Sample vs bulk: why your second batch looks different

A sample different from a bulk order almost always means the production process changed, not that the factory tried to cheat you. Hand-built samples get more attention per piece than a full run does. The fix is not a better factory. It is a sample you can point to, a spec you wrote down, and a check on the second and third runs.

This is for a Shopify merchant who has already run one production order from a China supplier and is placing a second or third. You know the product. You are trying to work out why unit two looks a little off from unit one.

Why quality drifts after the sample

A sample is usually built by a supervisor or a skilled operator working alone, on one piece, with no line speed to hit. Bulk production runs on a line, with a quota, and often with a mix of workers who did not touch the sample at all.

A few honest causes explain most of the gap.

• A factory near its capacity limit sends part of your order to a subcontractor you never approved.
• Injection molds and stamping tools wear down after tens of thousands of cycles, so dimensions drift slightly over a long run.
• Staff turnover moves your order to someone less experienced with that particular product.
• A material shortage leads to a substitute resin, fabric, or component that looks close on paper but performs differently.

None of this requires bad intent. It requires nobody checking after the sample was approved. A version of the same problem, deliberate and progressive corner-cutting on a long-standing order, is sometimes called quality fade. It usually shows up as a steady decline over many repeat orders, not a jump on order two. It is also rarer than plain process drift.

The practical question is not whether your factory is honest. It is whether anyone would catch the difference before it reaches a customer.

The golden sample and how to use it

A golden sample is the physical unit both sides agree is correct: the one you approved, kept as the reference for every run after it. Without one, "matches the sample" has nothing to point to.

Keep three golden samples. One stays with you. One stays at the factory, signed and dated by both sides. One goes to whoever inspects your goods, whether that is you, an agent, or your fulfillment partner. Seal each one in a bag so nobody can quietly swap it for a nicer unit later.

When a run arrives, hold a production unit next to the golden sample. Check weight, color, stitching, seams, or fit, whichever applies to your product. A kitchen scale is enough for weight. A five minute comparison on the second run catches most drift before it becomes a pattern on the tenth run.

What to write in the order spec

A sample proves the design works. A spec is what makes the next 500 units match it. Verbal agreement over chat is not a spec, because nobody can check a batch against a chat log six weeks later.

A usable spec names, in writing:

• The exact material or component grade, not a general description like "plastic" or "cotton".
• Weight, with a tolerance in grams or percent, not just a target number.
• Key dimensions the sample proves out, with a tolerance for each.
• Finish and color, ideally against a Pantone or a physical swatch.
• What counts as a critical, major, and minor defect for this specific product.

If your product includes a material the destination country restricts, batteries, liquids, or certain cosmetics, put that requirement in the spec too. See products you cannot ship from China for what triggers extra scrutiny at customs before you even get to a quality problem.

Send the spec with the golden sample, not instead of it. A supplier who has both has no room to claim they misunderstood.

Spotting drift early

Most sellers inspect the first run closely, because it is new, and then stop checking once the process feels routine. That is backwards. The first run is the one built most carefully. Drift tends to show up on the second or third run, after the factory has settled into a rhythm and attention has moved on.

A workable inspection approach borrows from AQL sampling, the ISO 2859-1 acceptance sampling standard used across contract manufacturing. Instead of checking every unit, you check a sample sized to the lot and decide based on how many defects turn up. Two rejected lots out of five consecutive runs moves you to tighter inspection on the next ones. Ten clean runs in a row lets you ease off.

You do not need a formal inspection firm to use the logic. Pick a handful of units from each new batch, not just the first few off the line. Compare them to the golden sample and the spec. Ask specifically about anything unusual: a supplier who answers a direct question about a substitution is behaving differently from one who deflects it.

A price cut you did not ask for is also a signal. If a factory drops its price on its own, ask what changed on their end to make that possible. The answer is sometimes a legitimate efficiency gain. It is sometimes a thinner spec.

What to do when it happens

If a batch does not match the golden sample, document it first. Take photos next to the sample, name the spec section it violates, and count the units affected. Vague feedback like "quality dropped" gives a factory nothing to fix.

Decide what you are asking for before you write to the supplier. A discount on the affected units, a remake at their cost, or credit against the next order are the usual options. Factories respond faster to a specific, documented ask than to a general complaint.

Weigh the cost of drift against your margin, not just against the unit price. A batch of defective units that gets returned or refunded costs you the product, the shipping both ways, and the customer's trust. Add the time spent resolving it on top. If you have not run this math for your own numbers, how to calculate dropshipping profit walks through the full cost side, including refunds.

If the same factory drifts twice after you raised it once, requote with another supplier rather than absorb a third bad batch. China dropshipping agent vs AliExpress covers what changes once a spec and a golden sample replace an unmanaged marketplace order.

Common mistakes

Approving a sample from photos instead of a physical unit. A photo cannot show weight, stitching quality, or how a hinge feels.

Treating the spec as optional once a factory is "known good". Trust does not stop a mold from wearing down or a subcontractor from getting your order.

Only inspecting run one. As covered above, that is usually the best run you will get, not the representative one.

Negotiating price down after the sample was approved. A factory can only build to a spec at a certain cost. Push below that and something in the spec gives.

Frequently asked questions

Is it normal for bulk orders to differ slightly from the sample?

Small variation is normal in any manufactured product. A defined tolerance in your spec, not a vague sense of "close enough", is what separates normal variation from a real problem.

How many units should I check from each new batch?

Enough to catch a pattern, not every unit. A sample sized to the batch, checked against your golden sample, beats checking a handful of random units with no reference point.

Should I hold a golden sample even for a small first order?

Yes. The cost is one extra unit and a sealed bag. Without it, you have no fixed reference for the second order, which is exactly when drift tends to appear.

What if the factory refuses to sign off on a golden sample?

Treat that as information. A factory confident in its process has no reason to avoid a shared reference point that protects both sides.

Does a lower price on the next order mean quality will drop?

Not automatically, but it is worth asking why the price moved. An unexplained price cut on an unchanged spec is one of the more reliable warning signs of a coming substitution.

Getting ahead of it before the next order

None of this requires distrust of your supplier. It requires a reference unit, a written spec, and a habit of checking runs two and three as closely as run one. Most quality drift is caught by someone comparing a real unit to a real sample, not by a stronger contract.

If you are still deciding which factory or product to commit to, run the numbers first. Our product sourcing and landed cost calculator helps you set a spec that matches a cost and margin that actually works.