Carnegie Fabrics Logo
Contract textile specification article cover

Material Notes

Why Fabric Specs Fail After Approval: Field Notes on Chenille, Velvet, and Bulk Curtain Fabric

2026-09-23 · Sahana Iyer

Seventy-two dining chairs in a chenille. Sample approved, PO signed, deposit wired. Eleven weeks later the goods landed, my client walked the room, stopped at the fourth chair, and asked me why it was a different color.

It wasn't. Not technically. Not in any way I could prove on paper.

That one cost $6,400 in reupholstery, three weeks of schedule, and four years of goodwill with a designer who has never fully trusted my sign-off since. And it took me a long time to say this out loud: it wasn't a fabric problem. It was a spec problem. I approved a swatch. I never approved a system.

If you buy fabric for a living — upholstery, drapery, any of it — you've probably been told the hard part is choosing. Comparing price per yard. Weighing one velvet against another. Deciding whether chenille fabric wholesale pricing justifies the lead time. That framing is comfortable because it turns the job into a decision. It isn't one. It's a chain of about six handoffs, and any of them can fail quietly.

What everyone assumes the problem is

The usual story is that the fabric was bad. Inconsistent dye. A mill that shipped something off-spec. Velvet that crushed in the truck.

I've documented 23 mistakes that cost real money — somewhere around $41,000 across nine years of contract and hospitality work. Maybe four of those were genuinely the mill's fault. The rest were mine, and nearly all of them came from the same three gaps: things that don't fit neatly on a spec sheet, so they never get written down.

Gap 1: Abrasion numbers don't travel

"30,000 double rubs" is not a specification. It's a number that came from a test, on a machine, using a method, on a particular version of that fabric. If you don't write down which one, you've written down nothing.

Wyzenbeek (ASTM D4157) abrades with cotton duck on an oscillating cylinder. Martindale (ASTM D4966) uses a completely different motion and abradant. ACT's performance guidelines treat heavy-duty upholstery as roughly 30,000 Wyzenbeek double rubs or 40,000 Martindale cycles — different figures for a comparable bar, because the tests aren't interchangeable. A mill can publish whichever one looks better and both will be accurate.

Then there's what nobody asks: was the test run on greige goods or finished goods? Did the backing go on before or after? Abrasion is also only one failure mode out of five or six. I've watched a fabric sail past 50,000 double rubs and then transfer dye onto about forty people's clothing over one summer, because nobody specified a crocking requirement (AATCC 8). Rub count is the number everyone compares and the one that explains the fewest real failures.

Gap 2: You approved a color, not a color tolerance

So what does "matching" actually mean in writing? It means a number.

The working reference most of us use: a Delta E under roughly 2 reads as the same color to almost anyone; 2 to 4 is noticeable to trained eyes; above 4 is visible to most people standing in the room. For brand-critical color, the industry default sits under 2. That figure traces back to Pantone's color matching guidance, and it's the tolerance I now put in writing on every order.

But the tolerance isn't the part that bites you. Lighting is. A lab dip approved under cool white fluorescents in an office and then installed under 2700K warm LEDs in a hotel corridor will read differently — metamerism, where two swatches match under one light source and mismatch under another because of how their pigments reflect. I had that happen on a job that was already 60% installed. Nothing was wrong with the fabric. Everything was wrong with my approval process.

And dye lots are their own quiet catastrophe. The difference between two production runs of the same colorway can be invisible in a two-inch square and obvious across 72 chair backs sitting side by side.

Gap 3: Nobody owns the handoff

Bulk curtain fabric is the clearest example I know. You buy 340 yards. A workroom cuts and sews it. An installer hangs it. Everyone did their job correctly.

The rails, the pattern repeat, the usable width, whether it's railroaded, whether the pile runs up or down, where the seam lands relative to the window mullion — all of that lives in the gap between "we bought the fabric" and "the drape is hanging." If your spec covers the yardage but not the finished goods, you've split accountability across three vendors who will each be technically right and collectively wrong.

Same problem with "or equal." If your spec says "or equal" and you never define what "equal" means, you've quietly handed that judgment to whoever is pricing the job. Guess which direction the substitution usually goes.

The assumption I had to unlearn

For years I carried a belief that came from an era when "performance textile" meant vinyl or PVC-coated polyester. That was largely true once. It isn't now.

