The container glass post-mortem stage nobody runs

It's six in the morning and the night crew has already logged off. The colour change that started at ten the night before finished somewhere around two, the ware has been running clean for four hours, and the scrap bin from the changeover is still sitting by section four, half full, untouched. Nobody is going to look at it. The line is up, the pack looks fine, and the day shift has its own job change to prep for by noon.

Why stage nine is the first thing the schedule eats

The Job Change Lifecycle runs nine stages: plan, prep, line-down, mould change, recipe load, ignition, first ware, stabilise, post-mortem. Eight of those have someone standing over them with a stopwatch, because if they slip, the line doesn't run. Stage nine has no such pressure. The ware is already selling. Nobody loses their afternoon if the post-mortem doesn't happen, so on a busy roster it's the first thing that gets cut. Not because anyone decided it wasn't important. Because nothing forces it.

A routine mould swap on a double or triple-gob IS machine is a 45-90 minute physical job. Getting to first good ware, with gob temperature stable and the mould properly through its pre-heat window, usually runs two to four hours on a line that's set up well. Add a colour or feeder-forehearth change and you're looking at six to eight-plus hours before you'd trust the pack. That's a long stretch of variable operation, and it's exactly the stretch a post-mortem is supposed to capture. Skip the review and you've thrown away the only record of what happened during the riskiest part of the run.

It's not complicated. It's just work most rosters don't leave time for.

What the first hour after restart actually looks like

Baseline reject rate on a well-run line sits at 2-4%. In the first hour after a job change, that number commonly jumps to 10-18%, and the defect tally is rarely a mystery if anyone bothers to pull it.

• Cold mould checks from thermal-shock cracking, usually traced to a pre-heat that never hit the 480-550°C target
• Birdswing and neck-ring marks from mechanism timing that wasn't reset after the swap
• Gob weight drift beyond the ±1-2% job card tolerance, the single most common trigger for a stretched stabilisation window
• Seeds and blisters carried over from glass left standing in the feeder bowl
• Wash-board or lap marks from lubrication reapplied on a compressed schedule

Every one of those is checkable in real time. The mould shop technician verifies pre-heat with an infrared gun or an embedded thermocouple, not by feel (and yes, I know the fitter will tell you the mould's fine, check it anyway). The hot-end superintendent holds sign-off authority on first-good-ware release, which means nothing ships to the pack until that number sits inside spec, not just close to it. Holding the line at first ware isn't the same as reviewing why the line needed holding in the first place. That review is stage nine, and it's the one nobody's job depends on finishing.

The same three defects, plant after plant

In 2014 I was running a triple-gob line outside Brisbane on an old Emhart still working off 1980s cam timing, and we had the same checking defect turn up on three separate colour changes inside five weeks. Nobody had written down what caused it the first time, so nobody knew whether it was mechanism wear, a pre-heat curve that had drifted, or the cullet ratio in that batch. We fixed it the same way three times, by trial and error on the floor.

And that's the actual failure mode. Not that plants lack smart people on the hot end. They do. It's that the knowledge lives in someone's head, or a notebook in the mould shop, or a shift-handover conversation at six in the morning that misses the swabbing data more often than it captures it, on the order of 70% of handovers on a line I haven't audited yet. Not a furnace problem. A handover problem.

A post-mortem that doesn't get written down isn't a post-mortem. It's a shrug.

Job-change downtime is commonly benchmarked at $6,000-15,000 USD in lost production per furnace-line hour, depending on container size and market. Three hours over target. Real money, not an abstraction. Most plants never reconcile that figure against the specific changeover that caused it, which means the same root cause gets to bill the business again next month.

What the SKU Library is meant to capture

Our founder, Zaid Hassoneh, started on the floor at O-I Brisbane in 2005 and won O-I's Most Improved Job Change Plant, Globally award in 2016. That result didn't come from new equipment. It came from making stage nine mandatory, every time, on every line.

This is the gap the Job Change Tool is built to close, starting with the SKU Library. Every recipe, mould set, and forming spec gets versioned and locked, so the mechanism settings used on a given job aren't a guess reconstructed from memory. When a post-mortem does run, the defect tally, the actual changeover time against target, and the mechanism settings in play get written back into that record and carried into the next job card for the same SKU. The 9-stage Job Change Lifecycle only compounds if stage nine actually closes the loop back to stage one.

This matters more in some regions than others right now. GCC container glass plants, particularly the fragmented export bottling programmes running out of Saudi Arabia and the UAE, are seeing more unplanned job changes than their beverage-heavy European counterparts, simply because the SKU runs are shorter. Add rising Gulf pharma localisation policy pushing Type I and Type II vial work onto lines that were set up for long beverage campaigns, and you've got plants running two to three times the job changes per month they were designed around. A plant doing that many changeovers without a systemised post-mortem isn't managing variance. It's accumulating it.

Why generic advice doesn't fix this

An OEM-affiliated consultancy will talk to you about mechanism settings and machine upgrades, because that's what they sell. A generic Lean or Six Sigma boutique will hand you an SMED framework built for a stamping press and expect it to work on a furnace, and it won't, because it ignores the minimum safe mould pre-heat and gob-stabilisation physics that actually govern first-good-ware time. Neither one owns the cross-functional handoff between mould shop, quality, and the hot-end team, which is the exact point where the same defect gets re-introduced job after job.

That's the gap a vendor-neutral container glass consultant is supposed to close. Not selling equipment. Not importing a framework built for another industry. Just making sure the post-mortem happens, gets written down, and actually changes the next job card.

So next time your line hits stable pack after a job change, ask the question that matters. Not "did we get there," but "did anyone write down why it took as long as it did?" If the honest answer is no, stage nine is still the one costing you money nobody's tracking.