Garment Quality Control for Small Orders (50–500 Pieces): A Buyer’s Guide

A 50-piece order and a 5,000-piece order do not fail in the same way. When a factory runs 5,000 pieces, one systematic error — a wrong shade, a mis-set pattern, a faulty zip — can be caught at the first inspection and corrected before most of the order is sewn. When the order is 50 pieces, that same error can be most of the order before anyone notices. Small orders are not simply scaled-down big orders. They need different quality control: decisions made before sewing starts, a physical reference sample both sides have signed, and inspection rules written into the purchase order rather than assumed.

This guide is written for the buyer: the small brand, the emerging label, or the sourcing manager placing a first or small order. It covers the four gates that protect a small lot — incoming fabric, the pre-production sample, inline inspection, and final inspection — and it ends with a one-page QC clause you can paste into a purchase order. Inspection is not the same as laboratory testing; if you need that distinction, our guide on garment inspection versus testing sets it out.

Why small orders need different QC

Sampling is statistics. An AQL inspection judges a lot from a sample: it tells you whether the lot is statistically acceptable, not whether every unchecked piece is clean. A passed lot can still contain defective units, because AQL controls the average, not each piece. That logic works reasonably well at volume, where a sample of 200 pieces represents a large order. At 50 pieces, the same logic breaks down: the sample is a large share of the order, and the cost of checking every piece is small — as an illustration, a visual and measurement check of 100 simple garments is typically a matter of hours rather than days. That is an illustration, not a standard time: the actual time depends on garment complexity, inspection method, reporting requirements and packing condition.

The second difference is leverage. On a big order, the buyer can afford a dedicated QC team, a full inspection plan, and several rounds of correction. On a small order, there is usually no QC department, no second chance, and very thin tolerance for error — the margin that absorbs a remake on a big order does not exist. That is why the decisions that matter most for a small order are made before production: the sealed sample, the written tolerances, and the sampling rule.

Our guide to garment inspection plans covers the general framework; this article applies it to the 50–500 piece range, where the numbers and the judgement calls are different.

The four gates: fabric, PP sample, inline, final

Think of small-order QC as four gates. Each gate protects something the later gates cannot fix.

If a gate is skipped, the cost of the failure moves to the next gate — and usually grows. Fabric defects found at final inspection mean re-cutting finished garments; a shade problem found after packing means opening cartons. The cheapest place to find every one of these problems is the gate designed for it. A note on timing: the 20–80% inline window and the 100%-produced / 80%-packed final trigger below are suggested buyer-defined checkpoints used by inspection companies, not requirements in any standard — agree the exact timing in your inspection plan, reflecting order size, production risk and product complexity.

Gate 1: incoming fabric and the four-point system (ASTM D5430)

Small lots are often cut from a small number of fabric rolls, so one defective roll can be a large share of the order. Incoming fabric inspection is therefore proportionately more important on a small order, not less. The standard method is the four-point system, described in ASTM D5430, Standard Test Methods for Visually Inspecting and Grading Fabrics. The current edition is D5430-26. The standard’s published Scope describes how to establish a numerical grade from visual inspection; it may be used for delivery or acceptance of fabrics where the requirements are mutually agreed by purchaser and supplier. What the standard’s text itself says about scoring is limited to what its published Scope discloses — notably §5.2, which warns that penalty points ‘may vary considerably when using each of the options listed herein’. The scoring numbers in this section do not come from the D5430-26 text, which is a paid standard and was not directly consulted. They come from one published version of the scoring — Clearload’s four-point table (14 October 2026), which matches Textile School and Ruhrose — and are reproduced here as that published example, not as the standard’s text. Name the edition and the scoring table in your contract and confirm the numbers against it; if your contract names a different table, that table governs.

