Record an approved toy sample by sealing one physical piece as the golden sample, labelling it with tool number and revision, and issuing a signed one-page sample record that lists photos, version, accessories and where the retained reference is stored. That single document — not the email thread that produced it — becomes the production standard, so any later change to colour, texture, logo, assembly or packaging is visible as a version step rather than a silent substitution.
A buyer's RFQ situation makes the point fast. An importer approves a sample from photos, places the PO, and receives a second lot that looks slightly off. Nobody can prove which detail was agreed, because the approval lived in chat. This page walks through the sample approval checklist a toy supplier can actually be held to, and how to keep the record usable months later.
Key Takeaways
- A sealed golden sample holding the approved material, colour, decoration, dimensions and photos is a practical production-QC reference — but it does not replace required market testing (approved_fact: golden_sample_lock).
- Production release should reference an identified approved sample, never a loose reference photo, especially for colour, texture, logo and assembly (approved_fact: production_release_sample).
- Any later change — colour, function, accessory, carton or retail packaging — should create a new sample version, so the change is visible instead of argued about.
- Colour matching only holds when resin grade, surface finish, lighting condition and the approved master sample stay the same; change any one and appearance moves (approved_fact: plastic_color_same_grade).
- Numeric defect limits come from the agreed inspection plan on the PO, not from a catalogue sentence or a blog.
Why does a photo-only approval fail on the second lot?
A photo cannot carry the four variables that decide how a toy looks in bulk. Emailed images flatten lighting, hide surface finish, and freeze a single angle, so the factory's second run has no fixed target to hit. Buyers who approve only a photo pay for it on the second lot.
The fix is physical. Seal the approved piece and keep it on the QC shelf, then compare bulk against that piece on colour, fit, flash and function. This is the reference that ends a colour dispute; email cannot, because email is a description and the sealed sample is the thing itself. Note the limit clearly: the golden sample is a production-QC reference, not a substitute for the market testing your destination requires.
What must be written on the sample before it is approved?
Label the sample with its tool number and revision at the moment of capture, and keep that label attached. This matters most at trade fairs, where a booth sample is often a hero unit built for display. Fair lighting lies; the revision card does not. A Fair photo should be explicitly excluded as the production standard.
Record the approval date and the identity of the sample itself, so a later claim of 'that's the one we approved' can be checked against a physical object instead of memory. Production release should point at that identified approved sample and not at a loose reference photo, particularly for colour, texture, logo and assembly (approved_fact: production_release_sample).
What belongs in the signed sample record?
Keep it to one page: product reference and tool number, revision, approval date, approver name and signature, a photo set, the approved details, and the storage location of the retained reference. Then list open points separately and mark them as pending — silence is not approval.
The photo set should cover the angles that later disputes actually turn on: overall front, colour-critical surfaces, decoration and logo placement, accessory layout, label and instruction booklet, and retail packaging. Documented observations (what you saw on the piece in front of you) stay separate from recommendations (what you want changed before release). Mixing the two is how a 'minor note' becomes a de facto requirement nobody agreed to.
Sample record fields: what to check and what a bad entry looks like
| Physical reference | Check: a sealed golden sample held at a named location. Red flag: approval rests on a photo, a chat message, or a Fair hero unit. |
|---|---|
| Version / revision | Check: tool number plus revision printed on the sample and repeated in the record. Red flag: no version, so any later change is invisible. |
| Signed approval | Check: approver name, date and signature on the one-page record. Red flag: a verbal 'sample approved' with nothing attached. |
| Accessories and labels | Check: accessory list, label and instruction booklet captured in the photo set. Red flag: accessories treated as 'obvious' and left out. |
| Packaging and carton data | Check: retail packaging, carton quantity and carton data confirmed against the physical sample. Red flag: packaging settled later by email. |
| Market documents | Check: which file covers which test, for which market and which lot. Red flag: one certificate reused to cover two different requirements. |
| Open points | Check: pending items listed and marked, not blank. Red flag: an empty open-points field interpreted as full approval. |
How do you make a later change visible instead of arguable?
Treat every change as a new version. If the factory proposes a substitute resin, a different decoration method, a reworked accessory or revised retail packaging, the record gets a new revision number and the change is compared against the retained approved piece before acceptance.
Colour is the usual flashpoint. Appearance shifts with resin grade, surface finish, lighting condition and the approved master sample — all four variables (approved_fact: plastic_color_same_grade). So a substitute that is 'the same colour code' is not automatically the same appearance. Put the piece next to the sealed sample under the same light before you sign off.
