Three words get used interchangeably for three different objects. A pre-production sample proves the design can be built. A first article proves the first units off the line match it. A golden sample is the signed reference every later batch is judged against. Only one of the three is a contract document.
Quick answer
- Pre-production sample, first article and golden sample are different objects at different moments in the same programme.
- Only the golden sample is signed by both parties and retained on both sides as the acceptance reference.
- A batch that does not match the golden sample is handled by an AQL sampling plan, not by argument. Put it in the purchase order.
What is a pre-production sample?
A pre-production sample (PPS) is made before the bulk run begins, from tooling that has not yet been proven. Its question is “can this be built at all”, and the answer arrives as a list of things that changed: a draft angle adjusted, a gate mark moved, a colour pulled a shade. Nobody signs a PPS as an acceptance standard.
| Attribute | Pre-production sample | First article | Golden sample |
|---|---|---|---|
| When it is made | Before the bulk run, from new or corrected tooling | At the start of the run, from the production line | After approval, before or alongside the run |
| Primary question | Can the design be produced? | Does the first production output match the spec? | What exactly did both sides agree to? |
| Typical quantity | Enough to test and handle, often 5–20 units | A small first-off group | One to a few units, signed |
| Signed by both parties | No | Sometimes, as a record of inspection | Yes, by design |
| Retained after approval | No | Usually no | Yes, one copy each side |
| Binding on a later batch | No | No | Yes, this is its function |
| Used to reject goods | No | Weak evidence only | Yes, together with the AQL plan |
The practical difference is timing. A PPS is made while the tooling is still being corrected, so it reflects a bottle that is about to change. A golden sample is made when the bottle has stopped changing.
What is a first article?
A first article, or first article inspection (FAI), is the first unit or small group produced on the actual production tooling and the actual line. It catches the difference between the sample room and the machine: different colour density from a longer residence time, a heavier wall, a mark from the ejection cycle.
The first article is a production record. It is measured against a drawing and a specification, documented, and usually consumed or packed rather than kept. Under automotive and aerospace quality systems the FAI is a formal report with numbered characteristics; in general packaging practice it is lighter, but the principle is identical — verify the line before the line produces the order.
| What a first article checks | Why a pre-production sample cannot cover it |
|---|---|
| Wall thickness and weight per unit | Resin lot and cooling on the real line differ from the sample machine |
| Colour against the approved Pantone | Masterbatch carryover and residence time differ at volume |
| Decoration registration and coverage | Screen or plate setup is done for the run, not for a sample |
| Neck finish and closure torque | Tolerance stacks only appear when closure and bottle meet in volume |
| Capacity and weight per carton | Packing density and carton pattern are set for the actual order |
| Drop and leak performance | Only meaningful on production units, not on hand-finished ones |
A first article nobody looked at has no protective value. It must be examined and recorded before the run continues, so it has to exist as a document, not a memory. Our article on what to do when a sample is almost right covers the call that usually precedes one.
What is a golden sample, and why is it the only binding one?
A golden sample — also called an approved sample, standard sample or sealed sample — is the single physical unit both parties sign and keep, which becomes the reference the production batch is measured against. Its authority comes from three things: it is signed, dated, and held in identical copies on both sides.
| Property | What it does | What happens without it |
|---|---|---|
| Signed by both parties | Moves the agreement from opinion to evidence | Each side recalls a different bottle |
| Numbered | Makes it identifiable in a dispute months later | “The one you sent” is not a reference |
| Dated | Fixes which tooling and which artwork version it represents | A later run is compared to an outdated reference |
| One copy each side | Neither side can quietly modify it | The factory’s copy becomes the only copy |
| Photographed | Gives a record of the surface, colour and decoration | A dispute turns into a debate about memory |
| Linked to the drawing revision | Connects the physical unit to the written spec | Paper and sample drift apart |
This is why the golden sample, and only the golden sample, appears in a purchase order as an acceptance reference. Our guide to how to sign off the golden sample before mass production covers the mechanics.
How should a sign-off be documented so it holds up later?
A sign-off is a small evidence file, and the only thing that settles a quality argument eight months after the goods were made. Build it like this:
- A unique reference for the approved unit, written on the unit and the paperwork — model number plus sample code.
- The drawing revision number the sample was made against, so nobody approves a sample from an old drawing.
- A date, which fixes which production run the approval applies to.
- Both signatures, on the paperwork and on a label attached to the unit.
- One copy retained by the factory, sealed and marked, stored so it cannot be mistaken for production stock.
- One copy returned to you, kept for the life of the programme.
- Photographs of each side, the neck, the closure fit and the decoration.
- A written defect definition, stating what counts as critical, major and minor in your inspection plan.
- A statement of what the sample does not cover, such as batch colour variance within tolerance, which the sampling plan governs.
- A return condition: returning the golden sample for production without written notice voids the sign-off.
Our note on bottle capacity and the numbers that belong on a spec is the companion here: the capacity block is the part of a specification most often approved verbally.
What happens when a batch does not match the golden sample?
