Measurement Tolerance in Bulk Linen: What AQL 2.5 Actually Allows

Measurement Tolerance in Bulk Linen: What AQL 2.5 Actually Allows

AQL 2.5 does not give you a measurement tolerance. It gives you an acceptance number. Those are two different numbers, and confusing them is the single most common reason a linen bulk order gets rejected at final inspection even though the factory "followed the spec".

In practice, a bulk linen order passes at AQL 2.5 when the number of defective units in the sample stays at or below the acceptance number for that lot size. A unit becomes defective because it failed a measurement point that you defined, against a tolerance band that you set. If your spec sheet says "chest 56 cm" and nothing else, no inspector in the world can make a defensible pass or fail decision.

This guide translates the two systems into each other: which measurement points deserve which band, how linen's shrinkage and relaxation behaviour shifts those bands, and what a verdict looks like when you apply the numbers to real inspection sheets. If you are still at sample stage, pair this with our 12-point fit testing checklist before bulk cutting begins.

What AQL 2.5 Actually Governs

Acceptable Quality Limit 2.5 is a sampling standard, not a dimensional standard. It tells you how many defective units you are willing to accept in a shipment, expressed as a percentage of defective units in a sample drawn from the lot. Under ISO 2859-1 at general inspection level II, AQL 2.5 produces a fixed acceptance number for each lot size range. For a lot of 501 to 1,200 units, you inspect 80 pieces and accept the lot at 5 defects or fewer, rejecting at 6.

Nothing in that sentence mentions centimetres. The standard is silent on how far a chest measurement may drift. That silence is where disputes are born. AQL tells the inspector how strict to be about counting; your specification tells the inspector what counts as a defect in the first place.

This is why two buyers can run the same AQL 2.5 inspection on the same lot and reach opposite conclusions. One has a spec sheet with a tolerance table and class assignments. The other has a tech pack with nominal measurements only. The first gets a defensible verdict and a clear rework instruction. The second gets an argument.

There is a second layer that buyers often miss. Measurement defects and visual defects are usually counted in the same inspection, but they behave very differently. A colour shading issue on one panel is a single visible defect. A chest that measures 2 cm over nominal is a defect that appears on every unit cut from the same pattern block, same marker, and same operator set. Measurement errors cluster. Visual errors scatter. Treating them with one undifferentiated AQL bucket hides the clustering, which is exactly the signal you need for corrective action.

Where Measurement Drift Comes From in Linen

Linen is a dimensionally restless fibre. Before you can set a tolerance, you need to know which process step is eating the millimetres, because the fix differs by cause. In a linen garment factory, measurement drift almost always traces to one of five places.

• Relaxation after cutting. Fabric is wound under tension. Once cut and left to rest, it recovers. A panel cut at 60 cm can settle to 59.3 cm within 24 hours. If you measure immediately after cutting, you are measuring a stressed state, not the final state.
• First-wash shrinkage. Linen commonly moves 3% to 7% on the first wash depending on weave, yarn count, and finish. A garment-washed order and a piece-dyed order do not share a tolerance band, because one has already paid the shrinkage cost and the other has not.
• Stitching and seam take-up. A 1 cm seam allowance consumed slightly unevenly across a long shoulder seam produces a 3 to 5 mm shift at the armhole end. On lightweight linen, seam grin and take-up compound this.
• Pressing and steaming. Final pressing can add or remove 5 to 8 mm on a body length depending on how much steam and stretch the operator applies. Pressing is the least documented and most variable step in the chain.
• Measurement method itself. Flat vs hanging, garment closed vs open, tape tension, and whether you measure to the seam line or the seam edge. Two inspectors measuring the same garment can differ by 5 mm purely on method.

The last point deserves emphasis because it is free to fix. A tolerance table without a measurement method is only half a specification. Write down whether the garment is measured flat, closed, and relaxed, and whether points are taken seam-to-seam. Our guide to pre-shrunk and garment-washed linen covers how to lock the shrinkage variable before it reaches the cutting table.

Fit testing a sample before bulk gives you the empirical drift for that specific fabric and construction. Without it, you are setting tolerances from a template, which is better than nothing but will be wrong in a predictable direction: too tight on length, too loose on width, for garment-washed linen.

A Working Tolerance Table by Measurement Point

The table below is the band structure we apply to linen woven tops, dresses, and light outerwear in bulk production. It is a starting point, calibrated to a 100–260 GSM linen or linen-blend woven, measured flat after relaxation, seam to seam. Adjust the numbers to your construction, but keep the three-tier logic.