Where tooling is involved, freeze the commercial terms in the same document rather than a handshake: fee, ownership, storage period, release conditions, and what happens if the tool sits idle. A buyer once asked who owned the tool after paying a fee, and the contract had not answered the storage years or the release conditions. Tooling policy is a document.
Where does packaging and booking instruction fit in the record?
Into the same file. Corner-sensitive goods die in LCL when the booking instruction is a thank-you note. Write double-wall, corner boards, ECT and pallet height into the sample record and the booking file, so a packaging change is as visible as a colour change.
Claims are won on the booking file. If the record says the carton is double-wall with corner boards and the shipment arrives single-wall, the signed page is what makes the gap provable.
What happens when a lot fails against the approved sample?
Isolate the failed lot, record the defect class, and release only after rework. Do not reach for a number that was never agreed. Numeric limits belong to the inspection plan on the PO; buyers who want a figure should get it from that agreed plan rather than from a catalogue sentence or a supplier blog.
The signed record makes this workable. Instead of debating whether a defect is 'acceptable', you compare the lot against the sealed reference and the approved details, and the conversation moves to rework terms.
Worked example (illustrative, not a real shipment)
An importer approves an ABS-bodied toy with a printed logo and a retail window box. The record is signed, the golden sample is sealed on the QC shelf, and the booking file says double-wall carton with corner boards for one destination while the second destination takes the same product in a polybag with a different label. One lot, roughly a container's worth, two packaging variants.
The hold point is the label and packaging step. During production follow-up, the bulk pieces match the sealed sample on colour, fit and function, but the second destination's labels arrive printed on a lighter stock than the approved booklet. Because the record separated accessories and labels from appearance and named the retained reference, the variance is visible on the page. The lot is isolated at packing and the labels are reprinted before release. Nothing here is a real order, and no defect rate or AQL figure is claimed — the numeric limit would come from the agreed inspection plan on the PO.
What to ask suppliers before you sign the sample record
Can you provide a signed sample record with photos, version, accessories and the retained reference?
Which physical piece is the golden sample, where is it stored, and who holds it?
What is the tool number and revision on this sample, and will you confirm the Fair photo is not the production standard?
For this market and this lot, which file covers which test — and can you show the mapping?
If tooling is involved, what are the fee, ownership, storage period, release conditions, and what happens if the tool sits idle?
What are the packaging specs for this shipment — double-wall, corner boards, ECT, pallet height?
What is the agreed inspection plan on the PO, and how are defect classes recorded and lots released after rework?
Compliance box: keep testing and the sample record separate
The sample record is a production control document. It is not a certificate. A toy conforming to harmonised standards whose references are published in the Official Journal is presumed to conform to the covered essential safety requirements (approved_fact: eu_harmonised_standard_presumption), and electrical properties are addressed separately for toys containing electrical functions — mechanical EN 71 evidence alone is not a complete electrical assessment (approved_fact: eu_electrical_requirement).
Also keep document types apart. UKCA and a VOC file on an electric bubble machine are different papers; a buyer once tried to use one certificate for both. Map market, lot, and which file covers which test. Paperwork that does not match the lot is not paperwork.
FAQ
Can I approve a toy sample from photos alone?
No. Approving from photos leaves the second lot without a fixed target. Seal one physical piece as the golden sample with the approved material, colour, decoration and dimensions, and keep the photos as supporting evidence rather than the standard itself.
What is a golden sample in toy sourcing?
It is the sealed, retained physical piece that production is compared against on colour, fit, flash and function. It is a practical production-QC reference, but it does not replace the market testing your destination requires.
How should a factory change be recorded after approval?
As a new version. Any change to resin, decoration, accessory, packaging or carton data gets a new revision number and is compared against the retained approved piece before acceptance — particularly for colour, texture, logo and assembly.
Do I need a separate record for each destination market?
Yes, when documents differ. Map which file covers which test for each market and lot. Reusing one certificate across two different requirements is a common and avoidable failure.
Where do I get the defect number for inspection?
From the agreed inspection plan on the PO. A supplier blog or catalogue should not supply it. On the floor, failed lots are isolated, the defect class is recorded, and release happens only after rework.
How long should the approved sample be kept?
For as long as the SKU is being ordered and for any claim window you have agreed. If tooling is involved, also freeze the storage period and release conditions in writing, including what happens if the tool sits idle.
Sources
Turn your next sample approval into a signed record
Before the next PO goes out, take one physical piece, label it with tool number and revision, photograph the details that cause disputes, and sign a one-page record naming the retained reference. If you want a second set of eyes on the fields that matter for your market and packaging, send the draft record with your next RFQ and we will tell you where the gaps are.