The Sample Approval Workflow: From First Sample to Golden Sample

Almost every quality dispute we're asked to untangle has the same root cause: the buyer and the factory approved different things. The buyer approved a photo, a verbal description, or an earlier revision; the factory produced whatever was on the workbench. Both sides are certain.

The sample approval workflow exists to prevent exactly this. Done properly, it converts your specification — words and drawings — into a physical object both sides have agreed is correct, then makes that object the standard every production unit is judged against. This guide walks the full sequence: what each stage is for, what to check, how to give feedback that works, what samples cost, and why the golden sample is the most important quality document you own.

Table of contents

1. Why samples exist: spec validation, not free products
2. The sample sequence at a glance
3. Stage 1: the factory or catalog sample
4. Stage 2: the custom development sample
5. Stage 3: the pre-production sample
6. Stage 4: the golden sample
7. Production against the golden sample
8. Giving feedback a factory can act on
9. What samples cost, and who pays
10. Shipping samples: use air courier, always
11. The golden sample as the QC contract
12. Common failure modes in sampling

1. Why samples exist: spec validation, not free products

A sample is not a gift or a formality. It's a test of translation: your specification lives in English; the factory's process lives in Mandarin, machine settings, material orders, and the habits of whoever runs the line. A sample is the only reliable way to check the translation before committing real money. Each stage answers a specific question:

• Can this factory make this kind of product at all? (factory sample)
• Can they make my product to my spec? (development sample)
• Can they make it on the actual line, with production tooling and materials? (pre-production sample)
• Do we both agree, in writing and in plastic, what "correct" looks like? (golden sample)

If a stage isn't answering a question you need answered, compress it. If it is, skipping it means inheriting the risk.

2. The sample sequence at a glance

The full sequence for a customised product:

1. Factory / catalog sample — an existing product off the shelf. Tests capability and general quality level.
2. Custom / development sample — a first attempt at your product, often on prototype tooling. Tests spec understanding; usually iterates two or three times.
3. Pre-production (PP) sample — made with production tooling, materials, and process. Tests the actual manufacturing setup.
4. Golden sample (signed sample) — a PP sample you formally approve, sign, and date. One stays at the factory, one with you.
5. Production — with in-line and pre-shipment inspection comparing units against the golden sample.

For a stock product with only logo or packaging changes, stages 2 and 3 often collapse into a single "modified sample" round. For a fully custom product, skipping the PP sample is the single most expensive shortcut available.

3. Stage 1: the factory or catalog sample

The first sample from a shortlisted supplier should be something they already make — the closest catalogue item to your product. You're not evaluating your product yet; you're evaluating the factory.

What to check: baseline workmanship (seams, joints, mould lines, solder quality — this is the quality level the factory shows a prospect, and production will not exceed it); materials feel; packaging and presentation; and consistency — get two or three of the same item and compare, because variation between supposedly identical units is a preview of your production run.

Order catalogue samples from your final two or three candidates in parallel — comparison is far more informative than any single sample — alongside the 30-point supplier verification checklist.

4. Stage 2: the custom development sample

Once a supplier is selected, the development sample is their first attempt at your product. It may be made on prototype tooling, adapted from an existing mould, or partly hand-finished — normal, because this stage tests design and spec conformance, not process conformance.

What to check:

• Every line of your specification, literally. Print the spec and tick each requirement against the physical sample: dimensions with a caliper, weight on a scale, colours against your Pantone/RAL reference, materials against the stated grade.
• Function. Whatever the product does, do it — repeatedly. Open and close it dozens of times, charge it, drop it, wash it. Samples that survive only careful handling fail in customers' hands.
• Fit with the rest of your system. Packaging, labels, accessories, mating parts.
• What was silently changed. Factories routinely substitute at this stage — a different fastener, a "similar" material — to fit what's available. All substitutions should be surfaced and agreed, not discovered.

