An approved sample proves a factory can make one good piece — not eight hundred. Bulk quality is not held by goodwill or by a promise in the quotation. It is held by three checkpoints, one statistical sampling standard, and three documents that nobody is allowed to change by phone call. Here is the system we run on every order, and how to audit it from the buyer's side.
Inspection only works if the standard was agreed early — the spec sheet defines what is measured, and sailing dates define when.
Why Samples Pass and Bulk Fails
Almost every "the bulk is not like the sample" complaint traces back to one of three causes, none of which is a bad sewer:
- Operator drift. The sample is made by the sample room — the most skilled team in the factory. Bulk runs on a line where the operator changes after lunch.
- Fabric lot variation. The sample used 3 metres from a hand-picked roll. Bulk uses 400 metres across 12 rolls, and roll 7 is half a shade darker.
- End-of-run rush. The last 10% of an order is produced under shipping pressure. Without a final inspection gate, that 10% is where the defects concentrate.
Each cause has a specific countermeasure, which is why a single "final check" at the end is not a quality system.
The Three Checkpoints
| Checkpoint | When | What is verified | Stop rule |
|---|---|---|---|
| 1. First article | First 3–5 pieces off the line, before the line is released | Measurements against the spec sheet, construction, trims, label placement | Any measurement outside tolerance stops the line until the pattern or the setting is corrected |
| 2. Inline (DUPRO) | At roughly 20–30% and again at 60–70% of production | Spot measurements, workmanship, colour continuity between rolls, stitch density | More than 2 major defects in a 30-piece spot check triggers a full re-check |
| 3. Final inspection | 100% packed, before shipment | AQL 2.5 sampling, packing method, label and carton accuracy, metal detection where required | Rejection at AQL means rework and re-inspection; the shipment date moves, not the standard |
The economics are simple: a defect caught at first article costs minutes. The same defect caught at final inspection costs a rework shift. Caught by the buyer at destination, it costs air freight and a season.
How to Read an AQL 2.5 Table
AQL 2.5 (ISO 2859-1, general inspection level II, single sampling, normal severity) is the default standard for apparel. It does not require checking every piece — it defines how many pieces to pull and how many defects are tolerated in that sample.
| Order quantity | Pieces inspected | Accept (Ac) | Reject (Re) |
|---|---|---|---|
| 151 – 280 | 32 | 2 | 3 |
| 281 – 500 | 50 | 3 | 4 |
| 501 – 1,200 | 80 | 5 | 6 |
| 1,201 – 3,200 | 125 | 7 | 8 |
| 3,201 – 10,000 | 200 | 10 | 11 |
Read it as: for an 800-piece order, 80 pieces are pulled at random and measured. Finding 5 major defects in those 80 is still an accept — that is what 2.5% means statistically. Finding 6 rejects the lot, even though 74 pieces were fine. Two things buyers usually miss: the sample must be random (not the QC's favourite carton), and the reject decision applies to the whole lot, not just the sampled pieces.
Defect Classes
- Critical (zero tolerance): broken needle fragments, sharp hardware, incorrect fibre content on the care label, restricted substances. One finding stops shipment regardless of AQL.
- Major: holes, a run of skipped stitches, measurement deviation beyond 1 cm, a visible shade band between panels, wrong trim colour. These drive the AQL count.
- Minor: loose threads, slight puckering at a seam, a 2 mm measurement deviation. Recorded, not counted against the lot.
Confusing major with minor is the most common dispute in third-party inspection. The defence is a written defect classification agreed before production, not after.
Keeping Bulk Identical to the Sample
Three documents do the work. If a factory cannot produce all three on request, the quality system is verbal.
- Golden sample, signed and sealed. One approved piece stays at the factory, signed by both sides, dated. It is the physical reference when words fail.
- Colour standard with lot numbers. Lab dips approved under D65 light, with the dye lot recorded. Reorders pull from the same lot where possible.
- Version-controlled spec sheet. Measurements, construction and trims on one page with a version number. Any change is written and re-signed — never agreed in a chat message.
What We Ask Buyers to Do
- Approve the measurement sheet (point of measure) in writing, with tolerances — typically ±1 cm on body measurements, ±0.5 cm on critical points.
- Consolidate comments into one round. Three rounds of small changes usually mean the fit was never defined.
- Book third-party inspection (SGS, BV, Intertek) at least 5 working days before the planned ship date. Last-minute bookings are the main reason inspections get skipped.
A Realistic Fix
On an 800-piece dress order, the inline check at 30% found the side seam running 1.2 cm over tolerance on one line — a machine tension setting changed at shift handover. The line stopped, 100% of the affected panels were re-measured, 60 panels were re-cut from reserved fabric, and the ship date held. Cost of catching it inline: two days of one line. Cost of catching it at final inspection: the whole lot back through finishing, or air freight to protect the delivery window.
Quality is a process with gates, not a final mood. If you want to see how the gates map onto a real order, read the reorder process (where the same documents prevent drift on repeat runs), check the FAQ on inspection and AQL, or look at how we handle each category. To discuss inspection terms for your order, send us the details.
Need the whole range rather than one category? Download the 2026 product catalogue (PDF) — 19 apparel categories and 13 footwear series, with MOQ and lead times for each.
Have a style to produce?
Send your reference, quantity and destination — we reply with a full quote within 24 hours.
Get a Quote Open the catalogue