Not matching is not the same as being defective. Colour and surface vary within any run, and the question is whether the variation is inside the agreed band. That band comes from an acceptance sampling plan — the one governing almost all consumer goods inspection is ISO 2859-1, published in the US as ANSI/ASQ Z1.4.
| Defect class | Typical AQL | Examples on a plastic bottle |
|---|---|---|
| Critical | Zero acceptance | Sharp mould flash, a cracked neck, a foreign body sealed in the wall |
| Major | 2.5 | Wrong colour family, a leaking closure, print not registered, deformation |
| Minor | 4.0 | Light flow marks, small cosmetic spots, a slightly off-centre label |
| Lot size | Code letter (General II) | Sample size | AQL 2.5 accept / reject | AQL 4.0 accept / reject |
|---|---|---|---|---|
| 3,201 – 10,000 | L | 200 | 10 / 11 | 14 / 15 |
| 10,001 – 35,000 | M | 315 | 14 / 15 | 21 / 22 |
| 35,001 – 150,000 | N | 500 | 21 / 22 | 21 / 22 |
So for a 5,000-piece run at General Inspection Level II, an inspector draws 200 units. Ten or fewer major defects accept the lot; eleven or more reject it. Fourteen or fewer minor defects accept; fifteen reject. Both sides can look these up in the standard’s tables.
When a lot is rejected, the options are stated in advance in the contract: sort to the plan, rework and re-inspect under tightened inspection, accept with a written concession, or return. Which one applies is a commercial decision made against a fixed rule, not an argument about whether the bottles “look right”. Our article on how many defective bottles are acceptable under AQL works through the arithmetic.
Is a drawing approval or a photo approval as good as signing the physical bottle?
No, and the difference is what each format can carry. A signed drawing or signed 3D file is effective for anything it expresses precisely: dimensions, tolerances, material, nominal capacity, thread. It cannot express surface feel, translucency, print registration on a curved wall, or the colour a masterbatch produces at a real wall thickness — exactly the attributes buyers most want to approve, and exactly the ones that produce disputes later.
| Approval format | Strong for | Weak for | Legal weight |
|---|---|---|---|
| Dimensioned drawing, signed | Geometry, tolerances, thread, capacity, neck finish | Anything visual or tactile | Strong as a specification |
| 3D render, signed | Shape intent and overall proportions | Manufacturing feasibility, finish, colour | Weak as a specification; not producible as drawn |
| Digital photograph, emailed | Rough appearance reference | Colour accuracy, scale, geometry | Very weak; screen-dependent |
| Pre-production sample, physical | Buildability, first impressions | Production repeatability | Weak; not signed |
| First article, physical and recorded | Line output against the spec | Use as a long-term reference | Moderate; a production record |
| Golden sample, physical and signed | Appearance, colour, surface, decoration and fit | Anything not visible, such as resin grade | Strongest; the acceptance reference |
The correct sequence is layered: a signed drawing fixes the numbers, a first article confirms the line, a signed golden sample fixes the appearance. Our guide to the checks to run before approving a sample covers what to look at on the unit that becomes the golden sample.
Which three sample mistakes create the most rework?
1. Approving the pre-production sample instead of a golden sample. A PPS is made from tooling that is still moving. Approving it as the acceptance reference means the batch is later judged against a bottle the factory itself intended to change.
2. Signing without numbering, dating or retaining. Approval recorded only as an email thread produces no usable reference. Number the unit, date the record, send one copy back, and photograph it.
3. Leaving the AQL plan out of the purchase order. Without a sampling plan, “acceptable” means whatever the two parties argue on the day. With one, the batch is measured against numbers both sides can look up in ISO 2859-1 — arithmetic instead of argument.
Frequently asked questions
Is a golden sample the same as a pre-production sample?
No. A pre-production sample tests whether the design can be built; a golden sample is the signed reference the production batch is judged against. Approving the first and calling it the second is the most common confusion in packaging.
Do I need a golden sample on a stock mould?
Yes. The tooling is proven, but colour, decoration and closure fit are not. A stock mould removes the tooling argument, not the appearance one.
Can the factory keep only the golden sample?
Not safely. Both sides hold a signed copy. If only one side holds it, the other has no evidence, and the reference can be lost or altered unnoticed.
What does a first article look like in general packaging?
Lighter than in automotive practice, same idea: the first units off the line, measured against the drawing, recorded before the run continues. A production record, not an acceptance standard.
Does the golden sample set the acceptable defect rate?
No. It sets what conforming looks like. The acceptable rate comes from the AQL plan in the purchase order — commonly zero for critical, 2.5 for major and 4.0 for minor defects.
What if the golden sample gets lost before the next run?
Ask for a new one from retained production stock, signed under the same reference with a note that it supersedes the earlier unit. Do not run a batch measured against an unverified stand-in.
About Shijin Packaging
Shijin Packaging Products Co., Ltd. has been manufacturing plastic packaging in Shenzhen, Huizhou and Guangzhou since 2003. We hold more than 2,000 existing moulds, run ISO 9001, GRS and SGS audited production, and ship roughly 500,000 bottles a day. Every approval we run is written down: a numbered unit, a drawing revision, a date, two signatures, one copy returned.
Send us your model and we will set up the approval chain: name the bottle and closure, and we will quote the sample rounds, the first article check and the golden sample sign-off as dated steps on the production schedule.