2.5 looks like this:

Sample Quantity	Major AQL 2.5	Minor AQL 4.0
50	Ac 3 / Re 4	Ac 5 / Re 6
80	Ac 5 / Re 6	Ac 7 / Re 8
125	Ac 7 / Re 8	Ac 10 / Re 11
200	Ac 10 / Re 11	Ac 14 / Re 15

Take a 5,000-piece order under General Level II.

The inspection quantity is commonly 200 units.

If Major AQL 2.5 applies:

0–10 major-defective units → within the major acceptance criterion

11 major-defective units → rejection threshold reached

Minor defects are evaluated independently against the minor limit.

Critical defects should also be treated separately.

Is “Critical AQL 0” Technically Correct?

“Critical AQL 0” is commonly used in commercial communication, but Ac 0 / Re 1 is more precise when the intended requirement is zero acceptance of a discovered critical defect.

The reason is simple: AQL should not be misunderstood as a direct percentage allowance.

A critical rule is usually intended to mean:

If one critical defect is found in the defined inspection, normal shipment release stops.

That is different from saying:

“0% critical defects have been statistically proven throughout every piece.”

Sampling cannot make that guarantee.

For plush, critical examples may include:

• needle fragments;
• sharp metal;
• dangerous loose magnets;
• serious small-part hazards;
• leaking weighted media;
• unsafe battery access.

Heyzizi’s documented quality policy applies zero acceptance to critical conditions and does not allow one discovered critical defect to proceed through ordinary shipment release.

Does AQL 2.5 Mean 2.5% Defective Plush?

No.

This is one of the most common misunderstandings in acceptance sampling.

AQL 2.5 is an index used with the defined sampling plan. It does not mean that exactly 2.5% of the full shipment has been inspected and accepted as defective.

Suppose 200 units are inspected under Major AQL 2.5.

The applicable plan may allow Ac 10 and reject at Re 11.

Ten defective units out of 200 equals 5% of that particular sample, yet the lot can still remain within the Major 2.5 acceptance criterion.

That alone proves why the AQL number should not be interpreted as a simple permitted defect percentage.

ISO describes AQL as an index used to determine sampling and acceptance/rejection thresholds for continuing lot inspection, while ASQ describes acceptance sampling as a system using defined sample sizes and acceptance criteria.

Commercial documents should therefore state the sampling plan rather than simply writing:

“Maximum defects allowed: 2.5%.”

How Should Multi-SKU Plush Orders Be Sampled?

Multi-SKU orders need more planning than single-design lots.

Consider 8,000 pieces containing:

• Character A: 2,000
• Character B: 2,000
• Character C: 1,500
• Character D: 1,000
• Character E: 500
• Character F: 500
• Character G: 300
• Character H: 200

A single 200-piece random sample from the entire shipment might include only a handful of Character H.

That may be statistically acceptable for the combined lot but weak for evaluating H’s unique embroidery, clothing or accessory.

Useful approaches include:

• define separate lots by SKU;
• stratify the total sample across designs;
• require a minimum inspection quantity per SKU;
• separately test unique high-risk structures.

Heyzizi’s controlled procedure supplements general AQL sampling by requiring ordinary dimensions and weight to be checked on at least five pieces per SKU in the applicable outgoing inspection.

The principle is more important than one fixed number: a low-volume design should not become almost invisible inside a large collection.

When Should AQL Be Tightened?

Inspection should become stricter when recent evidence gives less confidence in production stability.

Reasons may include:

• first production run;
• earlier failed lot;
• repeated major defects;
• new fabric;
• new sewing line;
• new embroidery setup;
• new subcontracted process;
• major structural revision;
• electronics, magnets or weighted components.

A tighter response can involve:

• moving from Level II to Level III;
• reducing the major AQL;
• increasing SKU-specific checks;
• adding 100% screening for the affected characteristic;
• moving from normal to tightened sampling where the adopted standard requires it.

ISO 2859-1:2026 and ANSI/ASQ Z1.4 both include switching concepts between normal, tightened and reduced inspection instead of assuming that every production lot deserves identical treatment forever.

This is one reason inspection history should be retained rather than discarded after shipment.

When Can Inspection Be Reduced?

Reduced inspection should come from demonstrated stability, not from schedule pressure.

Repeated accepted lots, stable materials, controlled processes and consistent defect history may support a less intensive sampling status where the adopted standard permits it.

Heyzizi’s documented internal policy similarly allows long-running stable repeats to be reviewed for General Level I or continued normal Level II, while first runs and higher-risk products move in the opposite direction.

Reduced inspection should never mean removing controls that are still required for every piece.

For example, a stable repeat run may justify lighter random outgoing sampling while still maintaining:

• 100% needle detection;
• 100% electronic function checks;
• 100% magnet checks where applicable;
• 100% weighted-product mass checks where specified;
• critical barcode verification.

Statistical history can reduce sampling effort. It does not eliminate product-specific risk.

Can AQL Replace Safety or Functional Controls?

No.

AQL is strongest when evaluating lot-level workmanship and conformity. It should not automatically replace checks where every unit needs individual confirmation.

Heyzizi’s recorded process separates outgoing AQL sampling from 100% checks for needle detection, electronic function, weighted-product mass, magnet position/function and selected barcode controls.

That creates a more complete structure:

AQL random sampling

for appearance, dimensions, workmanship and packing consistency.

Selected physical testing

for seams, attachments and pull strength.

100% checks where appropriate

for characteristics where individual failure carries disproportionate risk.

The strongest AQL specification for a bulk plush order is therefore not simply:

Major 2.5 / Minor 4.0.

It is:

named sampling standard + inspection level + lot definition + sample quantity + critical rule + major/minor limits + defect definitions + Golden Sample + SKU coverage + product-specific full checks.

When those elements are agreed before volume production, the final inspection becomes a repeatable lot decision rather than a negotiation over finished cartons.

How Are Defects Classified During Plush AQL Inspection?

Plush defects should be classified by the consequence they create, not simply by how large or visible they look. Critical defects involve unacceptable safety risk. Major defects materially affect function, durability, character appearance, labeling, SKU accuracy or commercial usability. Minor defects reduce finish quality without materially changing safety or normal use. The classification should be defined before final inspection and tied to the approved Golden Sample, technical specifications and product risk.