In Clearload’s published version of the scoring, the thresholds are straightforward: a defect up to 3 inches scores 1 point, over 3 to 6 inches scores 2, over 6 to 9 inches scores 3, and over 9 inches scores 4. No single defect scores more than 4 points, and no linear yard scores more than 4 points. A continuous defect scores 4 points for each yard it runs. One point of variation between published versions of the method: hole scoring. Clearload and Textile School score holes of 1 inch or less at 2 points and larger holes at 4 points; ForAllTex, Simibag and Novweave score any hole at 4 points regardless of size. As a default for the purchase order, use the Clearload/Textile School hole rule (holes of 1 inch or less = 2 points, larger holes = 4 points); if your supplier or inspector uses the any-hole-4-points version, name that version in the contract instead. Either way, the contract must name which option applies — the standard’s own Scope (§5.2) warns that penalty points ‘may vary considerably when using each of the options listed herein’ and that the same option should be used in cases of disagreement. The score is then expressed per 100 square yards:

Points per 100 sq yd = total points × 3,600 ÷ (yards inspected × fabric width in inches)

Here is the point most buyers get wrong: ASTM D5430 sets the method, not the pass mark. The figure most often quoted — 40 points per 100 square yards or fewer — is a trade convention and a contract term, not a value in the standard. Other sources quote different limits, including 25 points and a 28-point shipment average. That disagreement is itself the lesson: if you do not write your own limit into the purchase order, you have not agreed one. Write the number down, name the edition of the standard, and state whether the limit applies per roll, per shipment, or per fabric lot.

For the full method, including worked examples, see our guide to the four-point system for fabric inspection, and our types of fabric defects for what the defects actually look like.

Gate 2: the PP sample and the sealed (gold seal) sample

The pre-production (PP) sample is the single most important document in a small order — and it is not a document at all. It is a physical garment, made with the actual bulk fabric, the actual trims, the actual labels, and the actual packing method. Nothing substituted. It is normally made in two or three sets: one for the buyer, one for the factory sample room, and one for the QC or buying office.

Two rules follow from that definition. First, bulk cutting cannot start without written PP approval. Cutting fabric before approval converts a correctable sample problem into a finished-goods problem. Second, the approved PP sample is sealed and signed — the “gold seal” sample — and becomes the reference standard for the whole order. In any later dispute about shade, measurement, workmanship or trims, the sealed sample decides who is right. That is why it is sealed: so that neither side can quietly substitute a better or worse example after the fact.

Photographs cannot do this job. A product image cannot confirm actual measurements, tolerances, seam allowances, internal finishing or approval status. Photo approval is useful for confirming a style detail; it is not a substitute for a physical sample that both sides have held, measured and signed. Buyers who approve by photograph are, in effect, approving a picture of a garment rather than a garment.

What should be sealed? The garment itself, in the approved shade and wash; the fabric swatch; the trims (buttons, zips, labels, hangtags); and the packing method, including the polybag, the fold and the carton marking. If a trim changes after sealing, the seal is broken and a new approval is needed. Our guide to garment pre-production approval covers the approval meeting itself; this article covers what the approval must physically consist of.

Measurement: POMs, tolerances and the size set

A specification sheet is only as good as its points of measure (POMs). A POM is one defined measurement with named landmarks and a repeatable method — for example, “half chest, measured straight across 2 cm below the armhole, edge to edge” — not just “chest”. The spec sheet should list each POM with its how-to-measure definition, the finished measurement for each size, the base size, the grade between sizes, and the tolerance for that POM.

Tolerances are where small-order buyers most often accept numbers they have not read. There is no ISO or ASTM table of garment tolerances to fall back on: the numbers are trade practice, agreed per order. As a typical range, organised by point group so that no point appears in two categories at once:

These are typical figures from trade guides, not a standard — state your own per point, in writing. One design guideline is worth applying when you set them: keep each tolerance below half its grade increment, or adjacent sizes can measure the same. If the grade between sizes is 2 cm and the tolerance is ±1.5 cm, a size M can measure the same as a size L. That is a specification-design guideline, not a universal technical requirement and not a contractual requirement unless the parties expressly agree it — grading, fit intent and measurement methodology all affect it — but a specification that violates it has a built-in error, and it is the buyer’s error.

Finally, approve a physical size set — not just the sample size. The size set should cover the smallest, the base and the largest production sizes, with intermediate sizes added when the range is broad or the silhouette is sensitive. Grading rules should be approved as a complete size progression, not accepted as automatic software output. See our guides on garment measurement tolerances and the apparel size chart buyer specification guide for the detail.

Shade control before the cutting table

Shade variation is the defect most likely to ruin a small lot at the cutting table, because a small lot is often cut from few rolls and there is no spare fabric to re-cut a shade-off panel. The control is procedural, and it happens before spreading.

