Asset Positioning Across a Multi-Plant Container Glass Network

I sat in a control room in Sharjah in 2018 looking at pull-rate data from four furnaces across two sites, and the plant manager beside me was defending a furnace that had no business being defended. It was running at 84% of design pull, scrap was pushing 7%, and the SKU mix on it overlapped almost exactly with a better-performing furnace forty minutes down the road. He wanted a rebuild quote. What he needed was a reallocation plan. That gap, between what a single-plant view tells you and what the network actually needs, is the whole game in asset positioning.

Why furnace pull rate decides who keeps the SKU, not who asks first

Most groups still make asset decisions furnace by furnace, plant by plant, with each site fighting for its own capex. That's backwards. The core input to any serious network model is pull-rate utilisation, actual tonnes per day against rated design tonnes per day, and the pattern is consistent across regions I've audited: furnaces running sustained pull below 90% of design rating are the first candidates for reallocation or idling when a group finally rationalises its footprint. Below that number you're not looking at a maintenance problem. You're looking at a network-positioning decision dressed up as a maintenance problem.

Refractory life matters just as much as pull rate, and the two have to be read together. A typical end-fired regenerative container furnace runs a ten to fourteen year campaign between cold repairs, and once a furnace drops under eighteen months of remaining refractory life it becomes a binary choice: run it to failure on its current SKU mix, or commit real capital to a rebuild. Nobody should be making that call without first asking whether the SKUs on that furnace even belong there anymore.

A furnace with eighteen months of refractory life left isn't an engineering question. It's a portfolio question wearing an engineering coat.

Mould rotation and campaign planning aren't separate problems from SKU allocation

Job change downtime is the biggest scheduled-loss category most plants track against OEE, typically 45 to 90 minutes per section every time a mould set turns over. Networks that cluster similar bottle families onto the same IS machines, instead of scattering look-alike SKUs across sites for historical reasons nobody can explain anymore, cut annual job-change hours by 15-25%. That's not a hot-end tweak. That's a network design decision, made months before anyone touches a mould.

This is exactly where our asset positioning work sits, and it's why we built the Job Change Tool the way we did. A systemised Job Change Tool only pays off if the SKU sitting on that IS machine belongs there in the first place. If you're clustering the wrong bottle families because the campaign plan was set by whoever shouted loudest at the quarterly ops review, you're optimising the wrong changeover. Get the allocation right first, then let execution discipline do the rest.

Three things I check on every multi-plant campaign review, in this order:

• Hot-end scrap rate on the candidate furnace, against a 3% best-in-class benchmark
• Redraw and blank-mould temperature differential across the IS machine, against the 5-8°C threshold that drives section-to-section weight variation
• Forehearth thermal stability, since ±2°C swings are the leading cause of checks and blisters on light-weighted finishes

A plant that fails two of those three isn't a rebuild candidate. It's a reallocation candidate, and the sister plant taking the SKU had better be checked against the same three numbers before anyone signs off.

The regulatory and cost layer most asset models never touch

Gulf capacity exists partly because the economics are different, and if your model doesn't account for that you'll misprice every rebuild-versus-relocate decision in the region. Sharjah's cluster of container furnaces, including Gulf Glass and the Al Ahsa-linked capacity, runs on older Siemens-style regenerative end-fired designs that would be marginal almost anywhere else. They stay economically viable because UAE industrial gas pricing sits well below IOC benchmark rates. Layer Saudi Vision 2030's localisation push under Made in Saudi and NIDLP on top of that, and GCC producers are under real pressure to add flint capacity rather than lean on imports from Egypt or Turkey, with regional demand tracking roughly 5-6% CAGR into the late 2020s on soft drinks, water and the UAE's duty-free spirits channel.

Europe and the US are fighting a different battle, and it's worth knowing both if you're advising a group with assets in more than one region. Under EU ETS Phase IV, the linear reduction factor tightens free allocation by 4.3% a year from 2024, which is steadily pushing plants toward part-purchased allowances by 2027-2030, and it's a big part of why O-I's Fit to Win programme has closed and idled furnace capacity across France, Italy and Germany. In the US, groups weighing domestic furnace capex now have to price in Vitro's Mexican network as a genuine USMCA-qualifying nearshoring alternative, not a theoretical one. None of this changes the mechanics of pull rate and campaign life. It changes what a rebuild actually costs you to defend.

What generic consultancies get wrong on this

OEM-affiliated consultancies tend to frame asset positioning as a rebuild-timing and capex-sizing exercise, full stop. That's not surprising, their commercial relationship depends on capital projects, not on the slower, lower-fee work of re-optimising an existing footprint. And Lean or Six Sigma boutiques usually go the other direction, optimising single-plant OEE and changeover metrics in isolation and missing the network-level trade-off entirely: concentrate similar bottle families on fewer furnaces and you cut job-change losses, but you also add freight and breakage risk if that furnace sits three hundred kilometres from the filling line.

This is why we insist on staying vendor-neutral. A container glass consultant with no furnace-builder or OEM relationship to protect can put the SKU-to-plant allocation question on the table honestly, including the uncomfortable version where the answer is close a line, not rebuild it. Zaid Hassoneh ran this exact trade-off at O-I Brisbane before running the $220M USD Arglass Yamamura greenfield in the US, and the lesson carried across both: the furnace campaign clock and the SKU allocation clock have to be read on the same page, never separately.

Not every asset review needs a full network model on day one. Sometimes it starts with a single plant question, framed correctly, that turns into a network answer six months later.

Where to start

If your group is carrying a furnace with sub-90% pull utilisation and nobody's willing to say the word 'reallocate' out loud yet, that's usually the first conversation worth having. Our strategic advisory work exists for exactly that moment, before the rebuild quote lands on someone's desk and the decision gets made by capex deadline instead of by data.