Expect two or three iterations. Keep every sample, labelled with revision and date — when the factory says "but this is how sample two was," you want to be able to check.

5. Stage 3: the pre-production sample

The pre-production sample — PP sample — is the stage importers most often skip, and most often regret skipping. Its defining property is that it is made the same way production will be made: production tooling (the actual mould, not the prototype), production materials (batch-ordered stock, not sample-room stock), and the production line and workers (not the sample-room technicians, usually the most skilled people in the building).

Each of those shifts can change the product. New tooling introduces mould lines, sink marks, and dimensional drift the prototype didn't have. Production materials can differ in colour, texture, and behaviour. And the line works to time targets the sample room never faced — hand-finishing that quietly hid a defect on the development sample won't happen at line speed.

What to check: everything from the development stage, again, from scratch — plus tooling artefacts (flash, sink, weld lines, gate marks) and the final packaging as it will ship.

The PP sample is the last point at which fixing a problem is cheap. A flaw found here costs a tooling correction and days to weeks depending on the change; the same flaw at pre-shipment inspection costs a rework of the entire lot — or a fight about who pays. If your product involves custom moulds, this stage also interacts with tooling ownership, covered in our custom moulds, tooling and IP guide.

If anything changes after PP approval — a design tweak, a material substitution, a new packaging insert — you need a new PP sample. This is the rule most often broken; more on it in the failure modes.

6. Stage 4: the golden sample

The golden sample (also called the signed, sealed, or reference sample) is not a new physical stage — it's a legal one. You take an approved PP sample and formalise it:

1. Prepare at least two identical units from the approved PP run. Three is better: factory, buyer, and inspector or agent.
2. Sign and date each unit — physically, on the product or a tamper-evident tag. Both parties sign; include product name, revision, and date.
3. Photograph everything and attach the photo set to the approval email for a timestamped record.
4. Document approved deviations from the written spec ("colour slightly darker than Pantone reference — accepted").
5. Reference the golden sample in your PO and inspection instructions ("production to conform to signed golden sample rev C dated [date]"), stating which governs if sample and spec conflict.
6. Keep your counter-sample somewhere you can find it. The buyer who can produce their signed counter-sample in a dispute wins the argument; the buyer who can't is negotiating from memory.

That last point is the half importers most often drop. The factory's copy protects the factory. Your copy protects you.

7. Production against the golden sample

Once production starts, the golden sample becomes the reference for every quality decision. In-line checks compare units against it during the run. Pre-shipment inspection compares the random sample against it — the inspector should physically have it on the table, and your instructions should say so. In any dispute, the question is never "is this good quality?" but "does this match the signed sample?"

An AQL inspection without a golden sample is an inspector applying industry-default judgement; with one, it's a measurement against an agreed standard. Read how the inspection side works in our guide to AQL 2.5 pre-shipment inspection.

8. Giving feedback a factory can act on

Sample iterations stall mostly for one reason: feedback the factory can't convert into actions. "The quality isn't good enough" tells nobody what to change — the factory guesses, and the next sample is wrong differently.

Feedback that works has a consistent shape:

• One numbered item per issue — a list that can be answered, tracked, and confirmed fixed one by one, not a paragraph containing five complaints.
• A photo for every item, annotated — arrows, circles, measurements drawn on the image, ideally next to a drawing of what correct looks like.
• What's wrong and what correct is. "Logo is 4mm too far left; correct position per drawing p.3, centred on panel" beats "logo position wrong."
• Severity flags. Must-fix or nice-to-have. If everything is critical, nothing is.
• A written close-out. Have the factory reply item by item, confirming each fix, before making the next sample.

Send the list by email, not only WeChat, and keep every round — the numbered lists become the paper trail of exactly what was requested and when.

