A sample answers one question: will production look and behave like this? Getting there takes several samples, each with a different job. A pre-production sample is made to your brief and is where faults get corrected. The golden sample is the one that matters commercially: signed by both sides and held as the physical reference an inspector judges the shipment against.
Four samples, four different questions
Buyers often speak of the sample as a single object. In practice a project produces several, and confusing them causes real problems, most often when a reference sample is treated as an approval.
What is the difference between a prototype, a pre-production sample and a golden sample?
A prototype or reference sample is an existing product, or a rough first article, used to agree a starting point. It confirms the platform and the general standard. It is not made to your specification and approving it commits you to nothing.
A pre-production sample is made to your brief, often by hand or on soft tooling. It is where dimensions, function and finish direction get checked and corrected. Expect at least two rounds and treat the first as a conversation rather than a verdict.
A golden sample is the approved article, signed and sealed by both sides. It becomes the physical reference for inspection. Retain one copy and leave one at the factory, both signed and dated.
A shipment sample is drawn from the actual production run and confirms the line reproduced the approved sample at volume. It arrives too late to prevent a problem, which is precisely why the earlier stages matter.
How long does a pre-production sample take?
For a catalogue product with a logo change, days. For a product needing hand fabrication or soft tooling, two to four weeks is a realistic first round, and you should plan for two rounds rather than one. Where new hard tooling is involved the sample waits on the tool, which is a different order of time entirely. Ask for the sample lead time as a separate answer from the production lead time, because they are often quoted together and then assumed to overlap.
What to check, in order
Assess a sample in a fixed sequence, because the expensive faults are structural and the cheap ones are cosmetic. Buyers who start with the logo often miss a dimensional problem entirely.
Swipe or scroll the table to compare all columns.
| Check | What to look for |
|---|---|
| Dimensions | Measure against the drawing, not against your memory of it. Record actual figures rather than pass or fail. |
| Material and grade | Confirm the grade for each component. Request a mill certificate where corrosion performance matters. |
| Function | Use it as a customer would. Cycle hinges and closures repeatedly. Load bowls and trays to a realistic weight. |
| Weight and balance | A product that measures correctly can still feel wrong in the hand. Compare against the reference product. |
| Surface and edges | Run a hand over every edge that a user touches. Burrs on food-contact edges are a safety defect, not a cosmetic one. |
| Finish consistency | Check the finish across the whole part and between components, especially where two processes meet. |
| Decoration | Verify logo position, size, colour and durability. Test it against washing and abrasion, not just appearance. |
| Packaging | Assess the pack as a unit: fit, protection, print quality, barcode scannability and how it opens. |
Record measured values rather than verdicts. A note saying handle length 118 millimetres against a specified 120 is actionable. A note saying handle slightly short is not, and it will be interpreted differently by everyone who reads it later. Where the grade matters to corrosion performance, our 304 vs 430 comparison explains what to ask the factory to evidence.
Test the sample in its real conditions
A sample assessed on a desk tells you how it looks. Kitchenware fails in use, so put it through the conditions it will actually meet.
- Run it through a dishwasher cycle several times if the product will be described as dishwasher safe. Check for water spotting, loosened handles and finish degradation.
- Expose it to the foods it will meet. Acidic contact, lemon, vinegar, tomato, is the fastest way to reveal a material specified too optimistically.
- Leave a part wet overnight. Corrosion resistance is about prolonged moisture, not a brief rinse.
- For anything with a moving part, cycle it several hundred times rather than a dozen.
- Put the packed unit through a drop from shelf height, and if it ships direct to consumers, through a courier journey.
These tests take days, not minutes, which is why they need to be scheduled into the approval window rather than squeezed in when the factory is waiting for a decision.
Should the packed product be transit tested before the first shipment?
If it ships as an individual parcel, yes, and there is a defined procedure rather than a guess. ISTA publishes standard test procedures by distribution channel: ISTA 3A is the "Test for individual packaged-products shipped through a parcel delivery system... distributed as individual packages, either by air or ground", ISTA 1A is "Integrity Testing for Packaged-Products weighing 150 lb (68 kg) or Less", and ISTA 6-Amazon.com-SIOC covers products shipped to Amazon fulfilment centres without additional packaging.
