What Arglass Yamamura Taught US Container Glass Plants

It is late 2021 and the first commercial ware is coming off a line that, twelve months earlier, was a slab of concrete in Valdosta, Georgia. No tribal knowledge. No night-shift veteran who knows the forehearth runs three degrees hot in zone four when the ambient drops. No worn-out process sheets taped to the gob feeder with someone's handwriting on them. Nothing. You either had a system or you had chaos.

That was the reality of building Arglass Yamamura from the ground up. Zaid Hassoneh, the Lean Glass founder, led that $220M USD greenfield from commissioning through to stable production, and what it taught him about container glass operations took years to fully unpack. Most of it didn't show up in the capital budget. It showed up on the hot end.

A greenfield strips away every excuse you've been running on

In a settled plant, you can paper over a lot. The experienced operator on B-shift knows that section 4 runs 8ms slow and compensates without logging it. The mould shop keeps non-conforming blanks in rotation because replacing them sits at number seven on someone's work-order queue. You absorb the losses and call it the baseline.

On a greenfield, there is no baseline. The first time you see baffle marks on your early ware, you can't say "it's always been like that." You have to find the cause, because the cause was introduced somewhere in the last 90 days and you were there for all of it. Baffle marks at Arglass meant one thing to investigate first: alignment. Check the centring tool reading. Then check the baffle spring load, then confirm the blank mould temperature is holding to spec (and yes, I know the temptation is to go straight to IS machine timing, but alignment is almost always the faster diagnosis).

That discipline became non-negotiable from day one. Find the root cause without the shortcut. It produced operators who owned their defect rate instead of inheriting an explanation for it, and the difference showed up in first-ware quality within the first quarter of production.

What it also exposed was how much of normal plant operations rests on undocumented institutional memory. The greenfield had none of that to lean on. Every standard had to be written, tested, and signed off before it could be called a standard. That is slow and it is frustrating and it is exactly what most settled plants have never been forced to do.

The job change problem you inherit versus the one you build

Here is what US container glass incumbents usually carry into a job change: a spec sheet that hasn't been updated since 2017, a recipe loaded by the previous shift with three undocumented tweaks, and a mould preheat that someone eyeballed rather than held to the 480°C target because the preheat oven thermocouple was drifting and nobody had chased the maintenance order. None of these are catastrophic on their own. Together they produce 40-minute job change variance between shifts on the same SKU.

At Arglass, there was no 2017 spec sheet. The team had to write the spec the first time, which meant writing it right. Gob weight CV had a target of ≤0.4% and it was logged every shift from the beginning. The hot-end superintendent owned recipe lock. Operators could flag a concern but they could not change set points without sign-off. That wasn't a power structure decision. It was a quality decision, and the data backed it up within eight weeks of commissioning.

What that process produced, without anyone calling it by name at the time, was the foundation of what Lean Glass later formalised as the 9-stage Job Change Lifecycle. Stage by stage, named owner, locked spec, logged outcome. Not because a consultant told us to. Because a greenfield forces you to build the thing properly when you have no alternative.

When you build a plant from scratch, you can't blame history. Every spec you didn't write, every handover you didn't document, every drift you didn't catch. That's yours. Most settled plants never feel that pressure. They should.

What every US incumbent can steal from the greenfield mindset

The US container glass market has been through a tough stretch. Capacity rationalisation, labour-cost pressure, and a domestic demand picture that has left some operators running lines harder with fewer people. The controllable wins are not in new capital. They are inside the existing fleet, inside the existing shift structure, inside the handover that happens at 0600 every morning and takes 12 minutes when it should take five.

In 2019, before the Arglass build began in earnest, I spent time mapping what cross-shift variance looked like on identical SKUs across different US operations. The number that kept appearing was 30-60% variance in job change time for the same article, same mould set, same line configuration. Thirty to sixty percent. That is not a forming problem. That is a system problem.

And that system problem is completely fixable without spending a dollar on new equipment. The fix is documentation, version control, role clarity, and real-time execution tracking. It's not a sexy answer. It's the answer. A 1% OEE improvement at hot-end scale is worth millions in EBITDA, and most plants are leaving 4-8 OEE points on the floor simply because no one has ever forced the discipline that a greenfield commissioning demands.

The greenfield taught me that incumbents have one advantage Arglass didn't. Historical performance data. If you have been running a three-line furnace for 15 years, you have 15 years of evidence for where your variance lives. Use it. A vendor-neutral container glass consultant can help you read that data without trying to sell you a capital upgrade at the same time.

The number that doesn't show up in capital reviews

Twenty-three minutes. That was the average cross-shift gap in job change time at a two-furnace US plant I audited after Arglass. Same article. Same crew size. Same IS equipment. The difference between the best shift and the worst was 23 minutes of dead line time, every changeover.

At 30 changeovers a year on that article, that gap costs roughly 11 hours of production. On a line running 200 containers a minute, you do the maths. It is a number that never appears in a capital review because it doesn't look like a capital problem. It looks like "operational variance," which is how it stays invisible for years while the plant manager wonders why OEE won't move.

The Arglass build was not a normal operating environment. But the lessons from it translate directly to plants that have been running since the 1980s on Emhart 10-section machines with legacy AIS controls. The principles don't change with the equipment generation. Spec ownership, role clarity, real-time execution tracking. Those hold whether you commissioned yesterday or three decades ago.

The Job Change Tool Lean Glass built after Arglass is a direct product of that greenfield experience. It is a systemised methodology with a digital execution layer, built by operators who ran the lines. Not a generic Lean/SMED rebadge with a new cover sheet. If your plant's cross-shift variance on identical SKUs sits above 20%, you already know what the problem is. The question is whether you're going to build the system to fix it.