9. What samples cost, and who pays

The norms are consistent. Catalog samples are sometimes free, but expect to pay courier freight at minimum. Custom and development samples are normally charged, and reasonably so — they consume engineering time, sample-room labour, and materials. Crucially, sample fees are commonly credited back against the production order: the factory is protected against window-shoppers, and a serious buyer effectively pays nothing once the order lands. Ask for this in writing.

Refusing to pay any sample fee is a false economy — it filters you toward desperate factories and signals you may never order. Paying promptly is one of the cheapest credibility signals available, the same dynamic covered in negotiating MOQs with Chinese factories.

Actual figures vary too much with complexity, tooling, and courier weight to quote here — use our sample cost estimator tool for your product and iteration count. And budget for iterations, not one sample: a realistic custom-product programme is two or three development rounds plus a PP round.

10. Shipping samples: use air courier, always

Samples travel by air courier — DHL, FedEx, UPS express — essentially without exception. Slow postal options put a multi-week hole in every iteration loop, which across three rounds can add months to a launch.

Practical points: the factory arranges the courier and you typically pay, ideally on your own courier account. Have the paperwork marked accurately as samples with a truthful declared value. Under-declaring is a false customs declaration made in your name as importer of record, and it costs more in back duty and penalties than it ever saves in freight. Ask for photos and a short video of the exact unit being boxed, so transit damage can be distinguished from manufacturing defects. And ship one tracked, labelled box per revision, or you will lose track of which unit is which.

11. The golden sample as the QC contract

The golden sample is, plainly, the physical annex to your purchase contract. Your written spec says what the product should be; the golden sample shows it. Defect classifications like "colour deviation from reference sample = Major" only mean something because a reference sample exists. When we run inspections through our quality control service, the baseline is the approved sample we hold on file, and the inspector's first check is that the factory's own signed copy is present on the line and matches the revision on the PO. A missing or superseded factory copy is itself a finding — it usually means the line is building to something other than what you approved.

It also outlives the first order. Reorders, factory transfers, second sources — the signed sample and your counter-sample let you say "make this again, exactly," with something firmer than an old email thread.

12. Common failure modes in sampling

The same mistakes account for most sampling disasters:

Approving from photos. Factory sample photos are taken in good light, at flattering angles, by someone motivated to get approval, and cannot convey weight, texture, colour accuracy, or mechanical feel. Approve physical samples in hand, always.

Skipping the PP sample after design changes. PP sample approved, then a "small" change — new colour, thicker wall, different packaging — is agreed over chat and production proceeds against the old approval. The change interacts with tooling or materials in a way nobody predicted, and the whole lot carries the result. No exceptions for changes that "can't affect anything else"; those are the ones that do.

No retained counter-sample. The buyer signs the golden sample, leaves it at the factory, keeps nothing. A year later there's a dispute, the factory's sample has conveniently gone missing, and there is no independent reference.

Confusing sample-room work with production reality. Development samples are made by the factory's most skilled hands with unlimited time; approve one and skip straight to production and you've approved a quality level the line may not reproduce.

Vague approval. "Looks good, go ahead" in a WeChat message, with no revision number, date, or list of accepted deviations. Six weeks later nobody agrees which sample "looks good" referred to. Name all three, in writing, every time.

The bottom line

Sampling is where "we agreed on quality" becomes literally true or quietly false. The sequence — catalogue sample to judge the factory, development samples to converge on your spec, a PP sample to prove the real process, a signed golden sample to freeze the agreement — exists because each stage answers a question the others can't.

The habits that make it work are unglamorous: check the physical sample against the spec line by line, give numbered photographic feedback, pay sample fees and get them credited, re-sample after any change, and keep your own signed counter-sample. The golden sample then serves as the physical standard your pre-shipment inspection enforces, order after order.

If you'd rather have a China-based team manage the sample rounds, chase the iterations, and hold the counter-samples, get a quote — sampling management is one of the highest-leverage places to have people on the ground.

Related guides: AQL Inspection 2.5 Explained · Supplier Verification: 30-Point Checklist · First-Time Importer Mistakes