Golden Sample, Different Shipment: Where Quality Drifts

A buyer approves a sample coverall in week three. The fabric is right, the shade is right, the pockets sit where the drawing said. Production is released. Eight weeks later the shipment lands, the shade is visibly off across two of the four cartons, the main zip is a different brand from the one on the sample, and garments marked large measure closer to the sample medium. Nobody set out to substitute anything. The approved sample had been treated as the control, and it was never a control. It was a reference, and a reference enforces nothing on its own.

That gap is where most workwear quality complaints actually live. Not in a factory deciding to cheat, but in a specification that named a sample instead of naming numbers, and in a production run nobody checked until it was already in a container.

The four places drift starts

Fabric is the first. Woven and knitted cloth is dyed in lots, and lots differ. An order large enough to need three dye lots will show three shades unless one shade standard governs all of them, and the difference is often invisible beside a single garment but obvious across a stack. Weight drifts too. A fabric quoted at 240 g/m² is a target with a tolerance, and where the tolerance was never written down, the factory tolerance applies rather than yours.

Trims and components are the second, and they move faster than fabric. Zips, buckles, thread, reflective tape, snaps and labels are bought in by the garment factory, so when one is short on the day the line is scheduled, the quickest fix is a functional equivalent. On a plain polo that substitution is cosmetic. On a flame-resistant coverall it is not, because the thread and the zip tape are part of what was tested.

Sizing is the third and the most commonly mishandled. A sample is one size, usually a medium. Bulk is a graded set, and the grading rules rather than the sample decide what the extra large measures. Add shrinkage after the first wash and a garment that was correct at the factory is wrong on the body.

Substitution against a certified construction is the fourth, and it is a conformity question rather than a quality one. Protective performance belongs to a tested combination of fabric, thread and construction, not to a garment appearance. ISO 11612:2015 sets minimum performance requirements for clothing protecting against heat and flame, and those requirements are demonstrated on a specific made-up construction. A mill change mid-order does not reduce quality a little. It removes the basis on which the garment was certified.

What the approved sample actually proves

It proves the factory could make one correct garment once, under attention, from material it had in hand. That is worth having. It is not evidence about the thousandth garment.

For protective clothing the regulation is explicit that approval of a type and control of production are two different things. Under Article 19 of Regulation (EU) 2016/425, category III personal protective equipment requires EU type-examination (module B, Annex V) and then, in addition, either conformity to type based on internal production control plus supervised product checks at random intervals (module C2, Annex VII) or conformity to type based on quality assurance of the production process (module D, Annex VIII). Flame-resistant, arc-rated and chemical protective clothing sit in that category. The type-examination certificate on its own is therefore not the whole answer, and a buyer is entitled to ask which production-phase route the manufacturer actually runs.

That is the principle to carry into any order, certified or not. What you approved and what ships are governed by different controls, and only one of them is a sample.

Write the tolerance, not the adjective

Match the approved sample is not a specification. It cannot be measured, so it cannot be enforced, and on a marginal shipment it produces an argument rather than a decision. Every attribute that drifts has a method and a number available instead.

Shade is gradable. The grey scale in ISO 105-A02:1993 assesses change in colour on a five-step scale, extended to nine steps with half-steps, so a shade tolerance can be written as a minimum grade against the sealed sample rather than as an opinion. Dimensional change is measurable. ISO 5077:2007 specifies the determination of dimensional change in washing and drying, so a maximum percentage after a named wash procedure belongs in the specification next to the measurement chart.

Sizing and marking have a home standard. ISO 13688:2013 carries the general requirements for protective clothing, including size designation, ageing, compatibility and the information the manufacturer must supply, and it states that it is only to be used together with a standard containing requirements for specific protective performance. Naming it is not a substitute for naming the performance standard.

On high-visibility garments, colour is not a preference at all. ISO 20471:2013 and its 2016 amendment set performance requirements for colour and retroreflection, minimum areas of material and placement, so a shade that has drifted outside the permitted range is a non-conformity rather than a cosmetic complaint. A second edition of that standard is in development and has not been published, so the 2013 edition with its amendment is still what an order should cite today.

Lock it before the line starts

Five controls, all cheap, all before production. A sealed sample, dated and signed, one held by the buyer and one by the factory, with shade standards for every colour approved against it and governing all dye lots rather than the first.

A tech pack stating fabric composition, weight with tolerance, construction, stitch type and density, and a trim card naming every component by specification, with a line saying no substitution without written approval.

A measurement chart for every size with a tolerance at each point, and a size set approved before bulk cutting rather than one sample size.

An inline check at first output of the first size, which is the only point where a grading or shrinkage error is still correctable without remaking the order.

A pre-shipment inspection with a named acceptance quality limit and sampling plan. ISO 2859-1 is the usual reference, and the edition matters: a third edition was published in January 2026 and replaced the 1999 edition along with its amendment and corrigendum. Inspection clauses copied forward for years commonly still cite the 1999 text, which is withdrawn.

How we follow production up

This is the work we do between a buyer and a qualified manufacturer. We are the trading company and not the producer: the factory manufactures, and we select it, specify to it, inspect it and carry the consequence when it drifts.

In practice that means turning the operational requirement into a tech pack a factory can price and be held to, holding the sealed sample and the shade standards on the buyer's behalf, approving the size set before bulk cutting, placing the inline check at first output, and running the pre-shipment inspection against a written acceptance quality limit before anything is released to freight. Where the goods are certified protective clothing we also check that the certification covers the construction actually being made, and that the production-phase obligations above are real rather than assumed. One party is accountable for all of it, which is the practical difference between buying through us and buying direct.

What to send us

You do not need a tech pack to start. Send the garment type, the protection or performance the work requires, quantities with an approximate size breakdown, the destination country and the date the goods are needed on site. If you hold a sealed sample or a previous specification, send that too, including the version that went wrong. Where the last order drifted is the most useful document in the file, because it names the control that was missing.

We come back with the specification we would put in front of a factory, the standards it should cite, and the inspection points we would hold the order against.