OEE Benchmarks in Container Glass: What the Numbers Actually Mean

The OEE gap between the bottom quartile and the top in container glass is approximately 18 percentage points. On a single 200 TPD furnace, close that gap and you recover 50-60 additional production days per year without changing the glass composition, the mould set, or the pack specification.

The problem is that most plants I've worked with don't actually know where they sit. They're quoting OEE numbers calculated differently to the plant down the road, differently again to the benchmark survey they're comparing against, and differently a third time to the OEM-supplied dashboard the IS machine has been running since commissioning. Comparing those numbers is like comparing lap times where half the cars started from a different grid position.

This article works through what container glass OEE benchmarks actually mean, broken down by plant scale, and explains why the normalisation question matters more than the headline figure.

The scale bands that published benchmark data is actually telling you

Industry benchmarking from Glass International and FEVE operational surveys identifies four clear scale bands for container glass. Small furnaces below 150 TPD, typically running two to four IS sections with high colour-change frequency, achieve OEE in the range of 62-74%. Medium-scale plants at 150-300 TPD sit in the 74-82% band. Large, dedicated-ware operations at 300-500 TPD regularly reach 82-88%. Multi-furnace sites above 500 TPD, running a narrow ware range with stable job types, achieve 86-92%.

Those ranges feel obvious once you say them out loud. A small amber-to-flint plant running a 10-section IS machine with five colour changes a month cannot achieve the same OEE as Gallo Glass in Modesto, California, which operates at approximately 1,500 TPD across multiple furnaces on a tightly controlled single-ware programme. Gallo internally cites pack-to-melt ratios reportedly above 94%. At the small end of the scale, a plant achieving 70% OEE is likely at or near its structural ceiling.

The comparison only becomes meaningful within the same scale band. A 200 TPD mixed-colour plant at 78% is performing solidly. A 400 TPD dedicated wine bottle line at 78% has a problem that somebody should be actively looking for.

Why normalisation matters more than the number itself

Here is where benchmarks routinely mislead.

Generic OEE frameworks, the kind a Lean-Six Sigma boutique will arrive with, count colour changes, ware changes, IS section rebuilds, and furnace inspections as availability losses. In container glass, most of those are planned downtime categories that the industry conventionally excludes from the OEE denominator. Include them, and the denominator shrinks. The OEE figure rises. Two plants with identical real performance look different on paper because one operator counted the colour change and the other didn't.

The practical consequence: a plant reading its own "85% OEE" against an industry benchmark of 80% may believe it is outperforming. If the plant's internal figure excludes colour changes and the industry survey includes them, the plant may actually be running at 72-73% on a consistent basis (and I've sat in rooms where this exact conversation ran for two hours before anyone checked the methodology). I've seen that miscalculation justify twelve months of no capital investment, then surface in a furnace campaign audit as a gap nobody on the leadership team could explain.

And it compounds at the quality sub-component level. OEE has three parts: availability, performance rate, and quality. Most IS machine dashboards surface availability and performance prominently. Quality tends to get read after the shift handover, if it gets read at all. The 0600 handover is the moment when the night crew's finish-check data should be reviewed against the day crew's baseline. In my experience, it gets skipped or blended into a line average on the majority of plants without a structured handover protocol. For a precise OEE definition that includes the glass-specific quality denominator logic, it is worth auditing your calculation method before comparing anything externally.

Pack-to-melt is the number your OEE should be grounded in

Pack-to-melt ratio is the glass-specific quality sub-component that matters most. The range for well-run operations is 93-96%. The industry average sits at 87-91%. Each one-point improvement on a 200 TPD furnace at an average net selling price of €0.25 per bottle recovers approximately €430,000-€650,000 in EBITDA per year. Not OEE points. Euros.

In 2018 I was running a medium-scale plant in Southeast Queensland, two furnaces, seven lines, amber and flint on separate units. We were confident in our 80% OEE figure until we disaggregated pack-to-melt data by section and shift. The night crew on furnace one was running 88.4% pack-to-melt against the day crew's 91.7%. Nobody had raised it because the blended line average still looked acceptable. The composite OEE figure had masked a three-point quality gap for months. Twenty-three minutes. That was the extra time the proper section-by-section check added to the 0600 morning handover. It recovered three percentage points of pack-to-melt across the following quarter.