Two things about this table are deliberate. First, the bands are asymmetric in strictness, not in direction: tight control on short dimensions, looser control on long ones. A 0.5 cm error on a 14 cm cuff height is a 3.6% deviation and reads as wrong; the same 0.5 cm on a 110 cm body length is under half a percent and reads as normal. Second, every point has a class. A tolerance without a class cannot be counted as major or minor, which means it cannot be fed into the AQL arithmetic at all.

Assigning Points to Critical, Major, and Minor

The class assignment is a commercial decision, not a technical one, and it should be made by the buyer before the first sample is approved. The working definitions we use are below, and they map cleanly onto how an inspector fills in the defect columns.

• Critical. A deviation that compromises safety, regulatory compliance, or the garment's ability to function as designed. Broken needle parts, restricted substances, a placket that will not close. Measurement points land here when the error propagates into construction: shoulder, armhole, neck, sleeve length.
• Major. A deviation that is visible to a consumer or that changes the fit experience, but does not create a safety or function failure. Chest, waist, hip, body length, pocket placement. Most measurement defects live here.
• Minor. A deviation noticeable only under measurement, with no effect on wear or appearance. Hem width on a flared shape, a 2 mm difference in a hem turn-back.

Class assignment also determines your remedy. A critical measurement defect cannot be conceded by a discount; it either gets reworked or the unit is rejected. A major can sometimes be conceded with a price adjustment if the buyer agrees in writing. A minor is usually absorbed. Decide this in advance and put it in the purchase order, because negotiating remedy class by class during a dispute is expensive for both sides.

Turning Measurements into a Verdict: The Sampling Math

Here is where the two systems meet. You inspect a sample, you count defective units, and you compare the count to the acceptance number. A unit is defective if it fails any critical point, or accumulates enough major points to cross your threshold. The standard single-sampling values at AQL 2.5, general inspection level II, are below.

Note how unforgiving the step from accept to reject is: at the 501–1,200 lot size, five defective units in eighty passes, six fails. There is no grey zone, no "conditional pass". This is why the definition of a defective unit has to be unambiguous before inspection begins. If your spec lets an inspector debate whether a 1.2 cm chest deviation counts, you have handed the verdict to whoever argues harder.

Worked example. A 900-unit linen shirt order, sample of 80. The inspector measures all 80 against your table. Results: three units have chest width over the ±1.0 cm band, one unit has a sleeve length 0.7 cm short, and two units have pocket placement off by 0.6 cm. That is six defective units against an acceptance number of five. The lot is rejected at AQL 2.5, even though the defect rate is only 7.5% and the individual deviations are small. The correct response is a 100% measurement sort on the chest and sleeve points, not a discount negotiation.

This clustering is the useful part. Because four of the six failures are chest or sleeve, the cause is almost certainly one pattern block or one cutting marker, not operator variance. A single rework pass on those two points fixes the lot. If the six failures had been spread across six different points, you would be looking at a systemic control problem instead. Our guide to linen garment quality inspection covers the sampling mechanics in more depth; this article stays on the dimensional side.

Three Worked Inspection Verdicts

Abstract numbers become clear when you run them through a real decision. The three cases below use the 501–1,200 lot size, sample 80, accept 5, reject 6.

Case 1: Clean pass with a minor drift

An 800-unit linen dress order. Forty of the 80 sampled pieces measure 0.4 cm short on body length, all in the same direction. Two pieces have a 1.2 cm hem width deviation. No critical failures. Total defective units: two. Verdict: pass. The body-length drift is within ±1.0 cm and is consistent, which is acceptable; the two hem deviations are minor and counted. Action: none required, but log the directional drift so the next order's pattern can be corrected by 0.4 cm.

Case 2: Reject on a single clustered point

A 1,000-unit linen blouse order. Seven units have shoulder width 0.8 cm over band. Everything else is clean. Total defective units: seven against accept 5. Verdict: reject. Because the failures cluster on one critical point, the fix is scoped: re-measure all units on shoulder, rework or reject the outliers, re-inspect a reduced sample. Note that this rejects on seven units out of eighty when 73 pieces were perfect, which feels harsh but is exactly what the standard prescribes.

