Why First-Ware Quality Is Your Real Changeover KPI

It is 0215 on a Tuesday and section 4 has just taken its first gob on the new mould set. The machine restart time is logged at 23 minutes. The hot-end superintendent marks the job change complete on the wall board, clocks the time, and heads to the control room.

Nobody on that shift is tracking what happens next.

The KPI most container glass plants are not measuring

In most plants I audit, “changeover time” is measured from line-down to first gob delivery. It is a mechanical metric. It tells you how fast your crew can pull moulds and reload a recipe. It does not tell you how long it takes to produce ware you can palletise. Those are two very different numbers, and conflating them costs you money every single change.

The correct metric is time-to-spec (TTS): the elapsed time from first gob to the moment each section is producing ware within quality acceptance criteria. According to the most recent FEVE Container Glass Benchmarking survey, fewer than 30% of container glass plants track TTS as a formal shift KPI. The rest are measuring machine restarts and calling it efficiency.

Machine restart time tells you how fast your crew can swap moulds. Time-to-spec tells you how well they understand the glass.

TTS matters because the gap between first gob and first palletisable ware is where your changeover cullet accumulates. Industry data shows a ware-rejection spike of +18 to +35 percentage points above steady-state in the first 15-45 minutes of forming after a mould change. The range is that wide because it is driven primarily by mould temperature at the moment of first gob contact. If your blank-side temperature is below 480°C when that first gob lands, you are making cold-mould marks.

In 2017 I was running a two-furnace, six-line plant in Queensland. We had been measuring changeover performance by restart time for years and, by that metric, we were hitting target consistently. Then we started logging TTS by section across 12 consecutive job changes. Forty-three minutes. That was the real number. Our 23-minute changes were actually 43-minute quality events. Not a forming problem. A measurement problem.

What the first 45 minutes actually produce

The defect sequence during first ware follows a predictable pattern once you have seen it enough times. Minutes zero through five are dominated by bird-swing and gob weight instability. Bird-swing, those trailing glass threads between articles caused by mismatched shear timing or a gob viscosity shift, are the most visible failure mode at this stage. Shear blade gap must be at or below 0.5 mm, and gob weight CV must return to ≤0.8% before first-ware acceptance has any real meaning.

Minutes five through 30 are where checks and eccentricity take over. Checks, hairline fractures in the finish or body, spike hard whenever mould cooling-water inlet temperature sits outside 28-34°C or the delta-T across the mould exceeds 8°C. Wall-thickness eccentricity on first ware routinely runs ±25% in the first 10-20 cycles, well above the ±15% maximum under ISO 7458. Each section needs individual trimming. One experienced operator per four-section bank is the right resourcing level. Those check failures are exactly the rejects that get logged as “cold-end reject, cause unknown” at the 0600 handover, because the shift that ran the change has gone home and the data never made it to the log.

And then there is hot-end coating. SnO₂ coating weight target is 15-40 mg/m², but the first three to five minutes of production typically runs 20-30% light on coating because the glass surface temperature has not stabilised. That shows up at cold end as lehr whitening and reduced abrasion resistance. The bottles look fine until the palletiser or the retailer’s stacking machine disagrees. On older Hartford IS machines running mechanical section timers, this coating instability window is longer and harder to control because you don’t have the section-level speed trim that servo-driven takeout gives you on a current-generation machine.

First-ware defects by time window:

• Minutes 0-5: bird-swing, gob weight CV above 0.8%, shear timing instability
• Minutes 5-20: eccentricity, parison reheat asymmetry, plunger stroke drift
• Minutes 20-45: checks from cooling-water transients, mould seam flash as clamps bed in
• Throughout: SnO₂ coating weight deficit until glass surface temperature stabilises

Mould preheat: specify it to temperature, not to time

Cold-mould marks are preventable. The preheat protocol must bring blank and blow moulds to at least 460°C before first gob. On a 10-section IS machine, that takes 20-40 minutes depending on ambient temperature and mould alloy. The problem is that most plants run preheat off a clock. “Twenty minutes in the oven and you’re good to go.” That might be close enough at 40°C ambient in a plant running at capacity in Dammam. It is almost certainly wrong at 10°C ambient on a cold restart in January in northern Europe.

GCC plants carry an extra burden at this stage. Cullet return rates across the Gulf run 18-25%, compared to the EU average of 76%, which means those furnaces are running batch compositions with more than 70% virgin raw material. Higher virgin batch loads produce wider thermal variability through the forehearth, and that variability arrives at the IS machine as gob temperature instability in the critical first 15-30 minutes of forming (and yes, check the centring tool reading on blank-side alignment before you start the preheat clock, not after the change is already running).

A vendor-neutral forming audit will separate how much of your first-ware cullet is a mould preheat problem from how much is a gob stability issue or a recipe load problem. Those are three different corrective actions with three different cost profiles. Any independent container glass consultant worth talking to will tell you the same thing: a plant-wide first-ware percentage is a management number, not a diagnostic one. You need section-level resolution.

First-ware cullet costs more than you are booking

There is a unit-economics issue that carbon and sustainability teams are consistently missing. First-ware cullet returned to the batch re-processes the same tonne of glass a second time. That doubles its embedded energy and carbon intensity per net unit of packed output. Under CBAM, EU Regulation 2023/956, which moves into full financial force on 1 January 2026, exported containers from GCC and other non-EU plants will carry a carbon levy calculated at installation level. Extended first-ware stabilisation periods that generate significant cullet raise specific CO₂ per tonne of net packed product and directly inflate the CBAM liability on affected export batches.

Verallia’s 2023 Annual Report states that a 1% increase in cullet usage rate reduces CO₂ intensity by approximately 0.35%. Run that in reverse through your job change cullet data and you can put a carbon cost on every additional minute of first-ware stabilisation. US plants face the same arithmetic from the batch side. The Glass Packaging Institute reported a US container glass recycling rate of just 31.3% in 2022, meaning North American furnaces are running virgin batch ratios comparable to or higher than many GCC plants, widening the first-ware thermal stabilisation window and amplifying the cost of each changeover in both EBITDA and carbon terms.

Most plants have not done that calculation. The ones that have find it changes the conversation at VP level very quickly.

The operational fix is not complicated. Track TTS by section. Set a mould preheat spec to a thermocouple reading, not a kitchen timer. Run the machine at 85-90% of target BPM during first ware and don’t revert to full speed until wall thickness and weight are confirmed within spec at section level. The Job Change Tool was built to enforce exactly this sequence through the first ware stage of the changeover, with section-by-section sign-off before quality acceptance is declared. If you want to understand where your plant sits against plants that have already done this work, a forming audit is the right starting point.

If you pulled your last 20 job change records tomorrow and calculated TTS by section, what number would you find on the whiteboard?