We run cut-and-sew, wash, and decoration under one roof, so we inspect our own bulk every week.

This is the same framework we use on our floor and the same one your buyers will hold you to once you grow.

Here is the problem this guide solves.

You place a bulk order. A few weeks later, cartons land. You open two, the garments look fine, and you approve payment.

Then customers start emailing about crooked hems, a tag that itches, or a print that cracked after one wash. Now you are eating returns on a run you already paid for.

The fix is not “check harder.” The fix is a quality control (QC) standard, a written, agreed-upon set of rules that decides, on numbers alone, whether a production lot is accepted or rejected.

The backbone of that standard is a method called AQL. By the end of this guide, you will know what AQL means, where quality gets checked, how a lot passes or fails, who should be doing the checking, and how to write a standard your factory cannot argue with.

What this guide covers

  1. What AQL actually means and why acceptance is a number, not a feeling

  2. The Three Gates: where quality gets checked before a box is yours

  3. Sample sizes and defect classes: how a lot passes or fails

  4. Who actually inspects the factory: you, or a third party

  5. Setting a QC standard you can enforce

Chapter 1

What AQL Actually Means And Why Acceptance Is a Number, Not a Feeling

AQL stands for Acceptance Quality Limit. It is defined in two standards that say the same thing: the international ISO 2859-1 and the American ANSI/ASQ Z1.4.

Both trace back to a US military procurement standard (MIL-STD-105) built to check huge lots without opening every unit.

The whole apparel and consumer-goods world inherited it.

The official definition is dry: AQL is the “quality level that is the worst tolerable.” Translated for a founder: AQL is the maximum share of defective units you will accept in a lot before you reject the whole batch.

Why it exists: perfection is not the deal

Here is the uncomfortable truth every experienced manufacturer knows.

Visual inspection is not 100% reliable, and at streetwear volumes, virtually every production run contains a few defective pieces even after the factory has checked and repaired.

So the buyer and the factory do not agree on “zero defects.” They agree on a tolerance. AQL is the shared language for that tolerance.

It also does something quieter but important. By setting a limit and rejecting lots that miss it, you apply steady economic pressure on the factory to keep its average quality at or above your standard, run after run. That is the real point of the system.

The number that trips everyone up

Read this twice

“AQL 2.5” does not mean “2.5% defects allowed. “The number is a level that points into a lookup table, which then gives you an exact accept/reject count for your specific order size. The math rarely lands on a clean percentage. Treat 2.5 and 4.0 as labels, not fractions.

The three defect classes

Every defect gets sorted into one of three severity classes, and you set a separate AQL for each. This is the part founders skip, and it is the part that matters most:

  • Critical: anything unsafe or illegal, or that makes the product unsaleable. A broken needle left in a garment, a sharp pin, the wrong fiber content on the label, the wrong country of origin. AQL for critical is almost always 0: one is enough to fail the lot.

  • Major: a fault that would make a customer return the item: an open seam, a wrong size, a print in the wrong place, or a broken zipper. Usually set at AQL 2.5.

  • Minor a cosmetic flaw a trained inspector notices but most customers never would: a loose thread, a faint spot on an inside panel, a slight shade drift. Usually set at AQL 4.0.

The Streetwear Baseline

0 / 2.5 / 4.0. Zero critical, 2.5 major, 4.0 minor. This is the default AQL setting most consumer-goods and apparel programs use, and it is a sane starting point for a streetwear founder. Tighten major to 1.5 for premium pieces; loosen minor to 6.5 only if you truly do not care. Write these three numbers into every order, and you have the spine of a QC standard.

Different buyers use slightly different classifications, and big brands publish their own defect manuals, but the three-class structure is near-universal.

If you want to see how these classes get set during the sampling stage, where your golden sample locks the reference, that is covered in our companion guide.

Chapter 2

The Three Gates: Where Quality Gets Checked Before a Box Is Yours

AQL tells you how to judge a lot. This chapter tells you when. There are three moments in a production run where inspection actually changes the outcome.

Miss them, and you are left inspecting a finished, packed order you have already paid for at the worst possible time to find a problem.

The named framework

The Three Gates

Every bulk order passes three gates before it becomes yours. Each gate catches a different kind of failure, and each one that opens too late costs more than the one before it.

