Home / Journal / Sealed samples
Sealed samples: what to sign, store and send before bulk
A sealed sample is the physical object that decides what "the same as approved" means when a container arrives — so the seal has to fix four things before bulk starts: which object it is, who signed it, which order it belongs to, and where the second copy lives.
Every experienced buyer has had the same conversation at least once. The carton is opened, the pieces are held up against a photograph on a phone, and two people disagree about whether the ear is the right length. Nobody is lying. The reference object exists, but it was never sealed, or it was sealed by one side only, or it is sitting in a sample room with a sticky note that says "Coke, red, final?" — and a question mark is not a specification.
The cost of that ambiguity is asymmetric. A factory that cannot produce its own reference copy has no defence, even when the bulk is correct. A buyer who cannot produce one has no remedy, even when the bulk is wrong. The whole value of the exercise is that both sides can reach the same object within a day, and that the object carries enough information to be compared rather than remembered.
1. Five objects called "sample" · 2. What a seal proves · 3. Three copies, three locations · 4. How long to keep it · 5. Storage that survives · 6. Sending it across a border · 7. The sheet that makes it evidence · 8. Where the dispute actually lands · 9. Our own practice · 10. Questions
1. Five objects called "sample", and only one of them is a reference
Most sample disputes are really vocabulary disputes. The words below get used interchangeably in email, and they are not the same object, do not carry the same authority, and should not be stored in the same place.
| Object | What it is | Authority |
|---|---|---|
| Development sample | First physical attempt at a new design, made to test construction and cost | None. It is a question, not an answer |
| Sales sample | A presentable piece, often made before the order exists | Marketing only. Never the bulk reference |
| Counter sample | The factory's reproduction of a buyer's original, or the buyer's reproduction of the factory's | Binds only once both sides sign it |
| Pre-production (PP) sample | Made with the actual yarn lot, the actual hook size and the actual finishing route | The intended bulk reference |
| Sealed reference sample | The approved PP sample, sealed, signed, dated and linked to the purchase order | This is the one that settles arguments |
The step that gets skipped is usually the last one. A PP sample is approved by email — "looks good, go ahead" — and then goes back into the sample room as loose stock. Three weeks later it is indistinguishable from the six other colourways on the same shelf, and the factory is producing against a memory.
There is a sixth object worth naming because it is often confused with the reference: the top-of-production sample, pulled from the first finished cartons of the actual run. It is not the reference. It is the evidence that the run is being made to the reference, and it is worth photographing and keeping for exactly that reason.
2. What a seal proves — and the four things it must carry
A seal is not a signature on a document. It is a physical interruption of the object: a strip that crosses a fold, a seam or the neck of a bag, initialled across the join so that opening it destroys the evidence. If the strip can be lifted and replaced without leaving a mark, it is decoration.
Four fields have to be on it, written by hand at the moment of sealing, because a printed label can be produced at any time and by anyone:
- The date of sealing, not the date of approval. They are usually different, and the gap is where "which version did we approve" gets lost.
- A version identifier — PP1, PP2, Rev B — that also appears in the email thread and on the measurement sheet. Numbers, not adjectives. "Final" is not a version.
- Both signatories, with the initials crossing the seal itself. One signature means one side can open it.
- The link to the bulk: purchase order number, SKU, colourway name and, where relevant, the yarn lot or dye batch, because a sample made from lot A does not bind a bulk made from lot C.
What the seal does not prove is conformity. It proves identity — that this object is the object that was approved. Whether the bulk matches it is a separate question answered by measurement, by inspection records and, where the product is a toy, by test reports. A seal with no measurement sheet attached is a strong opinion with a sticker on it.
3. Three copies, three locations
The single most common structural failure is a reference copy that exists in exactly one place — usually the buyer's office, sometimes the factory's sample room, occasionally a photograph in someone's inbox. Anything that exists once can be lost, and losing it converts a specification into a negotiation.