Shade is segregated against buyer-approved swatches. In documented cutting-room practice — for example, the published cutting standard operating procedure of Azim Group (2020) — shade bands are labelled A, B and C (or D) and recorded per roll, with a sticker on both ends of every roll giving the shade number, roll number and width, and only rolls of one shade and one width are collected for a lay at a time, with separator (lay) sheets between different shades in the same lay. Other factories document the same practice differently, and shade grouping is ultimately a matter for the buyer’s requirements and the cutting room’s procedures — but whatever the local form, the discipline is the same: segregate before spreading, and record the grouping so that it can be checked. Lay-to-lay variation and centre-to-selvedge variation are checked before cutting starts. If two rolls are different shades, they are not mixed in one lay in the hope that nobody notices; they are cut as separate shade groups, and the garments are grouped accordingly.

On a 50-piece order cut from two rolls, a shade mismatch discovered after sewing is a 50-piece problem. Discovered at the cutting table, it is a re-lay. The difference is entirely a matter of when the check happens. See our guides on fabric colour approval and fabric spreading modes for the cutting-room detail.

Inline inspection: DHU on the sewing line

Inline inspection — during production, or DUPRO — is typically scheduled when roughly 20–80% of the order is complete; many inspectors prefer the earlier part of that window, for example 20–50%, because the earlier the check, the more of the order can still be corrected. Its purpose is to catch systematic errors while they can still be corrected: a wrong stitch density, a recurring seam pucker, a mis-set buttonhole. Final inspection cannot do this; by then the error is in every piece. The 20–80% window is a suggested checkpoint used by inspection companies, not a requirement in any standard — agree the timing in your inspection plan, reflecting order size and production risk.

The metric to track inline is DHU — defects per hundred units:

DHU = (total defects found ÷ total garments inspected) × 100

DHU counts every defect, so one garment can contribute several. That makes DHU different from defective percentage, which counts defective pieces: defective % = (defective pieces ÷ pieces checked) × 100. A worked example makes the difference concrete: 400 pieces checked, 22 defects found on 18 pieces — DHU is 5.5, defective percentage is 4.5%. DHU is always equal to or higher than defective percentage, and it can exceed 100. For example, 35 defects found in 250 garments gives a DHU of 14.

Track DHU hourly or daily on the line, not just at the end. A rising DHU is an early warning; a flat DHU after a corrective action is evidence the correction worked. One caution: do not treat any single DHU number as a universal pass mark. A figure of “2–3% acceptable DHU” appears in one published case study — a 2025 Bangladesh factory study of a short-pants line, where the line improved to 3.26% against a buyer requirement in that context. It is a case-study figure, not an industry benchmark, and this article does not present it as one. Your DHU target should be agreed with your supplier for your product, and recorded.

Final inspection: AQL sampling for small lots

Final inspection takes place when 100% of the order is produced and at least 80% is packed and cartonned. The standard framework is ISO 2859-1, Sampling procedures for inspection by attributes — Part 1: Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection. The current edition is ISO 2859-1:2026, the third edition, published in January 2026, replacing the 1999 edition and adding skip-lot sampling. The published list of main changes is not an exhaustive statement of every change in the edition. General Inspection Level II is the normal default buyers use for garment inspection — a buyer convention, not a requirement of the standard: the lot size and the inspection level give a code letter, and the code letter gives the sample size.

AQL — acceptance quality limit — is the worst tolerable process average of defective items, used to set sample sizes and accept/reject numbers. It is a scheme-indexing parameter: sampling plans are designed so that a process running at the AQL is accepted most of the time — conventionally about 95% of the time, a producer’s risk of about 5% — but the exact figure depends on the plan and its operating-characteristic curve. ReliabilityCalc’s published calculation for the code J / AQL 2.5 plan (Table II-A, normal inspection: sample 80 pieces, accept 5 / reject 6) puts acceptance at about 98% of lots when the process really runs at 2.5% defective, and about 86% under tightened inspection at the same defect rate. These are calculator estimates from ReliabilityCalc’s published tables, not values read from the standard’s operating-characteristic curves. AQL is not a guarantee that any particular lot will be accepted.