Naming the procedure that matches your channel turns a vague request to make sure it survives shipping into something a factory and a test house can both act on. The packaging guide covers how that decision interacts with carton quantity and freight cost.
Making approval mean something
What is a golden sample, and how is it sealed?
An approval is only useful if it creates a reference that a third party can apply months later. That requires more than an email saying the sample looks good.
- Sign and date the physical sample, and have the factory do the same. Seal it so it cannot be swapped.
- Photograph the approved sample from several angles and attach the images to the specification.
- Where a finish varies naturally, approve a limit sample showing the worst acceptable case as well as the target.
- Write the measured values from the approved sample into the specification, replacing target figures with achieved ones where they differ and you have accepted the difference.
- State explicitly what was not approved. If you approved the product but not the packaging, record that.
Keep your copy accessible rather than filed away. When an inspection is disputed six months later, the signed sample in your office is the document that settles it.
How does the approved sample relate to AQL inspection?
The golden sample is what an inspector compares the lot against, so it is the thing that makes a defect definition meaningful. An inspection plan sets how much deviation is tolerable: as the NIST/SEMATECH Engineering Statistics Handbook puts it, "The AQL is a percent defective that is the base line requirement for the quality of the producer's product."
What that percentage is measured against is your approved sample and your defect classes, not an abstract standard. Agree both before production starts rather than at the factory gate. Our AQL guide covers how to turn the checks above into inspection criteria, and the quality control page sets out how we run in-line checks against the sealed sample.
Sample costs and getting them to you
Who pays for samples, and are the costs refundable?
Samples are rarely free, and the terms are worth settling in the first conversation. A common arrangement is that the buyer pays for the sample and the courier, and the sample cost is credited against a subsequent production order. That is reasonable on both sides, but it only applies if it was agreed in writing beforehand.
Where a sample requires soft tooling or hand fabrication, the charge can be substantial and is usually not refundable, because the work is genuinely bespoke. Ask for the sample cost, the lead time and the refund terms as three separate answers rather than assuming any of them.
Shipping introduces its own friction. Courier charges on a single heavy sample can exceed the sample cost itself, and customs treatment varies: some destinations clear low-value samples easily while others assess duty regardless. Ask the factory to declare the shipment accurately, because an under-declared sample that is held by customs costs more time than the duty ever would.
For anything with a blade or a sharp edge, confirm the courier will carry it and how it must be packed. A sample refused at the depot can add a week to a schedule that had no room for it.
When tooling changes between sample and production
What if bulk production does not match the approved sample?
A pre-production sample made by hand or on soft tooling will not always behave like a part from the production tool. Soft tooling produces slightly different draw depths, radii and surface characteristics, and hand finishing can flatter a sample in ways a production line cannot repeat.
When you know the tooling will change, ask directly what that changes, and ask for a first-off sample from the production tool before the run proceeds. This is the single most useful request in the entire approval process, and it is frequently skipped because the schedule is tight by that point.
If a completed run does not match, the sealed golden sample is what turns a disagreement into a decision: the lot is measured against it, the deviation is classified against the agreed defect classes, and the options are sorting, rework, a negotiated discount or rejection. Without a sealed sample the discussion becomes two opinions, and the party holding the goods usually wins it.
Build the time in from the start. A first-off check that costs three days in week six is considerably cheaper than discovering at pre-shipment inspection that the whole run has a radius the approved sample never had.
Your checklist
- Sample type understood at each stage, and reference samples not mistaken for approvals
- Sample lead time quoted separately from production lead time
- Dimensions measured and recorded as values, not verdicts
- Material grade confirmed per component
- Dishwasher, acid and prolonged moisture testing completed where relevant
- Moving parts cycled to a realistic count
- Transit test procedure named for the actual distribution channel
- Golden sample signed, dated and sealed by both sides
- Limit sample approved where finish varies naturally
- Defect classes agreed against the sealed sample before production
- First-off sample requested from production tooling