The workable pattern is three sealed copies with three different custodians:
| Copy | Holder | Job |
|---|---|---|
| Buyer's reference | Buyer's office or QC team | Comparison at goods-in; the copy that answers a customer complaint |
| Factory's reference | Factory sample room, logged by PO | Line-side comparison during production; the copy that prevents deviation rather than detecting it |
| Neutral copy | Third-party inspector, testing lab or forwarder | Breaks a tie when the first two disagree, or when one side claims its copy was altered |
The neutral copy is the one people skip, and it is the one that matters most in a real dispute, because the first two copies belong to the two parties who are arguing. Where a third-party inspection is already planned — during production, or at pre-shipment — the inspector can hold or witness the neutral copy at almost no extra cost.
Two drafting points belong in the purchase order rather than in email. First, name the custodian of each copy and say that the buyer's copy prevails where a comparison is contested, or that the neutral copy does — pick one, because "the original" is meaningless when there are three. Second, state who pays the courier for the copies. It is a small amount of money and a surprisingly effective source of delay: samples held at a courier desk for unpaid duty are samples that are not being compared.
4. How long to keep it — the few numbers that are actually fixed
Retention is usually answered with a shrug ("two years? three?"). In fact there are three hard anchors and one free choice, and it helps to separate them, because they come from different legal instruments and run on different clocks.
4.1 The compliance clock: ten years in the EU
Under the General Product Safety Regulation, Regulation (EU) 2023/988, manufacturers must keep the technical documentation for a product at the disposal of market surveillance authorities for ten years after the product has been placed on the market (Article 9(3)). The technical documentation is not the same thing as the sample, but the sample is the physical evidence behind several of its entries — construction, materials, dimensions, colour — and an authority asking "how do you know this batch matches what you assessed" is asking for the sample chain, not for a PDF.
Where the toy has been through EC-type examination, there is a second clock: the certificate must be reviewed whenever the manufacturing process, the raw materials or the components change, and in any case every five years (Toy Safety Directive 2009/48/EC, Article 20(4)). A five-year review cycle is not a retention period, but it does mean the sample that supported the original assessment has to be identifiable five years later, or the review restarts from photographs.
4.2 The claims clock: two years at the outer edge
Under the UN Convention on Contracts for the International Sale of Goods, the buyer must examine the goods within as short a period as is practicable (Article 38) and give notice of non-conformity within a reasonable time, with an outer limit of two years from the date the goods were handed over unless that period is inconsistent with a contractual guarantee period (Article 39(2)). Chinese law runs in parallel: Article 621 of the Civil Code makes silence within the agreed inspection period an acceptance, and Article 622 lets a quality guarantee period override the inspection period where one is agreed.
Two years is therefore the minimum sensible retention for anything that could be the subject of a quality claim. In practice, most programmes outlive that: a seasonal line repeats for three or four years, and the complaint in year three is usually "it is not like the one we had before".
4.3 The free choice: programme life plus two years
The practical rule that covers all three clocks without over-engineering: keep the sealed reference for the life of the programme plus two years, and never less than the ten-year documentation period where the product is sold into the EU. For a one-off promotional order that is two to three years. For a repeat retail line it is five to seven. For a licensed character programme with a fixed end date, it is the end date plus two years, and the archive should be searchable by SKU rather than by date, because nobody remembers the month.
5. Storage that survives
Crochet and plush are textiles with a memory. What destroys a reference sample over three years is not handling, it is the environment:
- Light. Cotton and viscose yellow under UV; dyed acrylic shifts. Store in opaque bags in a closed cabinet, not on an open shelf in a window.
- Humidity. Above roughly 60% relative humidity, natural fibres take up moisture, stitch definition softens and mould becomes a possibility. A sealed polyethylene bag with a sachet of silica gel, replaced annually, is enough.
- Compression. A sample squashed under cartons for a year measures differently from one stored flat, and the first thing anyone measures in a dispute is a dimension.
- Pests. Wool blends especially. Cedar or a monitored trap set in the sample room; not mothballs, which leave an odour that contaminates the reference.