I've been ordering through Carnegie Fabrics LLC on and off since 2017, and the material that changed my default was Xorel — Carnegie Fabrics' PVC-free performance textile, woven from plant-based polyethylene rather than PVC-coated construction. I started specifying it on a library project in 2019 mostly because the client wanted bleach-cleanable surfaces, and I've kept it in rotation since, particularly for healthcare and higher-ed interiors where cleaning protocols are aggressive and replacement costs are brutal.

To be clear: vinyl still does specific jobs extremely well, and I'm not arguing against it. I'm arguing against assuming it's the only route to a performance target. That assumption is stale, and it costs people money.

Which leads to the related oversimplification. It's tempting to think you can compare two fabrics by putting their published numbers side by side. But those numbers were generated by different labs, on different equipment, against different standards, on different versions of the goods. Stacking them in a spreadsheet feels like analysis. Sometimes it's just decoration.

What the gap actually costs

Three from the file, with real numbers.

2018, chenille, 72 dining chairs. Two dye lots on one PO because the mill ran short and I told them to fill it from stock rather than wait. Difference was maybe Delta E 3. Approved it myself after a glance. $6,400 in reupholstery, three weeks of delay, and a very quiet conversation with a hotel GM about why his dining room looked patched.

September 2021, 340 yards of bulk curtain fabric. Specified by the yard, never specified by finished width. The workroom seamed it to make the width, which put a visible vertical seam every 54 inches across a wall of glass the architect had designed specifically to be uninterrupted. $5,200 of fabric effectively donated to a storage unit, plus roughly $1,900 in labor written off, plus a room that sat dark for a week while we re-solved it.

Summer 2023, restaurant banquette velvet. Genuinely beautiful cotton-blend velvet. I wrote an exhaustive spec for it — pile direction, pattern repeat, finish, cleaning code. I never wrote a crocking requirement, because I was focused on how it looked and not on what happens when people sit on it in August. $3,100 in customer dry cleaning claims and $9,800 to replace it six months after install.

The line items above don't include the stuff nobody invoices: storage on wrong goods you can't return because it's been cut, disposal fees, the labor hours spent re-explaining, and the credibility tax — the next time you say a sample is approved, somebody asks for a second opinion.

What I do now

The fixes are boring, which is probably why nobody writes about them.

  • Write the test method next to the number. Not "30,000 double rubs." Say "30,000 Wyzenbeek, ASTM D4157, finished goods." It takes four extra words.
  • Approve color with a number, not an adjective. Delta E tolerance in writing on the PO, and view the sample under the actual lighting of the space whenever you can get access to it.
  • Reserve dye lot yardage. Ten percent overage from the same run costs a fraction of a redo. I stopped arguing with clients about this one.
  • For drapery, spec the finished goods, not the yardage. Railroading, repeat matching, pile direction, seam placement. Put it in the workroom's scope in writing.
  • Add one unglamorous test for the real failure mode. Crocking for anything people sit on, lightfastness (AATCC 16) for anything near a window, pilling for anything high-traffic.
  • Define "or equal" before you go to bid. If you can't define it, remove it.

That's the whole list. We've caught 47 potential problems with it in the last 18 months — mostly dye lot mismatches and missing test methods that would have been invisible until install day.

One caveat on all of this. My experience comes from roughly 300 mid-to-upper-tier commercial orders — hospitality, higher ed, a little healthcare. I've worked almost entirely with domestic mills and a handful of converters. If you're specifying residential or ultra-luxury, or sourcing overseas where documentation practices differ, not all of this will map cleanly. Test what applies and throw out the rest.

Nobody gets fired over a spec sheet. They get fired over what the spec sheet didn't say.

Keep the method and product scope with every shared claim.
Sahana Iyer

Sahana Iyer

Sahana Iyer is an interior-textile analyst covering upholstery, curtain, blackout, chenille, velvet, jacquard, bedding, hospitality linen, awning, and outdoor-furniture fabrics. She applies ISO 12947 Martindale methods and ISO 105-B02 lightfastness testing while comparing abrasion cycles, pilling, colour change, seam slippage, pile direction, dimensional stability, usable width, stain response, and dye-lot variation. Her material guides help interior designers, contract specifiers, furniture makers, and hospitality buyers connect test conditions with traffic level, window exposure, cleaning regime, appearance retention, and replacement planning.

Recent material notes