The AQL numbers themselves — commonly 2.5 for major defects and 4.0 for minor defects, with 0 for critical defects — are buyer-selected defaults, not values ISO 2859-1 mandates for apparel. The standard defines the scheme; the numbers are whatever buyer and supplier agree and write into the purchase order. Each defect class (critical, major, minor) is judged separately against its own accept/reject numbers, and a passed lot can still contain defective units, because AQL controls the average, not each piece. For the deeper explanation, see our guides on garment AQL inspection limits and garment defect classification.

Here is the small-lot table, at General Inspection Level II, normal inspection, single sampling. Read the label carefully: this is the long-published ISO 2859-1:1999 / ANSI/ASQ Z1.4 single-sampling table. The current edition, ISO 2859-1:2026, is the third edition, published in January 2026; its published list of main changes covers skip-lot sampling procedures and the removal of the operating-characteristic curves, and it does not state whether the sampling plans themselves changed. No source available to this publication confirms the plans are identical, so the table below is labelled as the 1999 / Z1.4 table and is not presented as the 2026 table. To be explicit: this table is the 1999 edition, and the 2026 plans have not been confirmed as identical — name the edition and year in your contract and confirm the numbers against the current tables.

* The asterisk marks an arrow move. In the standard’s single-sampling table, some cells contain an arrow instead of numbers: the arrow points to the first sampling plan in that column in the direction shown, and its sample size replaces the code letter’s sample size. For a 26–50 lot, the AQL 2.5 plan moves up to 5 pieces (accept 0, reject 1) and the AQL 4.0 plan moves down to 13 pieces (accept 1, reject 2). For a 51–90 lot, the AQL 2.5 plan moves down to 20 pieces (accept 1, reject 2). The code letter still gives the starting point — but the arrow, not the code letter, gives the sample actually inspected. The table above is the long-published ISO 2859-1:1999 / ANSI/ASQ Z1.4 table, not the ISO 2859-1:2026 table: the 2026 text is sold by ISO, and at least one secondary source notes explicitly that it has not compared these tables line by line with the 2026 text. Name the edition and year in your contract and confirm the numbers against the current tables — do not assume a table found online matches the edition you have named. The entries above reproduce the single-sampling plans of ISO 2859-1:1999 Table 2-A (ANSI/ASQ Z1.4 Table II-A) as published in Clearload’s ‘AQL sampling explained’ (Table 2) and Ruhrose’s AQL table; the arrow moves follow the standard’s arrow rule as explained in those guides.

Read the table as a decision tool, not a ritual. For a 120-piece order, the lot falls in the 91–150 band: code F, 20 pieces inspected, and at AQL 2.5 the lot is accepted if no more than one major defect is found, rejected if two or more are found. That is a precise, checkable instruction — and it is only useful if it was written into the purchase order before production started. For the full analysis of when AQL 2.5 works and when it does not, see our article on how AQL 2.5 works for garment orders and when it doesn’t.

AQL or 100% inspection? How to decide

For a small order, the honest answer is often: check every piece. Sampling is a statistical judgement about the lot; 100% inspection checks every piece and ships only pieces that were actually looked at. When the AQL sample size equals or exceeds the lot size, sampling gives no practical efficiency advantage at all — there is nothing to save by sampling. As an illustration, a visual and measurement check of 100 simple garments is typically a matter of hours for an inspector, and the cost of that time is small against the cost of a customer complaint, a return, or a chargeback on the whole order. The exact time depends on garment complexity, inspection method, reporting and packing, so treat the illustration as such — not as a standard time.

Use this framework, and write the choice into the purchase order:

• 100% inspection when the lot is small (roughly the lower half of the 50–500 range), the unit value is high, the construction is complex, it is a first order with a new factory, or the product carries a safety or regulatory risk. The cost is inspector time; the benefit is that no unchecked piece ships.
• AQL sampling plus inline checks when the lot is larger (roughly 300–500 pieces), the factory is proven, the product is simple and stable, and inline DHU has been running clean. The sampling decision must still be written down — code letter, sample size, and accept/reject numbers for each defect class.
• Never sample by habit. If the purchase order does not name the sampling plan, the inspection company will use its own default, and the buyer will discover what that default was only when a dispute arises.

The break-even point between the two approaches is a matter of judgement, not a sourced threshold, and this article does not pretend otherwise. What matters is that the decision is made deliberately, before production, and recorded.

