Container Glass KPI Dashboards Operators Actually Use

A dashboard with forty tiles on it is not a KPI system. It's a plant manager's Christmas tree, and I've spent twenty years watching operators walk straight past every branch of it on their way to the swabbing log taped inside the section cabinet. The dashboard cost real money. Nobody on shift trusts it enough to make a call from it.

Most dashboards get built for the person who signed the purchase order

In 2019 I audited a five-line plant in the GCC running on subsidised natural gas feedstock, fuel costs 30-50% below what a European melter pays for the same tonne pulled. That's a real advantage. It's also a trap, because cheap gas masks poor thermal-efficiency numbers that would trigger a corrective-action review anywhere else. The plant's dashboard showed a clean 89% blended efficiency and a green furnace-health tile. Nobody had looked at pull-rate-versus-energy-use as a single trade-off in over a year, because the two lived on separate charts, which is how most OEM-supplied systems present them.

That's the pattern. OEM-affiliated consultancies bundle the dashboard with the furnace or IS-machine sale, so the metrics default to equipment-warranty and uptime concerns rather than anything a gob operator can act on mid-shift. Shift structure on a typical hot end runs a shift supervisor, batch-house operator, furnace/gob operator, IS-machine mechanics and a cold-end QA lead. The dashboards these plants buy get configured around the plant manager's weekly review cycle, not the decision window any of those five people actually work in.

Pack-to-melt is the KPI that tells you what happened to the glass

Pack-to-melt, saleable ware packed against tonnes of glass pulled from the furnace, is the number that cuts through the noise. It's not glamorous. It's the one number that can't be dressed up by a good week on the cold end.

Here's the part most dashboards hide: a plant can hold 95%+ pack-to-melt yield while quietly running redraw rates above 8-10%, and that gap erodes furnace campaign life long before it shows up as a yield problem anyone escalates. Gob weight control is normally specified at ±0.5-1.0% of target. Drift beyond that tolerance shows up downstream as overpress defects and check cracks well before it dents a top-level OEE figure. If your dashboard reports OEE and nothing underneath it, you're reading the outcome and missing the cause. Our OEE glossary entry breaks down why the blended number was never designed to carry this much weight on its own.

What the blended reject percentage is hiding

Cold-end inspection systems, Iris, Inex, Tiama-type units, generate reject data by defect code in real time. Checks, birdswing, spikes and thin wall are four completely different problems with four different fixes, and most purchased dashboards collapse them into one blended reject %. Hot-end supervisors rarely see the code-level breakdown at all.

• Checks point at cold-end handling and swing-plate timing, not the furnace
• Birdswing points at gob shape and takeout timing on the IS machine
• Spikes point at mould wear or lubricant residue in the finish
• Thin wall points at plunger drift or blank-side cooling imbalance

Cord and seed defects trace further back, to batch-house segregation and refiner temperature drift. A stable line holds seed count below roughly 0.3 seeds/cm² in the critical finish zone. A struggling line runs three to five times that without ever triggering a batch-house review, because nobody's dashboard flags seed count as a leading indicator. It just shows up two days later as a customer complaint.

A dashboard that shows you the defect happened is a record. A dashboard that shows you the defect three shifts before it happened is a KPI system.

One source of truth beats two disconnected systems

Forehearth conditioning is normally held to ±2°C across zones. A spread beyond 5°C between channels is a leading indicator of stone and blister defects 24-48 hours before cold-end rejects spike (and yes, I know your fitter will tell you the forehearth's within spec, pull the trend log anyway). But almost nobody reconciles that hot-end variable against the cold-end defect code in the same view. Two systems, two logins, and the cross-reference happens in someone's head at the 0600 handover, if it happens at all. On most lines I've walked, that handover misses the previous shift's swabbing data seven times out of ten.

Twenty-three minutes. That's what a Queensland plant lost every changeover to a step nobody had written down, because the record lived in a supervisor's head and not in the system of record. Not a furnace problem. A handover problem, and the dashboard never once flagged it because it wasn't built to track that stage at all.

And this is where regulatory pressure starts to matter even if you've never thought about it as a KPI issue. Under the EU ETS Fit for 55 revision, the emissions cap falls 4.3% a year from 2024 to 2027 and 4.4% a year from 2028 to 2030, tightening the free allowances available to energy-intensive melters. In the US, EPA's Greenhouse Gas Reporting Program under 40 CFR Part 98 Subpart N already forces facility-level furnace CO2 disclosure above the reporting threshold, which means the pull-rate-versus-energy trade-off your dashboard is hiding in two separate charts is now a public number somewhere. O-I's Fit to Win programme targeted roughly $150M USD in annualised savings partly through furnace-count reduction, and Ardagh closed its Madera, California plant in 2024 citing structurally lower wine-segment demand. Plants under that kind of margin pressure can't afford a dashboard that only tells them what already happened.

What one source of truth looks like on the floor

This is the same discipline behind the KPI Tracking layer in our Job Change Tool, which ties changeover time, first-ware quality and section-level variance to the same timestamp instead of three separate reports nobody cross-checks. It maps to the same stages as the 9-stage Job Change Lifecycle, so a hot-end superintendent who owns recipe lock sees the same data an operator sees on the floor, not a cleaned-up version three days later.

Look, the data says one thing and the floor says another most weeks, and that gap is exactly what a dashboard should be closing, not hiding. If your KPI system can't tell a gob operator why a section just tripped, it's a reporting tool, not a management tool. We built our digitalisation and reporting service around that distinction, because a vendor-neutral container glass consultant has no reason to protect an OEM's efficiency number over your actual furnace campaign life.

So the real question isn't whether your plant has a dashboard. It's whether anyone on the floor would notice if it went dark tomorrow.