Two pieces of silk on the desk, the same momme on the invoice, and only one is the one you approved.
Colour is fixed in a dye bath, weight in a greige lot, hand in a degumming kettle, shade in a steamer. Every one ran twice, and nothing obliged the second to land where the first did.
What decides your position is not the size of the gap but whether anyone wrote a number down first.
No figure a buyer can hold a supplier to says how far a bolt of mulberry silk may sit from nominal. The tolerance is whatever the purchase order says.
The bulk lot was coloured in a different batch from the one you approved, and shade is fixed at the batch, not the artwork file. A dye bath a degree or two off target moves colour uptake by 5% to 10%. The print paste is mixed fresh, and cloth from another greige lot takes dye differently.
One activewear mill holds batch-to-batch variation to grade 4 on the ISO 105-A02 grey scale your fastness reports use. That grade is the mill’s own convention, not a standard’s, so grade 4 against a sealed swatch is the floor I would write in.
Inside a written grade the shift is mill variation. Without one, there is nothing to hold the mill to.
A cloth measured at 0.8 Delta E under D65 daylight can read 3.2 under store lighting, so agree the lamp before either side looks. Then ask for a strike-off, the printed proof, made on the actual bulk lot.
Momme on an invoice is nominal, and the bulk bolt came off a different greige lot, the undyed woven cloth a mill buys and finishes in batches. Loom tension and finishing move weight, width, opacity and drape around it.
How far it may move is not a figure you can look up. ASTM D3776 tells a laboratory how to weigh cloth and sets no deviation limit, so a light bolt is not rejectable, only unspecified.
A greige lot changed without notice is a substitution, and rejectable whatever the scale reads.
The key factor for a buyer is not which tolerance you pick but that one exists. Write the momme with a tolerance, the finished width, and the whole order from one lot.
Degumming is the boil that strips sericin, the gum coating raw silk fibre, and how much comes off is a setting the mill chooses. Raw silk carries 25% to 30% sericin, and removing more leaves a softer, more lustrous cloth at the same weight.
Wöltje and colleagues ran two accepted laboratory protocols on the same silk and removed 35.6% and 32.8% of the sample weight, so the amount removed follows the method.
Over-degumming is worth catching: prolonged boiling fragments the fibroin molecule, so the cloth takes dye fast and anchors it shallowly. Fast dye uptake is not a virtue in silk.
A hand difference at identical weight sits inside normal variation and is hard to reject on paper. Over-degumming is not: it surfaces as a wash-fastness failure you can test for. Hold the degumming level to the sample in writing.
A sample printed digitally and a bulk run put on screens differ in colour gamut, edge definition and ink reaching the reverse. The gap between reactive digital and screen printing is why a screen run cannot reproduce every colour a digital proof showed.
Steam is next. On printed silk held at 102 °C, steaming time moved ink penetration in testing, and was the only factor changing how far ink spread along the warp. Steam acts on the print, not the hand of the cloth.
A print moved between processes without your approval is a substitution, rejectable whatever the shade reads. A steaming schedule that shifts depth slightly inside the same process is variation.
Running digital and screen lines in one building makes the fix cheap: name the method on the order, and hold the bulk to the strike-off’s schedule.
Shade drifts along a roll and across it, and that drift did not exist when your swatch was cut. The same activewear mill holds side-to-centre and head-to-tail variation tighter than batch-to-batch, at grade 4-5 rather than 4.
No swatch could have shown that: the roll had not been woven yet.
Hems at volume run on a line, not at the bench that made your sample.
Readings at head, middle and tail of every roll against the approved colour standard catch it. Pieces for one finished scarf then have to come from the same shade band.
Drift inside those grades is normal. Scarves cut across two bands and shipped as one order are not.
Three artefacts do the work a swatch cannot. A lab dip and a strike-off are both made on sample-stage cloth, so neither has ever met the lot your goods will be cut from.
Willingness to strike off on the actual bulk lot, not on whatever is on the shelf, is the cheapest test in vetting a silk supplier.
Sampling runs 7 to 10 days on our lines and bulk 20 to 25, so a strike-off on the bulk lot and a pre-production sample cut from it land before the clock starts. Both sides seal a swatch.
Colourfastness is the one line nothing else covers. Name a test method and a grade, because an OEKO-TEX Standard 100 certificate carries staining minimums and no wash-fastness requirement.
A supplier who argues about the tolerances beforehand beats one who agrees to everything.
Photograph both pieces under one lamp, keep the sealed swatch, and take the lot numbers off the rolls.
A re-dye sounds cheapest and carries the most risk on silk, because it puts fibre already degummed and dyed back through a bath. A remake costs the calendar.
A price adjustment against partial acceptance is where most of these settle.
Whatever else happens before the next order, put a number on the shade, the momme and the colourfastness. Nobody else will write them for you.
When you look at the numbers, the only question left is which of two categories your delivery falls into. A dye lot, a degumming batch or drift inside a roll is a tolerance question, and a tolerance question is a negotiation.
A greige lot swapped without notice, or a print moved between processes without approval, is a substitution, and a substitution is a claim.
If you are about to reorder, buy the whole quantity in one dye lot: a second lot cannot match the first perfectly.