Case 3: Borderline that should have been prevented

A 600-unit linen tunic order. Four units fail chest width by 1.1 cm, one unit fails armhole by 0.6 cm, one unit has a button spacing drift of 0.4 cm. Total: six, reject. Two of the six are marginal — 1.1 cm on a ±1.0 cm band is a 1 mm overage. This is the case that generates the argument. The buyer feels a 1 mm overage should not reject a lot; the factory feels the spec was clear. Both are right, and the failure happened months earlier, when nobody specified measurement method or rounding rules. Round to the nearest millimetre, state the rounding in the spec, and this case disappears.

Writing Tolerances into a Spec Sheet

A tolerance table is only usable if the spec sheet states the conditions under which it applies. Six items belong on every linen measurement page.

1. Measurement state. Flat, closed, relaxed for at least 24 hours after pressing, seam to seam. State it in one line at the top of the measurement page.
2. Nominal values with units. Centimetres or inches, not both, and specify which side of the seam.
3. Per-point band. A number for every point, not a blanket "±1 cm" for the whole sheet.
4. Class per point. Critical, major, or minor, mapped to your remedy rules.
5. Rounding rule. Measure to the nearest millimetre and round half up. This single line prevents most marginal disputes.
6. Shrinkage allowance basis. Whether nominal values are pre-wash or post-wash, and the accepted shrinkage range. For linen, 3% to 7% on first wash is the normal band; anything outside that needs a fabric conversation.

Size grading interacts with all six. If your grade rules are built for cotton and applied to linen, the tolerance bands will fight the grade increments. Our article on size grading for linen explains why natural-fibre garments need their own grade rules rather than borrowed ones.

When Linen Legitimately Exceeds Tolerance

Some out-of-band readings are not defects. Knowing which ones protects you from rejecting good goods and protects the factory from unfair claims.

• Unwashed vs washed state. A garment measured before garment washing and again after will differ by the shrinkage percentage. If the spec was written pre-wash and the inspection happens post-wash, the numbers will not match and the garment is not at fault.
• Relaxation window. Measuring within hours of pressing catches the fabric mid-recovery. Readings taken the same day can be 5 mm off the settled value.
• Natural-fibre variance within a roll. Linen yarns vary in count along the length. Two panels cut from different parts of a roll can behave differently under the same pattern. This is a fabric property, not a sewing error, and the mitigation is marker planning, not tighter sewing tolerance.
• Deliberate ease in oversized shapes. On a boxy or oversized silhouette, a ±1.5 cm band on chest and hem is commercially correct. Applying a fitted-garment band to an oversized garment creates false failures.

The honest limitation is this: a tolerance table cannot tell you whether a garment feels right. It tells you whether the garment matches the approved sample within a defined band. Fit judgement happens at sample stage, which is why the sample approval is the most consequential decision in the whole order and why the sampling window of 7 to 14 days should not be compressed to save a week.

When a deviation sits outside the band and outside the legitimate causes above, you have three options: rework to spec, concede the units with a written price adjustment, or reject the units. Whichever you choose, record the decision against the inspection report so the next order starts from evidence rather than memory.

Questions Buyers Ask About Linen Tolerance

Getting the Numbers Right Before Cutting

Measurement tolerance in bulk linen is a three-part agreement: a band per point, a class per point, and a stated measurement method. AQL 2.5 then does its job, which is counting defective units against an acceptance number. Neither part works alone. A tolerance table without an inspection standard gives you no verdict; an inspection standard without a tolerance table gives you no definition.

The practical sequence that avoids most disputes is short. Approve the sample with measured values recorded, not just a signed photo. Build the tolerance table from the sample's actual drift rather than a template. State method, rounding, and class on the measurement page. Run the final inspection against that page, and log directional drift for the next order's pattern correction.

If you are setting up a linen programme and want the bands reviewed against your construction before cutting, our team can work through the measurement page with you. Start with the contact page and include your tech pack, target market, and whether the order is garment-washed or piece-dyed — those three details change the recommended bands materially.

Linenwind is a Dongguan-based B2B linen garment manufacturer working with fashion brands on OEM, ODM, and private label programmes. MOQ is 60 pieces per style per colour, sampling runs 7 to 12 days, and bulk production 25 to 35 days with a 35 to 60 day first-order window. Fabric is held in 50+ in-stock specifications at 100 to 260 GSM, inspected to AQL 2.5, under ISO 9001, OEKO-TEX Standard 100, and SEDEX.