How to evaluate a factory sample, before it costs you a container.

First, be clear what a sample is: evidence about a factory's capability, offered on its best day. It is not a prototype (that's a build that answers a design question: a different discipline), and it is not a promise about the batch. Evaluated properly, it becomes the standard the batch is held to. Evaluated by vibes, it becomes the reason the dispute has no referee.

The checklist, in order

Dimensions against the drawing. Measure the dimensions that matter (interfaces, fits, wall thicknesses) with instruments, and log the numbers next to their tolerances. "Looks right" is how wrong parts pass. Materials and finish. Verify the material is what was specified, not what was convenient (weight is a quick tell, documentation is a better one), and judge finish where it will be seen and touched, including how it survives a fingernail and a solvent wipe. Function. Cycle it: every switch, thread, hinge and seal, more times than feels polite. Assembly and internals. Open it if you can; tidy internals are the best free signal of a factory's habits. Packaging. If retail packaging is in scope, the sample of the box matters as much as the box's contents.

Log it like you'll need it, because you will

Photograph everything against the drawing, write the deviations down, and send the factory a written disposition: approved, approved-with-changes (listed), or rejected (reasons attached). Log each measurement against its numbered characteristic on the drawing, with the tolerance, the instrument and the date. This paper trail is negotiating leverage, QC baseline and dispute referee in one. The trap of casual sampling isn't a bad sample. It's an undocumented good one, which proves nothing enforceable later.

The golden sample discipline

When a sample finally earns approval, it stops being mail and becomes the contract: signed, sealed, retained by both sides, and named in the purchase order as the reference standard. Every future inspection is judged against it. No golden sample means every batch argument is opinion versus opinion; with one, it's the batch versus the standard you both signed.

The agent-native lens: photos are claims, the record is evidence

When the buyer sends an assistant. More of the first look now happens at a distance. An AI assistant can read a factory's dimensional report against your drawing, flag the values sitting at the edge of tolerance, and compare a new photo set with the last round's for visible differences in finish or assembly. Vision tools can increasingly flag surface defects in images, too. What none of them can do from a photo is weigh the part, feel a hinge go soft after a long run of cycles or confirm what a resin actually is. Ask factories for evidence an assistant can check: measurements tied to numbered drawing characteristics, material certificates, photos with a scale in frame. Treat the rest as the factory describing its own best day.

When the business runs on agents. The sample log becomes the dataset the whole quality program runs on. Agents can draft the disposition from the measurements, list every deviation with its photo, and hold each later first-article, in-line and pre-shipment report up against the golden sample's record, so drift shows up as a trend rather than a surprise. A named person signs the disposition and the golden sample, and no agent edits the standard once it's signed. That record also points somewhere useful: buyers and regulators increasingly ask for a product's materials and how it was checked, and a sample log kept this way is where that answer starts.

The trap at the end

The sample they sent isn't the batch they'll ship. It's the ceiling, made carefully, possibly in a different building. The bridge between sample quality and batch quality is process: first articles inspected against the golden sample, in-line checks where risk demands, pre-shipment inspection before the balance payment. A perfect sample plus no inspection plan is half a system, and the half you paid for is the half that doesn't protect you. The tools for looking at samples will keep improving. The discipline that makes them worth anything stays the same: measure, write it down, sign it, and hold every batch to it.