  1. The Setup Gate before the line runs. Also called pre-production inspection. Confirm the right fabric, trims, and approved sample are actually on the floor.
  2. The line gate while the line runs, at roughly 20–80% complete. Also called inline or during-production inspection (DUPRO). Catch a mis-set machine before it ruins thousands of pieces.
  3. The Ship Gate after production, at 80%+ complete and packed. Also called final random inspection (FRI). The last accept/reject decision before goods leave the floor.

Gate 1 Setup (pre-production)

This happens before mass sewing starts.

The inspector or your factory’s QC lead checks that the raw materials and trims physically on the floor match the tech pack and the bill of materials: the right fabric weight, the right zippers, the right woven labels, and the correct care-label wording.

The approved pre-production (PP) sample sits on the table as the reference.

Why it matters: nearly every expensive bulk disaster is a setup failure that nobody caught.

Wrong GSM knit. A substituted trim. A care label with the wrong fiber percentage.

Fixing it here costs a phone call. Fixing it at the Ship Gate costs the whole run.

This gate is about the inputs: how the sampling rounds that produced that PB sample actually work (proto, fit, PP, and sealed sample) is its own process, covered in the inspection points companion guide.

Gate 2 Line (inline / DUPRO)

This is the gate founders don’t know exists, and it is the most valuable one.

It runs while the order is being made; the common window is 20% to 80% complete, with many inspectors targeting past the 50% mark so the sample reflects real output.

At this gate, the inspector is on the sewing and finishing floor.

They pull finished-but-not-packed units and check them against the spec: stitch density, seam strength, measurements against the size chart, print and embroidery registration, and label placement. The magic is timing.

A skipped-stitch problem on a flatlock machine, or a body length running 2 cm long, gets caught while there is still time to fix the machine and salvage the rest of the run.

Why the Line Gate pays for itself

Catching a defect at 20% complete is dramatically cheaper than finding it in a packed shipment. You rework a few hundred pieces instead of scrapping thousands. For a new factory relationship, one line gate check is cheap insurance against a run that could sink a small brand.

Gate 3 Ship (final random inspection)

This is the classic “inspection” people picture. It runs when 80% or more of the order is finished and packed.

A random AQL sample is pulled from the cartons and checked in full: appearance, workmanship, measurements, function (zippers, buttons, drawcords), and critical labeling and packaging, which are often a legal requirement in the destination market.

The Ship Gate produces the single pass/fail decision that governs whether you accept the lot. It is the final gate, not the only gate.

If it is the only time you inspect, you have no way to fix anything it finds, which is exactly why the first two gates exist.

Upload Image...

Chapter 3

Sample Sizes and Defect Classes: How a Lot Passes or Fails

Now the mechanics. When an inspector applies AQL, they are not inventing numbers; they are reading two tables from the standard.

You do not need to memorize them, but you need to understand how they turn your order size into a clear accept/reject line.

There is even a free calculator on most inspection-firm sites, but here is what is happening under the hood.

Step 1: Lot size becomes a code letter

The “lot” is your total order quantity for one product.

The first table maps that lot size, plus your chosen inspection level, to a single code letter.

Almost everyone uses General Inspection Level II, the default for consumer goods.

(Level I gives smaller samples and less confidence; Level III gives larger samples for higher-risk items.)

Step 2: The code letter becomes a sample size and accept/reject numbers

The second table takes your code letter and your AQL and returns two things: how many units to pull (the sample size) and the maximum defects allowed (the acceptance number). Go one over, and the lot is rejected.

Here is a worked example a founder can actually follow:

Worked example: General Inspection Level II, single sampling, normal


Your order (lot) Code letter Units inspected Max major @ 2.5 Max minor @ 4.0
151–280 G 32 2 3
1,201–3,200 K 125 7 10
3,201–10,000 L 200 10 14
10,001–35,000 M 315 14 21

Read the highlighted row. Order 5,000 hoodies. The inspector pulls 200 at random. If they find 10 or fewer major defects, the major batch passes; with 11 or more, the lot is rejected.

Minors run on a separate count; up to 14 pass. Critical defects sit outside this entirely: the acceptance number is 0, so a single critical finding fails the whole lot on its own.

The trap in the tables

The producer’s risk is calibrated near 5% and the consumer’s risk near 10%. In plain terms: sampling is a bet, not a guarantee.

