The sample arrives, the fit is off, the colour isn't yours, the print sits in the wrong place — and the obvious conclusion is "wrong factory". Sometimes it is. Far more often, the factory built exactly what the brief allowed them to build, and the gap lives in the spec.
Six causes explain most wrong samples. Find yours before you pay for round two — or switch factories and repeat it.
Nearly every wrong sample we've seen comes down to one of these six — and five of them start on the brand's side of the conversation:
Instructions live in chat threads, emails and a reference photo, and no single document holds the whole garment. The factory works from fragments and guesses the rest. Fix: one document carries every specification, so there's one source of truth.
What a tech pack actually contains →
"Heavyweight cotton" is a wish, not a spec — GSM, composition and finish decide how a garment feels, drapes and washes. Without numbers, you get the mill's closest stock fabric. Fix: specify weight, composition and finish, or approve a physical swatch before sampling.
"Oversized", "relaxed" and "true to size" mean something different in every factory on earth. Fix: measurements in numbers — chest, length, sleeve, shoulder — with a tolerance. Numbers travel; adjectives don't.
Print or embroidery positioned from a photo lands somewhere slightly new every round. Fix: the design's size in centimetres, and its position measured from a fixed point — centre chest, so many centimetres below the neckline.
The photo shows a ribbed cuff; your message says plain. The factory follows one of them, and it wasn't the one you meant. Fix: say which source wins when they conflict, and mark the details that matter on the reference itself.
The tweak you agreed on Tuesday isn't in the document the sample room works from on Friday. Fix: every agreed change goes into the pack and the revision list the same day it's agreed.
Lay the sample next to your spec and go line by line — measurements, fabric, construction, artwork, trims. Split every difference into one of two piles:
Most wrong samples contain both piles. Knowing which is which turns an argument into a revision list.

When you reply, send this — all of it, once:
Revision rounds cost time and money — the goal is one thorough round, not five hopeful ones.
What sampling and revisions cost →
My sample came back wrong — does that mean the factory is bad? Not on its own. Judge the factory on how they handle it: do they acknowledge clear errors, remake what was theirs, and communicate precisely? A factory that fixes a wrong sample well is a better sign than one that never gets tested.
How many sample rounds are normal? For a new style, one or two revision rounds is normal — first samples surface the gaps nobody saw on paper. What isn't normal is the same error repeating after a clear, written specification. That's the signal to worry about.
Who pays when a sample comes back wrong? It follows the cause. Where the spec was clear and the build differs, a reputable manufacturer remakes it at their cost. Where the brief was vague or changed, the new round is usually yours. Agree the revision terms before you pay for the first sample.
How should I give feedback on a sample? One document: photographs of what arrived, annotated at the problems; measured values beside the spec values; and each change written as an instruction, not an impression. Send it once, complete — dripped feedback across days gets half-applied.
Can I skip sampling and go straight to bulk? You can, and it's the most expensive way to learn what a sample would have taught you. Bulk multiplies every unresolved detail by your full order. Sampling exists so the mistakes happen at one unit, not one hundred.
What should I check first when a sample arrives? Measurements against the spec table first — that's the garment's skeleton. Then fabric weight and hand-feel, construction quality (seams, stitching), artwork size and placement, and trims and labels. Check in that order and write everything down before you reply.
Send the product, quantity and fabric you're planning — we'll cost the actual style, 100 per style, sample first.