Worked example — a 120-piece order. Lot size 120 falls in the 91–150 band: code F, sample 20, AQL 2.5 major accept 1 / reject 2, AQL 4.0 minor accept 2 / reject 3, critical 0. But the buyer in this example is a new brand, first order with this factory, mid-value woven shirts. The sensible instruction in the purchase order is: 100% inspection, with the AQL numbers retained as the defect-classification standard for deciding what counts as major and minor. The cost is an inspector’s morning, as an illustration for a simple garment; the benefit is that every shirt shipped has been looked at. That is the right trade for a first order.

The one-page QC clause: copy this into your purchase order

Most guides explain AQL, DHU or the four-point system in isolation and leave the buyer to assemble the pieces. The practical takeaway of this article is a single clause that assembles them. It is editorial guidance, built from the facts above — not a standard, and not legal advice. Fill in the blanks for your order, and have both sides sign it.

Quality clause — small order (50–500 pieces). (1) Sealed sample: the supplier shall produce a pre-production sample in bulk fabric, bulk trims, bulk labels and bulk packing; no bulk cutting shall start before the buyer’s written approval; the approved sample shall be sealed and signed by both parties, one copy held by each side, and shall be the reference for shade, measurement, workmanship and trims. (2) Tolerances: measurements shall follow the attached specification sheet, with a named point of measure, a finished measurement per size, and a tolerance per point as listed; tolerances shall be reviewed against grade increments and fit intent, and any rule that a tolerance must be less than half its grade increment shall apply only if expressly agreed by both parties. (3) Sampling: lots of [___] pieces and below shall be 100% inspected; larger lots shall be inspected at General Inspection Level II per ISO 2859-1:[year], AQL [___] major / [___] minor / 0 critical, with the code letter, sample size and accept/reject numbers stated in the inspection instruction. (4) Gates: incoming fabric shall be graded per ASTM D5430-[year] with a buyer limit of [___] points per 100 square yards; inline (DUPRO) inspection shall take place at about [___]% completion with DHU reported per production day; final inspection shall take place when 100% is produced and at least 80% is packed. (5) Reporting: the inspector shall report DHU, the defect list by class, and the measurement results for the size set (smallest, base and largest sizes) — not only a pass/fail.

Five blanks, one page, and every argument that usually happens after delivery has already been settled before cutting.

Common small-order QC mistakes

• Approving by photograph. A photo cannot confirm measurements, tolerances or construction. Approve the physical sample, in hand.
• Accepting AQL numbers without writing them down. “AQL 2.5” means nothing until the code letter, the sample size, the defect classes and the edition of the standard are all named in the purchase order.
• Checking only the sample size. A size M that fits does not prove that XS and XXL grade correctly. Approve the size set.
• Skipping inline inspection. A systematic error found at final inspection is already in every piece. DUPRO, typically at 20–80% completion, is the cheapest correction point in the whole order.
• Treating a passed lot as defect-free. AQL controls the average. A passed lot can still contain defective units; that is what the accept numbers mean.
• Treating a quoted limit as a standard. The 40-point four-point limit, the 2.5/4.0 AQL pair and the tolerance figures in this article are conventions and typical practice, agreed per order. None of them is an ISO or ASTM requirement, and none of them should be cited as one.

Frequently asked questions

What AQL should I use for a 100-piece order?
At 100 pieces, the lot falls in the 91–150 band: code F, sample 20, AQL 2.5 major accept 1 / reject 2, AQL 4.0 minor accept 2 / reject 3. But for a first order or a high-value product, consider 100% inspection instead — as an illustration, a visual and measurement check of 100 simple garments is typically a matter of hours for an inspector, and every piece shipped has then been looked at.

Who is responsible for inspection — the buyer, the factory, or a third party?
Whoever the purchase order says. In practice the factory runs its own inline checks, the buyer (or the buyer’s nominated third-party inspector) runs the DUPRO and final inspections, and the purchase order should name each party, the inspection gates, and who pays for reinspection after a failure. An inspection that nobody owns in writing is an inspection that does not happen.