The difficult part is rarely recognizing that something is wrong. The difficult part is deciding how serious that condition is.

A 2 mm eye shift may be barely noticeable on a generic animal but unacceptable on a licensed character whose expression depends on precise eye spacing. A small needle fragment may be almost invisible yet be far more serious than a large surface stain.

For that reason, plush defect classification should follow several dimensions at the same time:

Evaluation Area	Main Concern	Likely Severity
Safety	injury, choking, sharp metal, magnet or filling leakage	Critical
Function	sound, light, clasp, zipper or intended feature does not work	Major
Durability	seam or attachment likely to fail during normal use	Major
Character accuracy	face or silhouette no longer matches approved appearance	Major or Minor
Identification	incorrect SKU, label, barcode or warning	Major or Critical
Finish	small cosmetic issue without functional impact	Minor

AQL becomes meaningful only after those classifications are clear.

What Is the Difference Between a Defect and a Defective Unit?

A defect is one specific nonconformity.

A defective unit is one inspected plush containing one or more nonconformities.

This distinction matters during AQL counting.

Suppose one plush has:

• an open shoulder seam;
• crooked mouth embroidery;
• one loose thread.

That product contains three defects, but it is still one defective unit.

If the open seam and crooked mouth are Major and the loose thread is Minor, the unit should normally be counted once under the highest applicable severity while all three observations remain recorded.

Heyzizi’s controlled counting method follows this principle: when one unit contains several issues, the highest defect class determines how that unit enters the defect count, while all observed problems remain documented. Safety and functional conditions take priority over cosmetic ones.

This avoids two opposite errors:

• counting one poor unit several times and exaggerating the lot failure rate;
• recording only the highest issue and losing useful evidence about recurring workmanship problems.

The statistical count and the technical defect record should therefore be kept separately.

What Is a Critical Plush Defect?

A critical defect is a condition that creates an unacceptable safety risk or makes normal shipment release inappropriate even if every other sampled unit looks correct.

Examples include:

• broken sewing needle inside the plush;
• sharp metal fragment;
• accessible sharp edge;
• loose magnet;
• leaking glass beads or weighted material;
• serious small-part detachment risk;
• unsafe battery compartment;
• battery leakage;
• overheating or short circuit;
• serious incorrect warning or age information;
• use of an unapproved restricted material.

Heyzizi’s documented quality criteria classify needle fragments, accessible sharp edges, magnet detachment, weighted-material leakage, unsafe battery conditions and serious small-part risks within the critical category.

The most important rule is:

Critical severity is determined by consequence, not appearance.

A 1 mm steel fragment can be Critical.

A 30 mm stain can still be Major.

That is why critical findings should never be “balanced” against excellent cosmetic results elsewhere in the sample.

Where zero acceptance is required, it is more precise to state:

Ac 0 / Re 1

rather than thinking of critical issues as an ordinary percentage allowance.

Which Defects Are Major?

A major defect materially affects normal use, durability, expected appearance, identity, packaging usability or conformity to the approved plush.

Examples include:

• clearly distorted face;
• open seam;
• missing component;
• severe color mismatch;
• wrong embroidery;
• wrong main material;
• major stuffing imbalance;
• non-working sound or light function;
• detached decorative component;
• wrong care label;
• wrong barcode;
• incorrect SKU;
• obvious stain;
• heavily damaged retail packing;
• plush unable to hold the approved shape.

Heyzizi’s quality records include severe facial misalignment, out-of-spec dimensions or weight, major color difference, burst seams, detached parts, failed functions, incorrect labels, incorrect barcodes, wrong SKU packing and serious head deformation as Major examples.

Major does not mean physically large.

A character eye may move only 3 mm yet become Major because the face no longer resembles the approved Golden Sample.

A much longer thread hidden under clothing may remain Minor because it can be trimmed without changing function or appearance.

The useful test is:

Would this condition materially change how the finished plush is used, recognized, displayed, sold or accepted?

If yes, Major is usually the more appropriate classification.

What Counts as a Minor Defect?

A minor defect affects finish quality but does not materially affect safety, use, durability or the main identity of the plush.

Examples include:

• small removable thread ends;
• light pile disorder;
• slight non-critical dimensional variation;
• small finishing mark on a hidden area;
• minor packaging abrasion;
• slight print displacement that does not affect recognition;
• small cosmetic asymmetry away from controlled facial features.

Heyzizi’s documented minor examples include limited trimmable threads, light pile disorder, small non-critical size differences, light packaging scuffs and minor finishing marks outside prominent areas.

Minor does not mean “ignore.”

One loose thread may be Minor.

Thirty loose threads across a small inspection sample can show poor finishing discipline across the lot.

The classification of each thread remains Minor, but the frequency may cause the minor-defect acceptance limit to be exceeded.

This is one of the most important distinctions in AQL work:

Severity describes the effect of one defect.

Frequency describes how widely that defect appears in the lot.

Do not mix the two.

Can the Same Defect Be Major on One Plush and Minor on Another?

Yes.

Defect classification depends on context.

Consider a 2 mm mouth shift.

For a simple animal plush with a broad smile, the change may be difficult to notice.

Possible classification: Minor.

For a licensed character whose signature expression depends on a small angled mouth, the same 2 mm movement may visibly change the face.

Possible classification: Major.

The same logic applies to color.

A slight shade difference on an inner foot panel may be Minor.

The same shade difference across the main face fabric of a recognizable mascot may be Major.

Other context-sensitive areas include:

• embroidery placement;
• eye spacing;
• ear angle;
• print registration;
• accessory position;
• pile direction;
• clothing alignment;
• decorative hardware.

A reusable defect list is useful, but it should never replace project-specific judgment.

The physical product and approved reference decide the consequence.

How Should Character Accuracy Defects Be Classified?

Character accuracy deserves its own review because several small dimensional changes can combine into a large visual difference.

The most sensitive areas usually include:

• eye spacing;
• eye height;
• pupil direction;
• muzzle position;
• mouth angle;
• ear angle;
• forehead fullness;
• head width;
• embroidery size;
• pile direction;
• stuffing pressure.

Consider this example:

• left eye moves outward 2 mm;
• right eye moves outward 2 mm;
• muzzle drops 3 mm;
• head receives slightly more filling.

