The 90-Day Engagement Model for US Container Glass Plants

87%. That is the average pack-to-melt ratio I see at US container glass plants when I walk in for a first audit. The best-performing US plants target 93% or above. That six-point gap on a 300 t/day furnace is worth roughly $1.8–2.4M in annual revenue, sitting there every single year.

The market conditions right now make closing that gap urgent. O-I Glass's 'Fit for Growth' programme permanently idled furnaces at Lapel, Windsor and the Toledo complex, removing somewhere between 550,000 and 650,000 short tons of North American capacity. That should improve pricing dynamics for surviving plants. But it only does if you are running efficiently enough to capture the upside.

The competitive pressure inside USMCA isn't going away

USMCA Chapter 28 keeps tariffs at 0% on HTS 7010 container glass traded between the US, Canada and Mexico. Vitro Packaging runs cross-border supply chains from Monterrey that US food and spirits brands are happy to use. You compete on quality, lead time and SKU flexibility. Not price. That is the operating reality every US container glass plant faces right now.

Ardagh Glass Packaging North America carried total debt exceeding $9B entering 2025. Capital budgets at older furnace sites are compressed. You are not getting a new IS machine approved this year. The question is what yield you can recover from the asset you already have.

And the honest answer is: quite a lot.

The P/M gap is a process discipline problem, not a capital problem

In 2023 I was on the floor at a 3-line plant in the US Midwest that had been blaming an aging Emhart Glass 10-section triple-gob machine for a P/M stuck at 88.5%. The plant had submitted a capex request for a controls upgrade. We put that on hold, spent ten days doing section-by-section P/M forensics, and found that sections 3, 5 and 8 were generating 68% of the total reject load. Not a machine problem. A handover problem.

Roughly 60% of the P/M gap at US container glass plants is process-discipline-driven and 40% is mechanical. The process-discipline portion is recoverable without capital in a 90-day engagement, provided the diagnostics are done at section level rather than machine-average level.

Forehearth spout temperature is the highest-leverage, zero-capital fix. Glass needs to exit the spout at 1,040–1,100°C for standard soda-lime. A ±10°C deviation causes ±3–4% gob weight variance. US plants that have not calibrated spout thermocouples within the last six months routinely run 15–20°C hotter than logged. That suppresses viscosity, inflates gob weight and produces checks and seeds on the cold end four to eight hours later. By the time the AQL sheet flags it, the problem is three shifts deep.

Gob weight is the second lever. A 3-sigma spec of ±1.5 g on a 300 g container sounds generous until you watch a +3 g drift on an NNPB lightweight bottle push check defects up by 30–50 PPM. That drift typically originates at the spout needle valve and is correctable within one shift if the operator knows how to read the Agr ISGM or Emhart Glass weigher trace (and yes, most operators on US lines have been told to watch the number, not interpret the trace, which is not the same thing).

Night-shift P/M drift of -2% to -4% versus day shift is routine on unmanaged US lines. The night crew doesn't log the swabbing data. The incoming day superintendent doesn't know the forehearth has crept 12°C. By 0800 the machine is chasing a target that has already moved. Structured shift-handover protocols, covering section-level gob weight and temperature sign-off at shift start, recover one to two P/M points within the first 30 days. No capital. No maintenance work order.

The most expensive thing in a US container glass plant isn't natural gas. It's a 0600 handover where the outgoing superintendent doesn't transfer what actually happened overnight.

What a generic consultancy misses at the section level

An OEM-affiliated consultancy designs its recommendations around capital hardware. That is structurally predictable. A generic Lean/Six Sigma boutique applies DMAIC control charts calibrated for processes with near-instantaneous feedback loops. Container glass hot-end defect causes operate on a four-to-eight hour detection lag from forming event to cold-end AQL sampling. Control charts built without that lag embedded produce false-positive alarm rates that erode operator trust within weeks of implementation.

Neither category benchmarks at the section level. In practice, two or three underperforming sections on a 10-section IS machine account for 65–75% of the total defect and reject load. Section-level P/M forensics is the core differentiator of a specialist container glass consultant, and it is consistently absent from OEM-affiliated and generalist reports alike.

The US recycling rate adds structural cost pressure that batch-cost models rarely capture properly. At approximately 31% in 2023, US plants run far lower cullet ratios than European peers, where the EU average sits at 76%. Higher raw-material costs and reduced cullet buffer mean gob weight variability is harder to hold inside spec on batch chemistry alone. Waiting for a policy shift is not a plant-level strategy. Tightening what you already run is.

What thirty, sixty and ninety days actually deliver

Days one to 30: a 5-10 day hot end audit establishes the section-level P/M baseline. Spout thermocouple calibration, gob weight trace review, shift-handover gap analysis, mould wear condition assessment. The output is a ranked defect-cause list, a P/M recovery estimate with confidence intervals and a prioritised action list that separates process-discipline fixes from mechanical ones. The plant team owns execution from day one.

Days 30 to 60: process-discipline work goes live. Shift-handover protocols are deployed. Section-specific gob weight targets are set and tracked against the baseline. Spout temperature sign-off becomes part of the first-ware checklist at every SKU changeover. The Job Change Tool, a systemised job change methodology with a digital execution layer, is deployed at the first changeover in this window. Cross-shift variance on identical SKUs typically runs 30–60% before systemisation. That variance is the primary source of first-ware instability and the bulk of reject cost in the stabilisation window after every changeover.

Days 60 to 90: the P/M trend confirms whether improvements are holding across all shifts or regressing. Mechanical work orders are prioritised by section-level yield data, not historical incident logs. A 1% P/M improvement on a 300 t/day furnace generates approximately $300,000–$400,000 per year in additional saleable glass at current US container glass pricing. Two percentage points recovered is a material EBITDA movement inside a single quarter.

It's not glamorous work. But it is the work that moves the number on the report that lands on the VP's desk.

If your plant is running below 91% P/M and you want a vendor-neutral assessment of where the gap sits and what it costs per month, start the conversation with our US container glass advisory team.