A lot can pass AQL and still contain a few bad units. That is by design. AQL manages risk at a sane cost; it does not promise a flawless order. Set your numbers with that in mind.

Which defects land in which class

The accept/reject count is only useful if the factory and you agree on what counts as major versus minor. Here is how the classes shake out for real garments, drawn from standard industry defect manuals:

 

Garment defect classification: typical examples

Class What it means Streetwear examples
Critical Unsafe, illegal, or unsaleable. Zero tolerance. Broken needle or pin in the garment; wrong fiber content on label; wrong country of origin; missing safety trim.
Major Would cause a customer return; affects fit, function, or durability. Open or broken seam; skipped stitches; wrong size / out of tolerance; print in wrong position; broken zipper; clear off-shade.
Minor Cosmetic: visible to an inspector, rarely to a customer. Small uncut thread; faint spot on an inside panel; slight seam pucker on a non-stress seam; minor untidy stitch.

Borderline cases are where disputes start. A working rule from the floor: judge the flaw on placement, function, repeatability, customer visibility, and repairability.

If it fails even one of those tests, bump it up to major. A slight pucker on a hidden seam is minor; the same pucker across a chest panel is major.

How much size deviation is “out of tolerance” is its own topic; the exact +/- allowances per measurement point live in our dedicated measurement-tolerance guide, and the golden-sample workflow that anchors all of this is covered in the streetwear sampling process guide.

Chapter 4

Who Actually Inspects the Factory: You or a Third Party

A standard is only as good as the person applying it.

Three parties can inspect your order, and each has a different incentive. Understanding those incentives is the difference between real QC and theater.

Option 1: The factory’s own QC team

Every serious factory has in-house quality control.

Ours run continuous inline checks on the floor and a final check before packout.

This is the fastest, cheapest layer because it is built into production, and a good factory’s in-house QC catches the overwhelming majority of issues before anyone else sees them.

The honest caveat: the factory has an incentive to pass its own work.

That is not a reason to distrust a good manufacturer; it is a reason not to make in-house QC your only layer, especially in a new relationship where trust is not yet earned.

Ask any factory how their in-house QC works. A confident answer is itself a green flag.

Option 2: You inspect it yourself

You can pull and check the sample yourself, in person, against your AQL.

It is free, and it teaches you your own product faster than anything else.

The problem is obvious: most founders cannot be on a factory floor in Sialkot or anywhere else, on the right day, mid-production.

Self-inspection works for a local run or a founder who travels. For most, it is aspirational.

Option 3: A third-party inspection firm

Independent inspection companies exist for exactly this gap.

They send a trained inspector to the factory, apply your AQL by the book, and report back usually the same day, with photos of every defect and a clear pass/fail.

Their entire value is that they have no incentive to pass the shipment: no relationship with the factory, no reason to look away.

  • Cost: commonly around US$250–$300 per inspector-day, per product, per factory. For a run that could sink a small brand, that is cheap insurance.

  • Coverage: The large firms operate in 100+ countries and are trained to the same ISO 2859-1 sampling you are now reading about.

  • Output: a documented report with defect counts, root-cause notes, and photos; a corrective-action record you can hold the factory to.

The layered answer

It is not either/or. The strongest setup is layered: the factory’s in-house QC as the constant baseline, plus a third-party check at the line gate or ship gate for any run where a failure would hurt. New factory or large order → book the third party. Trusted partner, small repeat order → in-house QC plus your own spot check may be enough.

Upload Image...

Chapter 5

Setting a QC Standard You Can Enforce

Everything so far is knowledge. This chapter is the action: turning it into a standard that actually binds your factory.

The single most common QC failure among founders is not a bad factory; it is a standard that was never agreed upon in writing before production started.

The failure mode to avoid

The Dispute Trap

You inspect the finished lot, find problems, and try to reject it.

The factory pushes back: “This is normal. You never told us your limit.” And they are right: without an agreed AQL, defect classification, and inspection points on paper before the line ran, you have no standard to enforce.

You are negotiating from zero, holding a paid-for order you cannot return.

Every clause below exists to keep you out of this trap.

The five things your written standard must state

A QC standard does not need to be long. It needs to be specific and agreed before the deposit clears. Put these five into your production agreement or purchase order:

  1. The AQL, per class. Your critical/major/minor numbers (the streetwear baseline of 0 / 2.5 / 4.0 is a fine default) and the inspection level (usually General Level II).