Every single deviation is small.

Together, the face can look completely different.

This is why character inspection should not be reduced to one measurement.

The face should be compared against the Golden Sample from:

• front view;
• left side;
• right side;
• three-quarter view where needed.

For important IP or mascot work, facial zones can also have tighter dimensional controls than the rest of the body.

The correct classification depends on whether the deviation changes recognition or expression.

How Should the Golden Sample Be Used?

The Golden Sample is the physical reference for approved appearance, but it should be supported by technical records.

A strong inspection set can include:

• Golden Sample;
• final paper pattern;
• BOM;
• approved fabric codes;
• physical color reference;
• embroidery file;
• thread reference;
• print file;
• critical dimensions;
• filling requirement;
• accessory details;
• label files;
• barcode files;
• packaging specification.

Heyzizi’s consistency controls connect Golden Sample approval with locked pattern versions, BOM, color references, embroidery files, first-piece confirmation and production inspection.

The Golden Sample shows what the finished plush should look like.

The technical file explains how that result is controlled.

For example, “eye spacing must match the sample” is subjective.

A stronger specification might state:

• eye center spacing;
• eye-to-mouth distance;
• mouth width;
• head width;
• ear attachment location.

That reduces disagreement when inspection begins.

How Should Borderline Defects Be Classified?

Borderline defects should be defined before final inspection wherever possible.

Useful tools include:

• approved defect photographs;
• limit samples;
• critical dimension tolerances;
• visual zones;
• defect descriptions;
• acceptable and unacceptable examples.

A limit sample is particularly effective for plush because many surface and facial conditions are difficult to describe precisely.

Instead of writing:

“Slight facial asymmetry is acceptable,”

the project can define:

Acceptable: small cheek variation that does not change expression.

Unacceptable: eye, mouth or muzzle displacement that visibly changes the approved face.

The sequence for a borderline decision should be:

Safety → function → durability → identity → appearance → finish

If the defect affects several categories, use the highest meaningful consequence.

Can Minor Defects Cause the Entire Lot to Fail?

Yes.

Minor defects are individually low severity, but too many can fail the minor AQL criterion.

For example, assume a 200-piece sample has:

• Critical: 0
• Major: 5
• Minor: 16

If the agreed Major criterion allows 5 but the Minor rejection number is 15, the lot can fail because of the Minor count even though there are no serious safety findings.

This is exactly why small cosmetic defects are still recorded.

The reverse can also happen:

• Critical: 1
• Major: 0
• Minor: 0

The lot still fails under a zero-acceptance critical rule.

The defect categories operate independently.

Strong performance in one category does not cancel failure in another.

Does Repetition Make a Defect More Severe?

Usually, repetition changes the lot response, not the severity of each individual defect.

If one unit has a small trimmable thread, that remains Minor.

If 40 units have the same thread problem, those 40 issues are still Minor, but the lot may exceed the minor-defect acceptance limit.

Repeated identical defects also reveal something important: the issue may be systemic.

Examples include:

• same seam opening in the same body location;
• same mouth shifted left;
• same wrong barcode;
• same underfilled head;
• same accessory mounted backward.

A repeated pattern should trigger investigation even before the numerical AQL limit is reached if it suggests a process problem.

Ten unrelated imperfections tell one story.

Ten identical imperfections usually tell another.

When Should a Defect Trigger 100% Sorting?

AQL sampling gives evidence about the lot. It does not prevent expanded inspection when the evidence shows a concentrated problem.

100% sorting may become appropriate when:

• one repeated Major defect appears;
• an affected sewing line can be identified;
• one embroidery batch is suspect;
• one SKU has concentrated failure;
• labels or barcodes may be systematically wrong;
• safety-related attachment strength is uncertain.

Critical findings can require even stronger containment.

Heyzizi’s documented CAPA process includes temporary containment, root-cause analysis, corrective action, preventive action, 100% screening where required, reinspection and quality approval before production or shipment resumes.

The objective is not simply to make the next AQL sample pass.

The objective is to remove the cause that produced the defect.

How Should Defect Classification Affect the Final Lot Decision?

The complete decision chain should be:

Approved specification → defect observed → severity assigned → defective unit counted → Ac/Re compared → lot action decided

Possible lot actions include:

• Accept;
• Hold;
• Sort;
• Rework;
• Reinspect;
• Reject.

Consider a 200-unit inspection:

Defect Class	Result	Interpretation
Critical	0	within zero-acceptance rule
Major	9	below Re 11
Minor	13	below Re 15

The lot remains within those stated criteria.

Now change only the Major result:

Defect Class	Result	Interpretation
Critical	0	acceptable
Major	11	rejection threshold reached
Minor	6	acceptable

The lot fails even though Minor performance is good.

This is why defect classification must be completed before counting begins.

If the same crooked face is called Minor by one inspector and Major by another, the statistical result is no longer reliable.

A defensible plush AQL process therefore depends on a clear evidence chain:

Golden Sample + technical specification → defined defect → severity → count → AQL decision → corrective action where required.

When those links are agreed before shipment, Critical, Major and Minor become practical quality controls rather than subjective labels.

What Is Checked During AQL Inspection for Plush Toys?

A plush AQL inspection should verify the sampled units against the approved Golden Sample and final specifications for character appearance, dimensions, fabric, seams, filling, attachments, function, labels, barcodes and packaging. Some physical tests can be performed on selected units, while safety-sensitive controls such as needle detection, electronic function or certain weighted and magnetic features may require 100% checking outside the normal AQL sample.

AQL inspection should never begin with a generic checklist alone. The first step is establishing what the finished plush is supposed to be.

The inspection reference should normally include:

• approved Golden Sample;
• final dimensions;
• final fabric and color references;
• embroidery or print files;
• filling requirement;
• accessory specifications;
• care-label files;
• hangtag files;
• SKU and barcode files;
• retail packing instructions;
• carton assortment and shipping marks.

Without those references, obvious defects can still be found, but subtle production differences become difficult to judge consistently.

Plush is also different from rigid products because shape is created by several interacting variables. Paper pattern, fabric stretch, pile direction, seam construction and filling pressure all influence the finished result. A head can technically meet its width specification but still look wrong because the forehead is overfilled. Eye embroidery can use the correct file yet appear farther apart after stuffing stretches the face.