What happens if a lot fails inspection?
The lot is held, not shipped. The usual sequence is: the inspector reports the defect list by class; the factory sorts or reworks the affected pieces (or the whole lot, if the failure is systematic); and the lot is reinspected — normally a fresh sample under the same plan, not a re-count of the same pieces. The purchase order should state who pays for the rework and the reinspection, and how many reinspections are allowed before the order is cancelled or renegotiated. None of that is in the AQL standard; it is contract drafting.

What is a sealed sample?
The finally approved pre-production sample, sealed with a tag, stamp or seal and signed by the buyer or the nominated QC. It is the reference standard in later disputes about shade, measurement, workmanship or trims. Both sides hold a copy; neither side can substitute a different example afterwards.

What does DHU tell me that defective percentage doesn’t?
DHU counts every defect; defective percentage counts defective pieces. A garment with three defects contributes three to DHU but one to defective percentage. DHU therefore complements defective percentage: DHU shows the defect load on the line, defective percentage shows the share of affected garments, and neither alone fully describes quality performance — use both, alongside defect classification and corrective-action tracking.

Does ASTM D5430 set a pass limit for fabric?
No. ASTM D5430 sets the four-point method — how defects are scored and how the score is calculated. The pass limit, commonly quoted as 40 points per 100 square yards, is a trade convention and a contract term. Other sources quote different limits, so the buyer must write their own number into the purchase order.

Sources

• ISO 2859-1:2026, Sampling procedures for inspection by attributes — Part 1: Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection (3rd edition, January 2026) — catalogue page: iso.org/standard/2859-1
• ANSI, “ISO 2859-1:2026 — AQL sampling”: blog.ansi.org; preview of the standard: webstore.ansi.org
• ASTM D5430-26, Standard Test Methods for Visually Inspecting and Grading Fabrics — store page: store.astm.org/d5430-26.html
• Clearload, “Four-point fabric inspection system” (14 October 2026): cleartoload.com; and “AQL sampling explained” (Table 2 — single-sampling plans, ISO 2859-1 Table 2-A / ANSI/ASQ Z1.4 Table II-A): cleartoload.com
• Ruhrose, “Garment quality control — AQL inspection” and “AQL table — sample sizes guide” (sample sizes and accept/reject numbers as published): ruhrose.com
• AQI Service, “Acceptable Quality Level (AQL) in garment inspection”: aqiservice.com
• Textile School, “Fabric inspection and terms” (updated 17 October 2026): textileschool.com
• Online Clothing Study, “What is DHU? How to measure DHU” (updated 29 September 2026): onlineclothingstudy.com
• Textile Notes, “DHU” (updated 19 October 2026): textilenotes.com
• CITCE ERP, “Reduce DHU on the sewing line” (1 November 2026): citce-erp.com
• Bangladesh University of Fashion & Technology, “Defects per hundred units (DHU) in a short-pants line” (BJFT 2025, Vol. 9, pp. 50–65): buft.edu.bd
• Textile Merchandising, “Types of samples in the garment industry — 10 samples explained”: textilemerchandising.com
• Apparel Entrepreneurship, “Tech pack measurements” (updated 2 November 2026): apparelentrepreneurship.com
• TradeAider, “AQL sampling vs 100% full inspection”: tradeaiders.com
• Academia.edu, Azim Group, “Standard Operating Procedure of Cutting” (1 January 2020): academia.edu
• ECQA, “During Production vs Pre-Shipment Inspection” (updated 16 June 2026): ecqa.com
• ReliabilityCalc, “AQL — acceptance sampling plans” (Table II-A, code J / AQL 2.5: sample 80, accept 5 / reject 6; about 98% acceptance at 2.5% defective under normal inspection, about 86% under tightened): reliabilitycalc.com

All facts in this guide were checked against the sources above on 11 November 2026. Where a figure is trade practice or a contract convention rather than a standard requirement — notably the AQL 2.5/4.0 defaults, the 40-point four-point limit, the tolerance ranges and the DHU case-study figure — the text says so. The small-lot AQL table is the long-published ISO 2859-1:1999 / ANSI/ASQ Z1.4 single-sampling table, not the ISO 2859-1:2026 table: the 2026 text is sold by ISO, its published list of main changes does not state whether the sampling plans changed, and no source available to this publication confirms the plans are identical. Readers should confirm the numbers against the edition named in their contract. The four-point hole-scoring rule varies between published versions of the method; the contract must name the option used.