Getting First-Ware Quality to 90%+ in Eight Weeks

It is 2am on a Tuesday and your hot-end superintendent is standing at the cold-end inspector watching first ware tick through at 84%. The night-shift operator says the machine feels right. QA says it is the glass. Maintenance says check the moulds.

Nobody is wrong, exactly. And nobody has a number they can defend in the morning meeting. That moment, first ware ambiguity sitting across a shift handover with no agreed cause, is where most container glass plants lose the eight-week opportunity I want to describe here.

The audit gate most plants don't use

First ware quality in container glass is commonly treated as an outcome. You make bottles, you count the good ones, you report the number. What the evidence shows, across plant after plant in the Middle East and the US, is that first ware is better treated as a gate. Either you have earned the right to call production stable, or you haven't. That distinction changes everything about how you respond to variance.

The audit gate works like this. Before a forming line is logged as in production following a job change or restart, it must pass a defined quality threshold at the cold-end inspector. Not by feel. Not by the operator's read of the machine. A scored gate with named pass criteria: dimensional compliance, EVI acceptance rate, a pull sample inspected against the customer's AQL specification. If the gate fails, the line does not enter the pack count. Production time is not the incentive. Quality is.

Plants in the GCC region that have installed this gate as a formal forming sequence step typically recover 0.8–1.5 percentage points of first ware rate within the first four weeks, not from any process change, but from the simple discipline of not counting bad ware as good. The numbers make the case quickly. Raising first ware quality from 88% to 90% on a 300 TPD furnace recovers approximately six additional saleable tonnes per day. At USD 420–480 per tonne packed, that is USD 140,000–161,000 over an eight-week programme. No capital expenditure required.

What your defect categories are actually telling you

The second reason plants plateau at 84–87% is that they treat rejects as a single category. A reject on the cold-end inspector is not a data point. It is a question that needs a category before it can have an answer.

Seeds, gaseous micro-inclusions under 1 mm, are the dominant first ware defect mode in flint glass production. Root-cause ranking by frequency:

• Batch raw-material moisture above 3% equivalent
• Sodium sulphate fining dose below 0.2% as SO₃ on batch weight
• Furnace pull-rate increases exceeding 5% within a 24-hour period

Each mechanism raises seed count materially, but each requires a different intervention with a different response lag. A generic Lean/Six Sigma boutique without glass-physics literacy cannot differentiate a seed from furnace redox imbalance, which needs a 4–8 hour raw-material or fuel-ratio response, from a seed caused by a forming cold-spot you can address in the same shift. Getting that call wrong means corrective actions go in the wrong direction entirely. The improvement curve plateaus at week three and regresses by week five. I have seen it happen on plants that spent months on DMAIC before calling someone who had actually run the lines.

Stones are the second defect mode by severity. AZS refractory-derived stones appear as 0.3–2 mm octahedral zirconia crystals with characteristic UV fluorescence under a 365 nm lamp. A UV inspection station in the cold-end cabin lets operators separate refractory-origin stones from batch-origin stones within the same shift. Refractory origin means a furnace intervention is coming. Batch origin means look at cullet contamination. Those are different departments, different timelines, and different costs. Getting the diagnosis right on shift avoids a week of mis-directed corrective action.

First ware quality is not an outcome you report. It is a gate you earn the right to pass. Treating it as an outcome is why most plants have been stuck at the same number for three years.

Then there is the calibration blind spot (and yes, your EVI may be generating false rejects right now without your dashboard showing it). Electronic Vision Inspection systems running OEM-default reject thresholds, without explicit mapping to the end-customer's AQL specification, routinely generate 0.5–1.5 percentage points of unnecessary over-rejection. The Glass Packaging Institute reported in 2024 that US craft spirits and premium food customers are now embedding first ware minimums of 90% and AQL 1.0 into supplier qualification contracts, converting first ware from an internal KPI into a commercial contract requirement. If your EVI is calibrated to a different sensitivity, your reported first ware rate does not reflect your actual process performance. Fix the calibration before you touch the process.

Stabilisation is a discipline, not a recipe load

In 2021 I worked with a two-furnace plant in the GCC running a Heye 12-section machine on narrow-neck press-and-blow. Experienced operators, correctly loaded recipe, clean mould set. First ware after job change had been running at 85–86% for six months. The plant manager had accepted it as the machine's floor.

It wasn't.

Blank mould temperature was running at 535–540°C, well above the 480–510°C target for NNPB. At those temperatures you get parison skin chilling irregularities that produce cold fold surface defects, and those will cost you 0.5–2.0 percentage points per affected section before you have touched anything else. The night-shift had drifted the mould cooling while chasing a separate problem and nobody had re-locked the set point. The hot-end superintendent owns recipe lock. The operator does not change set points without sign-off. That is a system design statement, not a blame statement. And the system had a gap.

Gob weight control is the stabilisation variable that most consistently separates 85% plants from 90% plants. Target ±1.5 g on a 300 g gob (±0.5%). Drifts exceeding ±3 g correlate with a +0.8–1.2 percentage-point spike in dimensional rejection at the cold-end inspector within 20–30 minutes of onset. That makes gob-weight trending the fastest leading indicator a shift team has. If you are not trending it continuously on the forming line display, you are flying with instruments you are not reading.

Annealing lehr entry temperature must be maintained at 565–575°C for standard soda-lime container glass. A sustained drop below 545°C pushes residual stress above the acceptable threshold on a polarised-light stress meter, raising stress-related rejects by 1.5–3.0 percentage points. Those rejects show up at the cold-end inspector and sometimes do not appear until the customer's pasteurisation or carbonation trial. Neither is a good place to find them.

Mould swabbing is the third variable routinely under-controlled. Industry standard is every 20–30 minutes with a graphite-oil compound. Extending beyond 45 minutes on green or amber glass runs correlates with a +1.5 percentage-point increase in cold mould surface defect rates as the lubricant film burns off and bare iron contacts glass. At most plants I audit, the night-shift swabbing log is either missing or unrecorded by the time the morning meeting runs. Not a furnace problem. A handover problem.

The eight-week window is real, and it closes

Egypt's container glass producers, EGC and Sphinx Glass, are currently reporting first ware quality of 84–88%, sitting below the 90%+ threshold that European and North American beverage brands are writing into supplier qualification audits (Glass Worldwide, 2023). That gap is not a furnace limitation. It is a forming and inspection discipline gap of exactly the kind described above. And it has a commercial consequence that plays out in contract renewals, not in furnace readings.

But that window closes. Plants that close the gap get the qualification. Plants that don't lose the specification to someone who did.

The path from 84% to 90%+ in eight weeks requires four things working in sequence: a formal audit gate on the forming line, a defect taxonomy that distinguishes seed from stone from EVI artefact, a stabilisation protocol with locked set points and enforced swabbing intervals, and someone who can follow the defect back through forming, through the furnace, through the batch, wherever the data leads. An OEM-affiliated consultancy is constrained to the forming section it supplies. A generic Lean boutique cannot read the defect. Neither one is going to diagnose a batch moisture problem from a cold-end reject count. A vendor-neutral container glass consultant carries none of those constraints.

If you want a structured read on where your first ware is being lost, our forming audit gives you a five-day section-by-section analysis with a ranked defect contribution table. Where the Job Change Tool's Live Execution layer becomes relevant is in locking those recovered set points across shifts, so the improvement found in week one is still running in week eight.