That is why a strong final inspection combines four types of checks:

Control Type	What It Verifies	Common Plush Use
Visual inspection	appearance and workmanship	face, seams, stains, pile, embroidery
Measurement	numerical conformity	height, width, limb length, weight
Physical check	strength and durability	seams, eyes, accessories, hardware
Full-unit control	characteristics requiring individual verification	needle detection, selected functions, critical coding

The most reliable result comes from using each method where it makes sense rather than forcing every characteristic into the same AQL sample.

How Is Character Accuracy Inspected?

Character accuracy should be checked against the Golden Sample from several viewing angles, not only by measuring total height.

The areas most likely to change the visual identity include:

• eye spacing;
• eye height;
• pupil direction;
• mouth angle;
• muzzle position;
• ear position;
• head width;
• forehead volume;
• embroidery scale;
• filling balance.

A useful inspection combines visual comparison with several controlled measurements.

For example, instead of recording only “face should match approved sample,” the specification may define:

Feature	Control Method
Eye spacing	center-to-center measurement
Eye height	reference from head seam
Mouth width	embroidery dimension
Muzzle position	reference to eye line
Head width	defined measurement location
Ear placement	seam location + visual comparison

Small dimensional changes can combine into a large visual difference. Two eyes moving outward by 2 mm each and a muzzle dropping 3 mm may produce a clearly different expression even though each deviation appears small by itself.

Character inspection should therefore judge the whole face as well as individual dimensions.

How Are Seams and Stitching Checked?

Seam inspection should evaluate both visible workmanship and structural integrity.

Inspectors should look for:

• open seams;
• skipped stitches;
• broken thread;
• loose closing stitches;
• seam puckering;
• filling visible through the seam;
• excessive seam distortion;
• fabric trapped incorrectly;
• pile caught inside the seam;
• uneven closure areas.

Stuffing openings deserve special attention because they are closed after the product has already been filled. Internal pressure can expose weaknesses that are not obvious when the unit is simply lying on a table.

A seam can therefore look acceptable while relaxed but begin opening when the plush is squeezed or gently tensioned.

For selected units, manual stretching or seam-strength checking can be added to the visual inspection.

Where the same seam defect appears repeatedly in the sample—especially in the same body location—it should be treated as evidence of a possible process problem rather than several unrelated cosmetic faults.

How Is Stuffing Consistency Checked?

Stuffing affects handfeel, posture, dimensions and character appearance at the same time.

Inspection should compare:

• total unit weight;
• head firmness;
• body firmness;
• limb filling;
• left-right balance;
• cheek shape;
• belly volume;
• posture;
• recovery after compression.

Underfilled plush often looks narrow, flat or tired. Overfilled plush can become too round, stretch facial features and place unnecessary pressure on seams.

Weight measurement is useful because it can reveal hidden filling variation.

For example, if most sampled 25 cm plush units cluster close to the approved weight but several are noticeably lighter, those pieces should be checked for underfilled heads, limbs or bodies.

Weighted plush needs tighter control because the product mass may be part of its intended function. In that case, checking every finished unit can be more appropriate than relying only on an AQL weight sample.

How Are Eyes and Accessories Checked?

Eyes, noses and accessories should be inspected for both placement and attachment security.

Typical checks include:

• correct component;
• correct color;
• correct orientation;
• correct position;
• secure attachment;
• no sharp edge;
• no fabric damage around the connection.

Accessories can include:

• plastic eyes;
• noses;
• buttons;
• bows;
• key rings;
• lobster clasps;
• bells;
• zippers;
• clothing;
• magnets;
• vinyl parts;
• decorative hardware.

A visually correct eye is not automatically a secure eye.

Likewise, a metal clasp can look excellent while the fabric loop supporting it is too weak.

The inspection should therefore distinguish component appearance from connection strength.

For children’s products, attachment failure can become much more serious when a detached hard part becomes accessible. The final classification depends on the product structure, intended age and applicable safety requirements.

How Are Plush Pull-Tests Integrated Into AQL?

Pull tests are best treated as selected physical tests within the inspection plan rather than ordinary visual defect counts.

A practical sequence is:

1. Select representative units from the inspected lot.
2. Identify seams and attachments requiring strength verification.
3. Apply the defined pull or tension method.
4. Record detachment, seam opening, tearing or permanent damage.
5. Classify the result according to its consequence.
6. Expand checking if repeated weakness is found.

Possible pull-test locations include:

• plastic eyes;
• noses;
• buttons;
• decorative bows;
• keychain loops;
• metal fittings;
• clothing decorations;
• vinyl-to-fabric joints;
• stuffing closures.

The significance of a failure depends on what comes loose.

A decorative element on an adult collectible may mainly create an appearance problem. A detached small hard component on a children’s plush can create a much higher safety concern.

If several selected units fail at the same connection, simply recording a few Major defects may not be sufficient. The affected attachment process may require wider screening before shipment release.

How Are Size and Weight Checked?

Dimensions should be measured from defined landmarks so different inspectors obtain comparable results.

Possible measurements include:

• seated height;
• standing height;
• head width;
• body width;
• arm length;
• leg length;
• ear length;
• tail length;
• accessory position.

Long-pile fabric creates a common measurement problem. Measuring to the tip of the fur can produce inconsistent results depending on how the pile is lying.

The specification should state whether dimensions are measured:

• with pile naturally relaxed;
• with pile lightly compressed;
• from seam to seam;
• between defined anatomical landmarks.

Weight should also be checked on a calibrated scale.

Unexpected variation may indicate:

• inconsistent filling;
• missing internal part;
• excessive filling;
• incorrect weighted insert;
• wrong accessory.

Size and weight should not be judged independently from appearance. A plush can meet nominal height yet still look too thin, too round or improperly proportioned.

How Are Labels and Barcodes Checked?

Labels and codes should be treated as formal product requirements.

Inspection can include:

• care-label content;
• fiber information where applicable;
• country-of-origin information;
• age information;
• warning statements;
• tracking information;
• hangtag artwork;
• SKU code;
• UPC/EAN;
• FNSKU;
• carton barcode.

The inspector should verify both content and attachment.

A correct label that is sewn into the wrong SKU is still a failure.