Two habits make the physical object usable years later. Label it twice — once on the object's own tag, once on the outside of the bag — because the outer label is the one that gets read and the inner one is the one that survives a bag change. And photograph it at sealing, on a plain background, with a ruler and a colour reference in frame, from the four standard views plus a close-up of any feature that was specifically approved. The photograph is not the reference, but it is what tells you, years later, whether the object in the bag is the object in the picture.
6. Sending it across a border
Samples travel badly for administrative reasons rather than physical ones. Four things go wrong repeatedly:
Value declared as zero. A sample declared at zero value is a sample with zero insurance, and a parcel that is lost in transit then has no claim attached to it. Declare a defensible commercial value — the material and labour cost is normally enough — and mark the waybill "sample, not for resale".
No HS code, or a code that changes between shipments. Where samples are sent repeatedly, an inconsistent classification invites the parcel to be held while someone decides what it is. Use the same code that the bulk will use, and keep it on the invoice template.
The reference copy sent, no copy kept. The buyer's copy should be a copy. Sending the only sealed original and then asking the factory to produce "the same" is how the reference gets lost in transit.
No photograph of the parcel's contents at dispatch. A single photo of the open box with the sealed bags, the waybill and the date visible costs ten seconds and settles the "it arrived damaged" conversation immediately.
For buyers running a first order with a new factory, the sequence that fails least often is: factory seals three copies, sends two (buyer and neutral) with the measurement sheet, keeps one, and emails the photograph set the same day. Nothing about that is expensive. It simply has to be written down as a step, because it is nobody's favourite task.
7. The sheet that makes it evidence
A sealed sample with no data is a comparison of two opinions. The sheet that travels with it should carry the numbers that a dispute would otherwise argue about:
- Flat measurements at the points that matter for the product, with the measuring method stated — laid flat, unstretched, after a stated rest period. Height, width, limb length, circumference.
- Gauge: stitches and rows per 10 cm, because gauge is what changes when a factory substitutes a hook or a yarn count, and it changes the finished size before anything else does.
- Yarn and colour reference: count, composition, and a colour standard both sides can buy — a Pantone TCX reference, not a photograph.
- Weight per piece, and fill weight where the product is stuffed.
- Safety-critical values that the sample was approved against: attachment strength for eyes and noses, seam strength, and the small-parts judgement, each with the test method named.
The point of putting these on the sheet is not bureaucracy. It is that the sealed sample then defines a range, and ranges are what factories can hold across a production run. A sample with no numbers means any deviation is arguable; a sample with numbers plus an agreed tolerance band means a deviation is either inside the band or outside it. The tolerance band itself is a separate conversation — see spec-to-bulk tolerance for how to set it and what destroys it.
8. Where the dispute actually lands
When a sealed sample disagreement reaches lawyers, it is resolved by a surprisingly small number of rules.
The seller's sample is the default benchmark. Article 35(2)(c) of the CISG provides that, unless the parties have agreed otherwise, goods are non-conforming unless they possess the qualities of goods which the seller has held out to the buyer as a sample or model. Note whose sample: the seller's. A buyer's counter-sample binds the factory through the contract, not through the default rule — which is why the counter-sample has to be signed by both sides to mean anything.
In the United States, a sample becomes a warranty. Under the Uniform Commercial Code, §2-313(1)(c), any sample or model which is made part of the basis of the bargain creates an express warranty that the whole of the goods shall conform to it. "Part of the basis of the bargain" is the operative phrase, and a sealed, signed sample referenced in the purchase order is about as clear an example as exists.
Inspection standards do not override the sample. An AQL inspection plan such as ISO 2859-1 tells you how many pieces to examine and how many defects to accept. It says nothing about what counts as a defect — that comes from the specification, and the sealed sample is the specification's physical half. A shipment can pass a 2.5 major-defect AQL and still be visibly not the approved sample, because colour shift and hand-feel are judged against the reference, not counted.
The failure mode that produces the worst outcomes is three references that disagree: a sealed sample in one office, a photograph in an email thread, and a specification sheet with different numbers. When they conflict, the argument stops being about the goods and becomes about which document governs. One reference, one version number, one date — and the sheet attached to it.
