How to vet an OEM custom borosilicate glassware manufacturer: 9 checks, and what a real answer sounds like

Disclosure first, because this sub is strict about it: I work on the manufacturing side — our shop runs custom borosilicate glassware projects (lab glass, drinkware, cosmetic and packaging components) for brands that don't own a furnace. So this is a checklist written by someone who gets audited by it, not a buyer's revenge post. No links, no sales pitch.

The short version, if you only read one paragraph: vetting a custom glass supplier comes down to four gates — (1) certifiable 3.3 material, not "it looks clear"; (2) a forming process that actually matches your geometry; (3) documented annealing with residual-stress evidence; (4) a first-article inspection taken from the production run, not from the sample room. A supplier who can't put written data behind those four is a schedule risk no matter how good the price is. Everything below is how I'd test each one.

1. Is the glass actually borosilicate 3.3, or just "high borosilicate" marketing?

Borosilicate 3.3 has a coefficient of linear expansion of 3.3 × 10⁻⁶ K⁻¹, roughly a third of soda–lime, and tolerates a temperature differential of about 165 °C before fracturing versus roughly 40 °C for soda–lime (Borosilicate glass — Wikipedia). Those two numbers are why your hot-fill or dishwasher story works at all.

What to ask for: the raw glass/tubing batch certificate with the CTE figure and the glass designation (ISO 3585 for 3.3), or a dilatometer/CTE test report on a finished part. What a good answer sounds like: a PDF with a batch number and a date. What a red flag sounds like: "all our glass is high borosilicate" with no designation, no number, no document.

Cheap field test if you have no lab: two polarizing films (or a phone screen plus one polarizer) crossed behind the part. A well-annealed piece stays dark and uniform; a stressed piece shows bright fringes, especially at the base-to-wall junction. It's not a substitute for a report, but it tells you whether anyone is even looking.

2. Does the forming process match your geometry — or is it being forced?

Three processes, three different cost and tolerance profiles:

Press-blow / machine blowing — high cavity count, tight repeatability, but tooling is the entry ticket and geometry must be releasable from a mould.

Hand-blown on a bench — flexible for unusual shapes and low volumes, but wall-thickness spread is wider and lead time follows the gaffer, not the machine.

Tubing conversion (cut, fire-polish, shape a tube) — almost no tooling, excellent for cylinders, vials, straight-walled drinkware, and the natural route for lab glass.

If a supplier quotes a press-blown price on a shape that can only be hand-blown, or promises machine tolerance on a tubing conversion, the quote is fiction and the samples will be the correction. Ask directly: "Which process will you run this on, how many cavities, and what's the wall-thickness tolerance you hold on that process?"

3. Is the annealing real, or is "we anneal everything" the whole answer?

Annealing is where most "it cracked in the field" problems are born, and it's the cheapest thing to fake in conversation. Ask for the annealing/lehr schedule and how they verify that residual stress is inside spec. A serious shop has a strain viewer or polariscope at the end of the lehr and keeps the records; a shop that only says "we put it through the oven" is telling you the oven exists, not that the curve is controlled.

4. What does first-article inspection actually measure?

A usable FAIR on glassware covers more than a drawing: outer dimensions, capacity at the fill line, wall-thickness distribution (min, max, and where), concentricity/roundness, rim flatness and fire-polish quality, base thickness, and — if decorated — decoration placement and colour. Ask what instrument measured each one and when the calipers were last calibrated. If the answer is "we check by eye", you've learned something valuable before spending on tooling.

5. Decoration: fired-on ceramic ink or organic ink that just looks fired-on?

This single question separates two very different products. Fired-on decoration is fused into the surface in a decorating lehr at high temperature; organic (UV/epoxy) inks sit on top and are cheaper and faster. Both can look identical on day one. Ask: which ink system, what firing temperature, dishwasher cycles tested, and against what acceptance criterion (gloss or colour change — not just a tape pull). And ask for heavy-metal migration (lead/cadmium) reports on decorated ware: the risk lives in the ink, not in the glass.

6. Are you inspecting a production sample, or a sample-room sample?

The sample room is a different building from the shop floor, metaphorically and often literally. Ask for samples pulled from the actual production cavity that will run your order, with the cavity number marked. Then ask for the QC data that came with that run — sample size, AQL level, and the acceptance criteria used (ISO 2859-1 is the usual attribute sampling reference).

7. Change control — the question almost nobody asks

"What happens if you change glass supplier or ink brand mid-order?" A supplier with a real quality system notifies you and re-runs first article. A supplier without one treats it as an internal detail. This is the clause that decides whether month 4 looks like month 1.

8. Packaging is part of the spec, not an afterthought

Glass is cheap to make and expensive to break in transit. Ask for the drop and vibration test standard (ISTA 1A or 3A), the inner-pack configuration, and carton compression figures. If they can't tell you how the last shipment that crossed an ocean arrived, ask for the claim rate.

9. The commercial gates

Tooling ownership (do you own the mould, and can you take it?), the MOQ ladder and what actually drives it (cavity count, colour change, decoration set-up), quoted lead time versus observed lead time, and whether the price holds if your volume halves. A supplier who won't discuss tooling ownership is a supplier you can't leave — which is a pricing problem, not a relationship problem.

The six questions I get asked every week

Borosilicate or soda-lime — do I actually need 3.3? If the product sees heat (hot fill, microwave, dishwasher at high temp), rapid temperature change, or chemicals, yes. If it's cold-fill, ambient, decorative and never sees a dishwasher cycle above 60 °C, soda–lime is a legitimate engineering choice, not a downgrade — it's just a different ΔT budget (~40 °C vs ~165 °C).

What MOQ should I expect? It's driven by three things, not by the supplier's mood: cavities in the mould, colour/batch changeover in the furnace, and decoration set-up. Tubing conversion is the lowest barrier. Anyone quoting a MOQ without telling you which of those three sets it is guessing.

Why did my samples pass and my production crack? Almost always one of three: lehr loading changed (annealing is per-piece, not per-batch), minimum wall thickness drifted at the shoulder or base, or the decoration firing introduced a new thermal cycle. Ask for the wall-thickness distribution on the production run, not the mean.

Can I test thermal shock without a lab? Yes, roughly: repeated 100 °C water → ice-bath quench on a sample set, recording where each crack starts (rim, base junction, decoration edge all mean different things). It won't give you a ΔT number, but it ranks your suppliers fast.

Who owns the mould if I pay for tooling? You should. Get it in writing, ask for the tooling serial/photos, and confirm you can have it shipped elsewhere. If that request is awkward, that's the answer.

Is "high borosilicate" the same as 3.3? No. 3.3 is a designation with a number attached (CTE 3.3 × 10⁻⁶ K⁻¹, ISO 3585). "High borosilicate" is a marketing phrase with no number attached — which is exactly why it's used.

Quick reference: what a real answer sounds like

The mistake I see most often

Buyers vet the sample and skip the process. A beautiful sample proves a skilled person can make one piece; it says nothing about cavity 7 at 3 a.m. in week six. If you only have budget for one verification step, spend it on a production-run first article with the QC data attached — not on a prettier sample.

Happy to answer specifics in the comments — geometry, wall thickness, decoration durability, whatever you're stuck on. If you're mid-quote and something in the numbers doesn't make sense, post it and I'll tell you which part is probably wrong.