  2. The defect classification. Attach a defect list or reference a standard one so “major” is not up for debate mid-inspection. Big buyers hand factories a defect manual for exactly this reason.

  3. The inspection points. Which of the three gates apply, and who attends each? State whether a third party has floor access.

  4. The reference sample. A sealed golden sample, signed by both sides, is the physical tie-breaker. When a dispute is about “does this match,” the sealed sample answers it.

  5. The remedy. What happens on a failure: rework and re-inspect at the factory’s cost, a price discount for accepted seconds, or full rejection. Spell it out now, not in anger later.

How to actually enforce it

A standard on paper still needs teeth. Three habits give it teeth:

  • Agree before you pay. Leverage exists before the deposit and evaporates after. The standard goes in the agreement, not in an email after the run.

  • Tie payment to the Ship Gate. Final balance is released on a passed final inspection. This is the strongest, cleanest incentive there is.

  • Document everything. Every inspection produces a dated report with photos and defect counts. A paper trail turns “I think it’s fine” into “here is the record,” and it is what a corrective-action conversation is built on.

Start here

If you do only one thing after reading this, write your three AQL numbers, name the three gates you will use, and get both into your next purchase order before you pay the deposit. That single page is a real QC standard. Everything else is refinement.

Quality control FAQs

No and this trips up almost everyone. The number is a level, not a percentage. It points into a standard lookup table that gives you an exact sample size and a maximum defect count for your specific order quantity. The resulting figure rarely equals a clean 2.5%. Treat 2.5 and 4.0 as labels that select a plan, not as fractions of your order.

A sane default is the Streetwear Baseline: 0 for critical, 2.5 for major, 4.0 for minor, inspected at General Level II. Tighten major to 1.5 for premium pieces; only loosen minor toward 6.5 if cosmetic flaws truly don’t matter for that product. Write all three numbers into your order.

Inline or during-production inspection happens while the order is being made (around 20–80% complete), so problems can still be fixed on the line. Final random inspection happens after production is finished and packed (80%+), and produces the accept/reject decision but leaves little room to correct anything. The strongest programs use both.

Ideally, both in layers. A good factory’s in-house QC is your constant baseline and catches most issues. A third-party inspector adds independence for runs where a failure would hurt a new factory or a large order. Third-party inspections commonly cost around US$250–$300 per inspector-day and come with a photo-documented report.

Anything unsafe, illegal, or that makes the product unsaleable: a broken needle or pin in the garment, a sharp trim, the wrong fiber content on a care label, or the wrong country of origin. Critical defects carry an acceptance number of zero; a single one fails the entire lot.

Only if you agree on on a standard first. Without an AQL, a defect classification, and inspection points documented in your agreement before production, the factory can reasonably dispute your rejection. Set the standard and tie final payment to a passed final inspection before the deposit clears.

Founder’s QC glossary


AQL (Acceptance Quality Limit)

The worst quality level you’ll tolerate in a lot before rejecting it; defined in ISO 2859-1 / ANSI/ASQ Z1.4.


Lot

The total order quantity of one product, treated as a single batch for inspection.


Sample size

The number of units an inspector pulls at random from the lot to check.


Code letter

A letter (A–R) that the lot size and inspection level map to; it selects your sample size.


Acceptance number

The maximum defects allowed in the sample before the lot is rejected.


Inspection level

How large a sample to take relative to lot size. General Level II is the consumer-goods default.


Critical / Major / Minor defect

The three severity classes each get their own AQL (typically 0 / 2.5 / 4.0).


Pre-production inspection (Setup Gate)

A check before mass sewing that materials, trims, and the approved sample are correct.


DUPRO / inline inspection (Line Gate)

A during-production check (≈20–80% complete) that catches issues while they can still be fixed.


Final random inspection / FRI (Ship Gate)

The post-production check (80%+, packed) that produces the accept/reject decision.


PP sample

Pre-production sample: the approved reference made before bulk starts.


Golden/sealed sample

A signed, sealed reference garment both sides agree defines “correct.”


Normal/tightened/reduced inspection

Switching modes: scrutiny tightens if quality slips and relaxes if it stays consistently good.


Corrective action report

The documented findings and root causes from an inspection are used to fix and recheck.


Third-party inspection

An independent inspector with no incentive to pass the shipment, applying your AQL for you.