On high-speed NNPB lines, finish checks (hairline cracks at the sealing surface of the bottle finish) are the single largest driver of quality OEE loss. Standard press-and-blow lines run a finish-check rate of 0.3-1.8% of pulls. On a maintained NNPB line with gob temperature controlled to ±5°C and plunger timing held within ±5ms, that rate drops to 0.1-0.5%. The gap between those two ranges is a forming-process problem, not a machine-speed problem. It belongs inside your quality sub-component calculation and stays invisible on a bottles-per-minute dashboard.

A plant that compares its OEE to an industry benchmark without first normalising the denominator is not measuring performance. It is measuring methodology.

The campaign-end OEE floor that forming-side work cannot fix

There is a structural OEE ceiling that drops as a furnace campaign ages, and most generic asset-management frameworks do not model it.

In the final two to three years of a furnace campaign, targeting 10-12 years total, AZS refractory spalling at the melt-glass interface accelerates. Stones PPM rises by approximately 0.5 PPM per year beyond campaign year eight or nine. The target for premium food-grade production is below 10 PPM. Late-campaign furnaces approaching that threshold on specific ware types during peak pulling seasons cannot address the gap from the forming section. You can tune the IS machine all week. The ceiling belongs to the furnace.

Usable pull rate in late campaign also typically falls 5-8% due to crown geometry changes and thermocouple drift in the working end. Older Emhart IS machines running 1990s-era pneumatic timing controls compound this further, because section timing repeatability degrades with the control hardware independently of tooling wear. A plant managing a 10-year-old furnace alongside 12-year-old section controls is tracking two independent degradation curves simultaneously. The OEE number will show both, but only if the quality sub-component is captured at section level rather than blended into a line composite.

Ardagh Glass Packaging's European estate of approximately 18 plants reported portfolio-average OEE in the low-to-mid 80s range per 2023 investor materials, with highest-performing dedicated-ware plants approaching 89% and the smallest multi-colour plants sitting at 72-76%. That spread is not purely a management outcome. A meaningful part of it is campaign age and furnace scale working as physics dictates. A vendor-neutral assessment separates the structural floor from the controllable losses on the forming side. An OEM-affiliated approach tends to frame everything as a machine intervention opportunity, which produces a structurally different kind of recommendation.

Four questions your OEE benchmark needs to answer first

Before you put an OEE number into a board presentation or a capital request, four questions need honest answers.

• What is in your denominator? Colour changes, planned maintenance, ware changes, IS section rebuilds. Are they in or out, and is that definition consistent across shifts and across sites?
• Which scale band are you benchmarking against? A 180 TPD mixed-colour operation belongs in the 74-82% peer group, not alongside a dedicated single-ware site running above 500 TPD.
• Is your quality sub-component tracking pack-to-melt by section and shift, or a line-end composite? A composite number hides the variance that drives cost.
• Where is your furnace in its campaign? Late-campaign furnaces carry an OEE ceiling that needs to be quantified and separated from controllable forming-side losses before you can set a credible improvement target.

Plants using a structured Job Change Tool have the changeover timing data to answer the first question reliably. Plants running on shift supervisor estimates typically don't, and that gap shows up the moment you try to build a consistent normalised denominator across sections or sites.

Gob weight deviation is worth tracking separately as one of the most controllable forming-side variables. Top-quartile forming sections hold gob weight to a standard deviation of ≤0.8g for lightweight wine and spirits bottles below 300g. Moving from a σ of ±2.5g to ±0.8g enables a 5-8g bottle weight reduction, which opens up 8-12% additional usable pull capacity on the same furnace without capital expenditure. That intervention shows up in pack-to-melt before it shows up anywhere else.

Gulf Glass Manufacturing in Kuwait reported average OEE of approximately 78% in 2023 against a 2025 target of 82%, per the GPCA Sustainability Report. That four-point gap on a two-furnace 350 TPD operation represents roughly 40-50 additional production days annually. Whether the gap lives in availability, rate, or quality determines everything about where to intervene. Without reliable KPI reporting that separates those three sub-components at section level, the 78% is a headline without a diagnosis.

Look, the data tells you where the top and middle quartiles sit. It doesn't tell you what is inside your number or why your plant sits where it does. If you want a properly normalised OEE benchmark against a peer group matched to your furnace scale and campaign position, our hot end audit builds that picture at section level, with a methodology that a vendor-neutral container glass consultant can defend at board level.