Multi-SKU collections create higher risk because similar characters may share:

• carton size;
• bag size;
• hangtag shape;
• outer packaging.

A single wrong barcode file can therefore affect hundreds of units without changing the physical plush.

For mixed-SKU or fulfillment-sensitive programs, 100% barcode scanning can be more appropriate than relying only on random visual checking.

How Is Packaging Inspected?

Packaging inspection should verify product identity, presentation, protection and shipping configuration.

Checks can include:

• correct polybag;
• correct retail box;
• correct insert;
• carton quantity;
• SKU assortment;
• carton dimensions;
• carton weight;
• carton marks;
• barcode;
• carton closure;
• internal dividers;
• compression condition.

Plush packaging has one additional concern: shape recovery.

A carton can arrive undamaged while the plush inside suffers:

• flattened ears;
• crushed muzzle;
• deformed head;
• compressed long pile;
• bent accessories.

For products packed under compression, several sampled units should be unpacked and allowed to recover before their final appearance is judged.

The packing method should therefore be assessed not only by how much space it saves but by whether the plush can recover to the approved appearance after unpacking.

Does AQL Testing Include Metal Detection for Plush?

Metal detection should not normally be reduced to ordinary AQL sampling.

AQL might require 80, 125 or 200 finished units to be inspected from a larger lot. If only those pieces are checked for needle contamination, thousands of untested units may remain.

For ordinary sewn plush without intentional fixed metal components, 100% needle detection before final packing is a stronger control.

Products intentionally containing metal, magnets, electronics or hardware need a different method because those parts may trigger conventional detectors.

Possible controls include:

• controlled metal registration;
• separated metal zones;
• handheld detectors;
• broken-needle records;
• equipment sensitivity verification;
• localized checking.

Metal control and final AQL inspection therefore serve different purposes.

AQL evaluates lot consistency.

Needle detection addresses metal contamination at individual-unit level.

Passing one does not replace the other.

AQL vs 100% Inspection: Which Is Better?

Neither method should be used for everything.

AQL is efficient when a representative sample can provide useful evidence about the condition of the production lot.

100% checking is more suitable when every individual unit must demonstrate a required characteristic.

A practical combination is:

Inspection Item	AQL	100% Control May Be Better
Overall appearance	✓	sometimes also checked during finishing
Facial accuracy	✓	special premium programs may add broader checking
Dimensions	✓
Standard unit weight	✓
Weighted-product mass		✓
Electronic function		✓
Needle detection		✓
Magnet position/function		✓
Mixed-SKU barcode		✓
Packaging appearance	✓

The strongest inspection system is therefore layered.

General workmanship can be evaluated through AQL sampling.

Specific attachments can receive physical tests.

Safety-sensitive or function-critical characteristics can receive full-unit verification.

If the AQL sample then reveals a repeated defect—such as the same open seam, wrong barcode or weak attachment—the response can escalate to 100% sorting for that characteristic.

The inspection chain becomes:

Golden Sample and technical files → random AQL sample → visual and measurement checks → selected physical tests → separate full-unit controls → defect classification → Ac/Re decision → sorting or rework where needed.

That process provides much stronger evidence of bulk consistency than a final visual check alone.

When and Who Should Perform AQL Inspection for Bulk Plush?

Final AQL inspection should be carried out late enough for the sampled units to represent the real shipment, but early enough to leave time for sorting, rework and reinspection if the lot fails. For most plush orders, the strongest timing is after production is complete or nearly complete and most export packing is finished. Internal QC controls the process; an independent inspection party can add a separate release check when required.

Timing is not a small administrative detail. It changes the value of the inspection.

If AQL inspection happens too early, the sampled units may not represent later sewing, filling or packing work. If it happens only a few hours before container loading, a failed result can create an impossible choice between delaying shipment and releasing goods that have not been corrected properly.

Two widely used third-party practices illustrate the timing:

• QIMA states that its pre-shipment inspection is performed when at least 80% of the order is complete, with packing already underway.
• SGS describes its Final Random Inspection as taking place when 100% of goods are produced and at least 80% are export packed.

These are service conditions used by those inspection organizations, not a universal legal rule. The better principle is:

enough finished and packed goods to represent the actual shipment + enough remaining time to correct a failure properly.

For custom plush, the final AQL inspection should sit at the end of a longer quality chain:

incoming materials → first-piece approval → in-process checks → finished checks → special 100% controls → final AQL inspection → corrective work if required → shipment release

That sequence is much stronger than treating final inspection as the first serious quality check.

When Should Pre-Shipment AQL Inspection Be Done?

A final random inspection is most useful when the production lot has reached a stable finished condition.

The inspector should be able to sample from:

• finished units;
• multiple cartons;
• different carton positions;
• all relevant SKUs;
• actual retail packing;
• actual shipping cartons.

If half the order is still on sewing lines, the available sample may overrepresent earlier production.

If retail packing has not started, the inspection cannot verify:

• wrong polybags;
• wrong display boxes;
• barcode errors;
• carton assortment;
• incorrect carton marks;
• packing compression;
• missing inserts.

For this reason, a plush PSI is normally scheduled close to completion rather than at the middle of sewing production.

A useful practical schedule is:

Stage	Inspection Purpose
Materials received	fabric, color, accessories, labels
First production pieces	shape, embroidery, filling, construction
During production	repeated workmanship and process stability
80–100% complete	final lot AQL inspection
After correction	reinspection if required
Before loading	quantity, cartons and shipment release

During-production inspection still has value. QIMA describes this type of check at roughly 20–80% completion, which provides time to identify repeated defects before the entire lot is finished.

For a difficult character, multi-SKU collection or first run, that earlier intervention can prevent a final AQL failure from becoming a large rework project.

How Much Time Should Be Left Before Shipment?

AQL inspection should not be booked directly against vessel cutoff, warehouse appointment or launch deadline.

Consider a 10,000-piece order with a 200-piece final sample.

If the inspection finds repeated open neck seams, the next steps may include:

• opening affected cartons;
• identifying the production group involved;
• 100% sorting;
• repairing open seams;
• checking repaired pieces independently;
• repacking;
• drawing a new random sample;
• issuing a new inspection result.

That cannot always be completed safely in one afternoon.

