In September 2022 we took delivery of 1,400 glass candle jars ahead of the holiday season. Production looked fine. Cartons were intact. Then, about 300 pieces into the pallet, I ran my thumb along the inside rim of a jar and felt it catch on a thin, ragged lip of glass. Barely visible on a shelf. Definitely visible on a customer's hand once the wax burns down.
We ended up hand-inspecting all 1,400 pieces over three days. Roughly 220 of them went straight into the dumpster. We air-freighted the balance at about 4x our sea freight rate. Total damage was probably in the $1,100-1,300 range plus nearly a week of lost time, right when everyone was already in Q4 mode.
My first reaction was the obvious one: "the factory's QC is terrible."
That conclusion felt good. It was also mostly wrong.
The Real Problem Rarely Shows Up in the Arrival Email
Since then I've repeated the same mistake with different products — a wildflower scatter dinnerware set where the print registration drifted between boxes, a reorder of vintage-style Libbey glassware, and a custom water bottle project for a client who kept asking how do you make your own water bottle without, you know, actually sourcing bottles. Same pattern every time.
The defect showed up at the dock. The decision that caused it was made weeks earlier, usually by me.
1. You're buying from a spec sheet, not from something you can hold
"Glass candle jar, flint glass, 80mm diameter, electroplated rose gold, 220ml." That sentence describes about ten thousand different objects. It says nothing about rim finish, wall thickness variance, bottom thickness, or how the plating holds up after 40 hours of warm wax sitting against it.
Our first order went through a full pre-production sample round. Photos, measurements, a physical sample on my desk. The second order — the one that failed — went through none of that. I reordered from the same supplier with the same PO and assumed the word "same" was doing heavy lifting.
When I put our approved sample photos next to the photos from that second delivery, I finally understood the thing I should have understood in year one: I had approved a sample. I received a production run. Nobody in that chain was ever asked to confirm those two things were the same.
2. "Supplier QC" doesn't mean what most buyers think it means
This is the one that stung. Most factories are not inspecting every unit. They're sampling — pulling X pieces per batch, checking against an agreed AQL level, and shipping. If your agreed standard is loose, or worse, if there is no written standard at all and the checklist lives in someone's head, their "QC passed" stamp means approximately nothing to you.
To be fair, this isn't a conspiracy. Full inspection on a 5,000-piece glass order is expensive and slow, and most buyers don't want to pay for it. The problem is that the buyer (me) assumed 100% inspection was happening and priced the order like it was.
I'm not a glass engineer, so I can't tell you the right annealing curve or what your furnace temperature profile should be. What I can tell you from a procurement seat is this: if you cannot write down, in plain language, what a defective unit looks like and how many defects you'll accept, you don't have a quality standard. You have a hope.
3. MOQ pressure pushes everyone into betting before testing
Here's the part that ties into something I feel strongly about. We once tried to place a 150-piece trial order on candle jars with three different suppliers. Two didn't reply. One replied with "our MOQ is 3,000." That's not a moral failing — volume pricing is real, setup costs are real. But it means the only path forward for a small buyer is to gamble the full 3,000 on an untested product.
We eventually found a fourth supplier who took the 150 pieces at about $9.40 each instead of the $6.20 we'd budgeted. Everyone on my side hated that price. Eighteen months later we placed a 6,000-piece order with that same supplier, and I never went back to the first three.
Small doesn't mean unimportant. It means untested. A $200 trial order is the cheapest quality audit you will ever run, and treating it as a nuisance is how buyers end up committing five figures to a product they've never held.
What It Actually Costs You
The visible cost is the scrap. That's the easy number. The invisible costs are worse:
- The sort. Three days of your team's labor to hand-check 1,400 pieces, plus the argument about who pays for the replacement freight.
- The season. Candle jars are a September-to-November business. A five-day delay in October is a five-day delay you cannot recover. Our retail accounts run on local searches — someone typing "scented jar candle nearby" is not going to wait for your restock.
- The reorder trap. Once a batch ships with marginal defects, you have no leverage. You can reject it (and eat the delay), or you can accept it (and eat the returns). Either way, the buyer absorbs it.
- The trust. This one's the most expensive. When you tell a retail account "it'll be there Tuesday" and it isn't, they don't remember the reason. They remember the Tuesday.
I don't have hard data on industry-wide defect rates for glassware — I wish I'd tracked my own numbers from the start. What I can say anecdotally, after eight years of purchase orders, is that something like 8-12% of our first deliveries from a new supplier needed some kind of rework or renegotiation. Roughly speaking, that's the ballpark for anyone buying from an unfamiliar factory for the first time.
What I Do Now (The Short Version)
I don't have a 40-page SOP. I have four rules that go on every PO, and they took me about three years and a few thousand dollars in scrap to arrive at.
- No physical sample, no PO. Not a photo. Not a "we always make it this way." A sample, in your hand, with the rim or edge or print you actually care about. Keep it. Seal it. Reference it on the PO.
- Write the defect definition yourself. Two sentences is fine: "A jar is rejected if any edge can catch on a cotton swab. We will accept up to 2% per 1,000 pieces." If the supplier won't agree to it in writing before you pay, that's a red flag, not a negotiating position.
- Price the trial separately. Assume the trial order will cost 30-50% more per unit than your target. That premium is your insurance policy. Budget it, don't fight it.
- Photograph the arrival. Every carton, before you open anything. It's not paranoia — it's the only thing that makes a claim conversation short instead of endless.
The bottom line is that almost none of the problems I've written down over the last eight years were caused by a factory doing something sneaky. They were caused by me assuming that "repeat order" meant "same product," that "QC passed" meant "inspected," and that a small trial order wasn't worth a supplier's time.
Two of those assumptions were mine. The third one is the industry's — and it's the one I'd push back on hardest.