Why Overseas Jewelry Orders Go Wrong: Six Places the Specification Breaks

When an overseas order goes wrong, both sides usually remember the same conversation differently, and neither is lying. The specification was never written in a language both sides measure the same way. These are the six places it breaks.

1. Units and sizing scales

Millimetres or inches is the easy one. Ring sizing is where it gets expensive: the same finger is a different number depending on which scale you named, and a workshop will use the one it knows unless told otherwise. Bracelet length, bangle inner diameter, earring drop — each has a convention that varies by market. Name the scale and the number on the order, and verify it on the approved sample rather than on the drawing.

2. Tolerance written in the wrong place

“Plus or minus something” means nothing until you say on which dimension. A tolerance on overall length does not control the diameter of a stone seat; a tolerance on weight does not control the wall that produces the weight. The dimensions that decide whether parts fit, whether stones seat and whether the piece survives finishing are not the ones that make the drawing legible. The numbers that actually decide it are usually the ones nobody annotated.

3. Colour described in words

“Rose gold” covers a range, and the range moves with the copper in the alloy — which is also the most variable ingredient in the mix. “Gold colour” does not say what alloy, what finish or what reference. Two batches can both be correct and still not sit together on a display. Colour needs a physical reference: a retained sample, or a stated alloy with a named finish level. The alloy comparison is where the range question starts; the reference piece is where it ends.

4. Colour judged from photographs

A phone camera adjusts white balance on its own; a workshop bench light is not a shop light; a screen is not either. The same piece photographed twice looks like two finishes. Photography is fine for progress and useless for approval — approve colour against a physical reference under light you control, or accept that you are approving an impression.

5. “Looks good” is not an approval

The loose approval is the most common single source of downstream argument. Which version was approved — the drawing as issued, the drawing as amended in messages, or the sample as it physically is? An approval should name the version, the revision, and anything explicitly accepted as a deviation. Without it, production follows one reference while the buyer evaluates against another, and both are certain they are right.

6. Verbal changes

A change agreed in a call is a change nobody can measure later. It also tends to be a change whose consequences were never discussed: move a stone and the wall changes; change the finish and the weight changes; change the clasp and the routing changes. Every revision needs a version number and a note on what it touches — geometry, cost, or both.

What holds an order together

Four mechanisms, none of them exotic. A written specification sheet with a version number. A retained physical sample that both sides can measure. A first production piece checked against both before the run continues. And a written record of every change with what it affects. What to send before quoting is the front half of this; the back half is refusing to let a revision exist only in a chat thread.

The payoff is not fewer conversations — it is that the conversations that remain are about the piece, not about who said what. Sample-versus-bulk drift is almost always traceable to one of the six breaks above, and every one of them is closable in a single email at the right moment.

How we handle the specification side

We ask for the four things that change the manufacturability answer — intended wear, metal, stone sizes, expected quantity — and we put what we understood back to you in writing before quoting. When something is genuinely open, we say it is open rather than filling it with a working assumption. That is slower on day one and considerably faster by the third revision.