A useful shipment plan therefore includes a quality buffer between inspection and dispatch.

The amount of buffer depends on:

• quantity;
• SKU count;
• carton count;
• defect type;
• repair difficulty;
• access to replacement materials;
• third-party availability.

Heyzizi’s factual records specifically state that third-party inspection timing depends on the inspection organization’s scheduling and should not be promised as an automatic fixed duration.

The inspection date should therefore be included in the production calendar rather than added only after packing is almost complete.

Internal QC vs Third-Party Inspection: Which Is Better?

They perform different roles.

Internal QC has access to the product throughout production. It can stop an issue when only dozens of units are affected rather than discovering the same fault after several thousand pieces have been packed.

Internal controls can include:

• incoming material inspection;
• first-piece approval;
• embroidery and print review;
• sewing-line checks;
• filling checks;
• finished appearance inspection;
• needle detection;
• functional checks;
• packing verification.

An independent third party adds separation between daily production activity and the final lot assessment.

It can be particularly useful for:

• first orders;
• large retail programs;
• licensed characters;
• high-value shipments;
• multi-SKU collections;
• orders requiring an independent report;
• lots with earlier quality instability.

The strongest arrangement is often not internal QC or third party.

It is:

internal process control + independent final verification where needed.

Heyzizi’s quality records support external inspection and retain third-party inspection reports together with incoming, first-piece, in-process, final and needle-detection records.

That creates evidence from several stages rather than one isolated inspection day.

Who Should Set the AQL Standard?

The inspection limits should be agreed before final production is completed.

The inspector should apply the requirement, not invent it after defects have already been found.

The inspection instruction should define:

• sampling standard;
• inspection level;
• critical acceptance rule;
• Major AQL;
• Minor AQL;
• defect definitions;
• special tests;
• Golden Sample reference;
• SKU-specific requirements.

For example:

ISO 2859-1 / General Level II / Critical Ac 0 Re 1 / Major 2.5 / Minor 4.0

is much clearer than:

“Use standard AQL.”

Premium licensed characters may require tighter facial and appearance criteria than an ordinary animal plush.

Baby, magnetic, weighted or electronic products may require additional controls beyond the AQL values.

If a retailer, license owner or brand already has an established inspection protocol, those requirements should be incorporated before volume production.

Late changes to AQL limits after the lot is finished create unnecessary disputes because the production process was not built around the same release criteria.

What Documents Should Inspectors Receive?

A custom plush inspector needs more than an order quantity.

The inspection set should normally include:

• Golden Sample;
• final product specification;
• final measurements;
• approved material references;
• embroidery files;
• print references;
• approved colors;
• accessory details;
• SKU list;
• quantity by SKU;
• care-label files;
• hangtags;
• barcode files;
• retail packing instructions;
• carton assortment;
• carton marks;
• defined AQL criteria.

Heyzizi’s third-party preparation records include the order information, final specifications, Golden Sample, AQL requirement, packing information and quantity list.

This information is essential because custom plush often contains intentional asymmetry.

One ear may be deliberately lower.

One eye may intentionally be larger.

A tail may be designed to curve only to one side.

Without the Golden Sample and approved files, an inspector may classify correct character features as defects—or accept incorrect ones because no precise reference exists.

How Long Does a Plush AQL Inspection Take?

The actual on-site time depends on the sample quantity and the amount of physical checking required.

A straightforward single-SKU order may fit within one inspector-day.

A complex inspection can take longer when it contains:

• 200–315 sampled units;
• many SKUs;
• extensive measurements;
• pull testing;
• electronic functions;
• magnetic parts;
• weighted inserts;
• several packing styles;
• carton assortment checks.

QIMA states that its inspection reports can be delivered the same day and that inspectors may be deployed quickly, while SGS currently states that booking generally requires about 5–7 working days for its FRI service.

Those figures refer to specific inspection services and should not be treated as a guaranteed duration for every plush order.

A 500-piece single-character plush and a 30-SKU collectible program should not be expected to require the same inspection effort.

Should Every SKU Be Sampled Separately?

Not every multi-SKU order must be divided into completely separate AQL lots, but every meaningful SKU needs enough inspection coverage to reveal SKU-specific problems.

Consider a 10,000-piece order:

SKU	Quantity
Character A	3,000
Character B	2,500
Character C	2,000
Character D	1,500
Character E	700
Character F	300

A 200-unit random sample drawn only in proportion to quantity may contain very few units of Character F.

If F has a unique hat, embroidery file or fabric, a handful of pieces provides weak evidence about those unique features.

Three practical methods are available:

• treat important SKUs as separate lots;
• stratify the overall AQL sample across SKUs;
• keep the combined lot but establish minimum checks per design.

SKU-specific measurement is particularly important because each character can have its own:

• paper pattern;
• embroidery;
• dimensions;
• accessories;
• packaging;
• barcode.

Heyzizi’s outgoing control requires ordinary size and weight checks by AQL while maintaining a minimum of five measured units per SKU in the applicable inspection batch.

When Should Third-Party Inspection Be Required?

There is no need to use an outside inspection organization for every order simply for appearance.

Independent inspection adds the most value when the commercial or technical risk justifies another layer of verification.

Situations can include:

• first production run;
• large shipment value;
• strict retail program;
• licensed character;
• several production locations;
• many SKUs;
• earlier failed inspection;
• repeated major defects;
• contractual requirement.

If internal QC already detects a serious problem, the correct action is not to hide it until the independent inspection.

The affected goods should be contained and corrected first.

An outside inspection works best as verification of a controlled lot, not as a substitute for daily production discipline.

Can AQL Replace Toy Safety Testing?

No.

AQL inspection and laboratory safety testing establish different types of evidence.

AQL evaluates whether a production lot conforms to approved workmanship, dimensions, functions, labeling and packing through sampling.

Toy testing evaluates applicable regulatory safety requirements.

A plush lot can therefore pass AQL and still fail a laboratory requirement.

It can also pass laboratory testing but fail AQL because:

• embroidery is visibly wrong;
• seams are open;
• SKU labels are incorrect;
• filling is inconsistent;
• packaging does not match approval.

The two results should not be used interchangeably.

Safety requirements depend on product design, age grading, components and destination requirements.