9. Our own practice
For context on how this reads from the manufacturing side: on our bulk orders we seal the approved PP sample in three copies, initialled across the seal by the merchandiser and the line supervisor, tagged with PO number, SKU, colourway and yarn lot. One copy travels to the buyer with the measurement sheet, one stays in the sample room filed by PO, one is held for third-party inspection where an inspection is booked. Photographs — four views, a ruler, a colour reference — are taken at sealing and sent the same day, and the top-of-production sample from the first finished cartons is photographed against the sealed reference and the two images kept together.
None of that is unusual. What is unusual is doing it on every order rather than on the difficult ones, because the difficult ones are only identifiable in hindsight. The reference is cheap to create at the moment of approval and impossible to recreate afterwards — and "impossible to recreate" is precisely what a buyer is paying for when they ask for a sealed sample. If you want the full document set that should travel with it, see verifying supplier certificates for the paperwork side and closing audit nonconformities for what happens when an inspection does find a gap.
10. Questions buyers ask about sealed samples
Should the buyer's counter-sample or the factory's sample be the sealed reference?
Either, provided both sides sign the same object and the purchase order names it as the reference. The distinction that matters legally is that Article 35(2)(c) of the CISG defaults to the sample the seller held out; a buyer's counter-sample binds through the contract, not through the default rule. Sealing one object jointly removes the question entirely.
How many sealed copies do we need?
Three: buyer, factory, and a neutral holder such as a third-party inspector or testing laboratory. Two copies belong to the two parties who will be arguing, so a neutral copy is what actually breaks a tie. Where an inspection is already booked, the inspector can witness or hold the neutral copy at negligible extra cost.
How long should a sealed sample be kept?
For products sold in the EU, the technical documentation behind the safety assessment must be available to market surveillance authorities for ten years after the product is placed on the market (GPSR Article 9(3)), and the sample is part of the evidence behind it. For claims, two years is the outer limit for notice of non-conformity under CISG Article 39(2). A practical rule is the life of the programme plus two years, and never less than the ten-year documentation period for EU-bound goods.
Does a sealed sample expire?
The object does not expire, but its usefulness decays: natural fibres yellow under light, moisture changes hand-feel and stitch definition, and compression changes measurements. A sample stored in an opaque sealed bag with silica, flat, in a dark cabinet, remains comparable for years. One stored on an open shelf near a window is a different object within eighteen months.
What if the bulk is correct but the sample drifted in storage?
This is why the measurement sheet and the sealing photographs matter. The sheet records what the sample measured when it was approved, and the photographs show what it looked like, so a comparison can be made against the record rather than against a piece that has spent two years in a warm room. Without those, a factory can be held to a drifted object it never produced.
Is a photograph enough as a reference?
No. A photograph does not carry gauge, weight, hand-feel, or colour under different lighting, and it cannot be measured. It is a supporting record — useful for proving which object was sealed, useless for deciding whether the bulk matches it.
Does passing an AQL inspection prove the bulk matches the sample?
No. An AQL plan such as ISO 2859-1 sets the sample size and the acceptance number for defects; it does not define what a defect is. Colour shift, gauge drift and hand-feel are judged against the sealed reference, and a lot can pass the count while failing the comparison.
A sealed reference sample is only useful if it is physically present when someone measures the bulk. The pre-arrival checklist puts the sample, its measurement sheet and the colour record on the same table before the inspector arrives.
The sealed sample is the physical evidence the conformity file points back to, and its reference number belongs on the invoice and the packing list as well. What travels with a shipment covers the paper side, including the numbers that now sit on the invoice itself.
Related: yarn dye lots and lot-change control — why a tolerance figure without an equation is not a measurement, the four places lots break, and the written notice that turns a complaint into a decision.
Want the sealing protocol as a one-page checklist? Send the product type and the number of colourways, and we will return the seal fields, the measurement sheet and the three-copy dispatch plan matched to your order.