The final inspection should confirm that the required reports and product versions correspond, but an on-site AQL inspection does not replace laboratory evidence.

When Is 100% Inspection More Appropriate?

Full-unit checking becomes more appropriate when every individual piece must demonstrate a required condition or when one missed failure carries unusually high risk.

Examples include:

• needle detection;
• electronic function;
• magnet position and function;
• weighted-product mass;
• critical barcode matching;
• special hardware function.

Heyzizi’s controlled process requires 100% needle detection for ordinary plush where applicable, along with category-specific full checks for electronics, weighted items, magnets and certain barcode situations.

AQL can also reveal when full sorting becomes necessary.

If a random sample finds the same open seam repeatedly, the response should not necessarily be “inspect another random sample.”

The repeated pattern may justify checking every affected unit first.

The practical distinction is:

AQL determines whether the lot shows acceptable overall consistency.

100% inspection verifies a required condition on every individual unit.

A well-controlled bulk plush order often needs both.

The most defensible release sequence is therefore:

process controls → finished-goods controls → applicable 100% checks → representative AQL inspection → corrective work if needed → independent reinspection where required → shipment release.

When the timing, inspection authority, reference files and release criteria are fixed before production reaches the packing stage, AQL becomes a useful shipment decision rather than a last-minute argument over cartons.

What Happens When a Bulk Plush Order Fails AQL?

When a bulk plush order fails AQL, the affected lot should be placed on hold rather than released for shipment. The inspection findings are then reviewed by defect class, SKU, carton and production source. Depending on the pattern, the lot may require targeted sorting, 100% screening, rework or replacement. After correction, a new representative sample should be drawn for reinspection before shipment is released.

An AQL failure does not automatically mean the entire order is unusable.

It means the sample provided enough evidence that the lot cannot be accepted under the agreed inspection plan in its current condition.

What happens next depends on three things:

• severity — Critical, Major or Minor;
• frequency — isolated or repeated;
• distribution — random across the lot or concentrated in one SKU, line, material batch or packing group.

Those three factors determine whether the correct response is simple repair, targeted sorting, full screening or shipment rejection.

A useful failure sequence is:

Fail → Hold → Confirm → Trace → Sort → Rework → Verify → Reinspect → Release

The aim is not merely to make the next sample look better. The affected condition has to be controlled across the actual production lot.

What Causes a Plush AQL Inspection to Fail?

A lot fails when one or more agreed rejection criteria are reached.

Common causes include:

• Critical defect found under an Ac 0 / Re 1 rule;
• Major-defective units reach the Major rejection number;
• Minor-defective units reach the Minor rejection number;
• required function fails;
• product version does not match the approved Golden Sample;
• serious SKU, barcode or packing errors;
• repeated physical-test failures.

Consider a 5,000-piece plush order inspected under General Level II.

Sample quantity: 200 units

Assume the agreed criteria are:

Defect Class	Acceptance	Rejection
Critical	0	1
Major AQL 2.5	10	11
Minor AQL 4.0	14	15

Now assume the inspection finds:

• Critical: 0
• Major: 12
• Minor: 8

The lot fails because the Major rejection threshold has been exceeded.

The eight Minor defects do not cause the failure, but they should still be recorded because they provide additional evidence about production consistency.

What Should Happen Immediately After a Fail?

The first action should be containment, not reinspection.

The affected lot should be held so cartons are not loaded, transferred or mixed with approved goods while the cause is still unclear.

The inspection report should then be reviewed for:

• SKU;
• defect type;
• defect severity;
• quantity found;
• carton number;
• photographed evidence;
• production date or batch;
• sewing line where traceable;
• embroidery or print batch;
• filling group;
• packing group.

The main task is to determine whether the findings are random or systematic.

For example:

12 unrelated Major defects

may indicate broad workmanship instability.

12 identical open neck seams

strongly suggest one repeated construction problem.

12 wrong barcodes all on Character C

suggest a SKU-specific packing or file-control failure.

The same AQL result can therefore lead to very different corrective actions.

What Happens After a Critical Defect Is Found?

A Critical defect requires stronger containment than an ordinary Major or Minor failure.

Examples include:

• needle fragment;
• accessible sharp metal;
• dangerous loose magnet;
• leaking weighted media;
• unsafe battery condition;
• serious small-part detachment risk.

If one Critical condition is discovered under a zero-acceptance rule, removing that one unit is not enough.

The next issue is whether other units may have been exposed to the same cause.

For example, a broken needle fragment may require:

• identifying the sewing workstation;
• reviewing broken-needle records;
• identifying production made during the affected period;
• isolating related cartons;
• running applicable units through metal detection again;
• confirming detector sensitivity;
• recording the investigation.

Likewise, one loose magnet may require a wider attachment-strength review because the problem may come from the same assembly method across hundreds of units.

Critical findings should be treated as evidence of possible broader exposure, not just one defective piece.

How Do You Know Whether 100% Sorting Is Needed?

Full sorting becomes appropriate when the sample indicates that a defect may exist repeatedly through the affected lot.

Common triggers include:

• repeated open seams;
• repeated facial misalignment;
• wrong barcode applied systematically;
• incorrect care labels;
• one embroidery batch shifted;
• one accessory installed backward;
• weak hardware connection;
• one SKU packed incorrectly.

Consider a 6,000-piece order containing six characters.

If 11 Major defects are found and nine of them are the same crooked mouth on Character D, it is usually more useful to isolate Character D and inspect that SKU thoroughly than to blindly sort all six designs in the same way.

Sorting should therefore follow the defect source.

Possible scope:

All units

when the cause could affect the entire order.

One SKU

when the defect is design-specific.

One production batch

when records identify a controlled production period.

One carton group

when the failure occurred during packing.

The purpose of traceability is to narrow the affected population without leaving suspect goods untreated.

Sorting vs Rework: What Is the Difference?

Sorting identifies conforming and nonconforming units.

Rework changes a nonconforming unit so it can meet the approved requirement.

Examples:

Defect	Sorting	Possible Rework
Wrong barcode	separate affected units	replace barcode
Open seam	isolate open-seam units	reopen/close correctly
Underfilled head	identify soft units	add filling
Light removable stain	separate stained units	approved cleaning
Wrong hangtag	separate affected SKU	replace hangtag
Crooked embroidery	isolate defective pieces	panel replacement may be required

Not every defect should be repaired.

Some corrections can damage the plush more than the original fault.

For example, trying to remove incorrect dense embroidery from short-pile fabric can leave needle holes, backing damage or visible surface marks. In such a case, replacing the affected panel or rejecting the unit may be safer than cosmetic repair.

Every rework method should therefore be defined before large-scale repair starts.

A few repaired units should be checked first to confirm that the method itself does not create a new defect.

Who Should Inspect Reworked Plush?

The person performing the repair should not be the only person deciding that the repaired unit is acceptable.

Independent verification is important because rework can introduce secondary problems.

After seam repair, check:

• seam appearance;
• strength;
• shape;
• trapped pile;
• filling distribution.

After filling adjustment, check:

• unit weight;
• firmness;
• face shape;
• seam pressure.

After label replacement, check:

• correct version;
• correct SKU;
• placement;
• barcode scanning.

The product should return to finished-goods status only after the correction has been verified.

When Should a Failed Lot Be Reinspected?

Reinspection should take place only after the corrective work has been completed and internally verified.

Simply drawing another random sample immediately after failure does not solve anything.

A stronger sequence is:

Failed AQL

→ isolate affected population

→ identify cause

→ sort or rework

→ verify correction

→ restore lot integrity

→ draw a new random sample

→ reinspect

The reinspection sample should be newly selected from the corrected production lot.

The original failed sample should not become the whole basis of the second inspection.

For example, if 200 units were inspected and 12 Major failures were repaired, checking only those 12 repaired units proves that those pieces were corrected. It does not prove that the remaining 4,800 units are under control.

A new representative random sample provides much stronger evidence.

Should Reinspection Use the Same AQL?

The commercial agreement should define this in advance.

In many cases, the same agreed AQL criteria can be used for reinspection after effective corrective action.

However, a serious or repeated failure can justify:

• tighter inspection;
• a larger sample;
• separate SKU inspection;
• 100% screening of the failed characteristic;
• additional physical tests.

The decision should reflect why the first inspection failed.

A widespread cosmetic finishing issue and a repeated attachment-strength failure should not necessarily receive the same reinspection treatment.

The more serious the evidence, the stronger the verification should become.

Who Pays for a Failed AQL Re-Inspection?

There is no universal rule assigning every reinspection charge to the same party.

Responsibility should follow the cause of the failure and, ideally, be written into the order or quality agreement before production begins.

A practical responsibility matrix is:

Cause of Failure	Cost Direction
Production does not match approved Golden Sample	production side
Wrong material used	production side
Workmanship failure	production side
Wrong packing executed	production side
Incorrect barcode supplied in approved file	file-originating side
Approved Golden Sample itself contains the issue	depends on approval responsibility
Late design change after production	requesting side
Ambiguous written approval	shared according to responsibility

This is particularly important with independent inspection services because a failed lot can create both a reinspection fee and additional sorting or repair costs.

The cleanest arrangement states before production:

• first inspection responsibility;
• reinspection responsibility;
• sorting cost responsibility;
• repair responsibility;
• effect on shipment date.

That prevents quality discussions from becoming cost negotiations after the failure has already happened.

Can Shipment Proceed After an AQL Failure?

Normal shipment release should remain on hold until the failure has been resolved.

Possible dispositions include:

• targeted sorting;
• 100% screening;
• repair;
• repacking;
• relabeling;
• partial replacement;
• reinspection.

For purely cosmetic deviations, a documented commercial acceptance may occasionally be possible if the relevant approval authority deliberately accepts the known deviation.

Safety-related conditions should be treated differently.

A discount does not make a needle fragment safe.

A lower price does not correct a loose magnet.

A shipping deadline does not eliminate the risk created by leaking weighted material.

Where the failure involves safety, critical function or serious regulatory information, physical correction should take priority over commercial convenience.

How Should Failed Units Be Recorded?

The inspection record should make it possible to trace what happened after the initial failure.

Useful fields include:

• order number;
• SKU;
• defect description;
• severity;
• affected quantity;
• sample quantity;
• carton number;
• batch;
• defect photograph;
• suspected cause;
• sorting quantity;
• repair quantity;
• rejected quantity;
• reinspection result.

Photographs should show enough context to understand the fault.

For example, a close-up photograph of a crooked eye is useful, but a second image showing the complete face makes it much easier to judge its effect on character appearance.

Data from the failed inspection should also be compared with the final sorting result.

If the sample suggested a high level of one specific defect but sorting finds very few affected pieces, that difference deserves review.

How Does CAPA Prevent the Same Failure From Returning?

Correcting defective units fixes the current lot.

Corrective and preventive action fixes the cause.

Consider repeated open neck seams.

A weak response is:

“Repair all open seams.”

A stronger investigation asks why they opened.

Possible causes include:

• seam allowance too narrow;
• incorrect stitch density;
• closure opening too short;
• excessive filling;
• weak thread;
• missing backstitch;
• inconsistent operator method.

The correction may therefore require:

• pattern adjustment;
• revised seam allowance;
• stitch specification change;
• new first-piece approval;
• operator retraining;
• additional in-line pull checking.

The same logic applies to crooked embroidery, barcode errors, loose accessories and packing mistakes.

A useful CAPA record should contain:

defect → containment → root cause → correction → prevention → verification

Without the final verification step, the action remains theoretical.

How Can Repeat Orders Benefit From a Failed AQL?

A failed lot can become valuable production evidence if its records are retained properly.

Useful repeat-order records include:

• defect photographs;
• final Golden Sample;
• corrected pattern;
• updated BOM;
• corrected embroidery file;
• revised filling requirement;
• rework instruction;
• sorting result;
• AQL report;
• CAPA record;
• final reinspection result.

When the same SKU is produced again, the earlier failure should become a specific control item during first-piece and in-process inspection.

For example:

Earlier failure: Character B mouth embroidery shifted left.

Next run control: first embroidery checked against final file before bulk embroidery begins.

The best result after an AQL failure is therefore not simply:

“The lot passed reinspection.”

It is:

the current shipment was corrected, the affected population was verified, the root cause was removed, and the next production run starts with stronger controls than the one before it.

That closes the quality loop from failed sample to stable repeat production.