What a Real Container Glass Plant Audit Deliverable Looks Like

Most audit decks are landfill by month two

Ask a plant manager what an outside audit leaves behind and most will describe the same thing: a 50-slide deck, an executive summary that reads like a press release, and a folder on a shared drive that nobody opens again after the kickoff meeting. That's not a deliverable. That's a receipt.

A real container glass plant audit deliverable does four jobs. It shows you where the variance actually lives, section by section, shift by shift. It ranks the fixes by EBITDA impact, not by whichever defect got the loudest complaint on the floor last month. It forecasts the KPI uplift in numbers a CFO can defend at board level. And it hands the plant a 30-60-90 day plan with a named owner against every line item, not a wish list somebody photographs off a whiteboard.

I've read audits from OEM-affiliated consultancies that amount to an equipment upgrade brochure with a logo swap, and generic Lean/Six Sigma boutiques that import an OEE framework built for an automotive line and never adjust it for refractory decay curves or redox drift. Neither leaves the hot-end superintendent with anything he can action on the next shift. Lean Glass builds it the other way, as a vendor-neutral container glass consultant, the way Zaid Hassoneh ran hot ends for two decades: on the floor at O-I Brisbane from 2005, plant manager by 2019, and running the $220M USD Arglass Yamamura greenfield build in the US before founding this business.

A variance map shows you where the line actually bleeds

The variance map is the part most outside audits skip, because it takes real shift-log reconciliation instead of a plant walk and a clipboard. You pull cold-end inspection data, gob weight logs and section timing against the schedule, then lay it out section by section, across three shifts, on the same SKU.

In 2021 I audited a two-furnace, five-line plant outside Dammam. Corporate OEE read 84% across the board. The variance map told a different story: section 4 was running 6% below the line average on the same job, on every shift, for 11 months straight. Nobody had flagged it because the plant-wide average absorbed it.

The root cause was gob shear timing drift of roughly 3 degrees on an older Emhart 8-section still running cam-driven cut sequencing. A newer Heye Smart H1 line on the same site had servo-driven section timing that would have self-corrected. Gob weight variance beyond ±0.5g target tolerance cascades straight into cold-end thickness rejects, and that's exactly what the inspection data showed once it was lined up section by section instead of averaged across the plant.

Twenty-three minutes. That's how long it took the mechanic to find the drift once we pointed him at the right cam lobe. It had been sitting in the reject data for the better part of a year.

Cord and seed clusters usually tell the same kind of story. They trace back to regenerator checker-pack degradation, and they show up as a slow 0.3-0.8% creep in cosmetic reject rate over six to 12 months as checker brick channels partially block and disturb glass-flow homogeneity. It's invisible on a monthly scrap summary. Plot it against days since the last cold repair and it's obvious.

A punch list ranked by EBITDA, not by who shouts loudest

Once the variance map is built, the punch list mostly writes itself. The discipline is in ranking every finding by EBITDA impact instead of by how dramatic it sounds on the floor. A stone or blister claim from a filler gets attention fast. A 0.4% creep in cosmetic rejects from checker-pack wear doesn't, even though over a full campaign it's often worth more.

Ranked by dollars, not by decibels. A punch list I'd put my name on scores each finding against four things:

• Dollar impact per year at current reject and downtime rates
• Capex required to fix it, from zero to a full furnace rebuild
• Time to payback, in months, not quarters
• A named owner who can action it on the next shift

Baffle marks point at baffle alignment drift almost every time. Check the centring tool reading first (and yes, I know your fitter will say it's fine, check it anyway). It's a five-minute fix that sits near the top of most punch lists precisely because it's cheap and it's being ignored, not because it's expensive.

Furnace pull rate matters more than most punch lists admit. Run consistently above 100-105% of design pull and you accelerate throat and breast wall wear, and a campaign that should run 10-14 years can lose 18 months to 2+ years off the back end. That's not a maintenance line item. That's a capital planning problem, and it belongs near the top of the list even though nobody on the floor is complaining about it yet.

An audit that doesn't rank its own findings by dollar impact is just an expensive opinion.

The KPI uplift forecast has to survive a CFO's questions

The KPI uplift forecast is where audits get sloppy, because it's tempting to promise a number that makes the report look good instead of one that survives scrutiny. The starting point has to be a true availability times performance times quality calculation, not the version reported to corporate.

That gap is bigger than most plants admit. Reported OEE routinely excludes planned job-change downtime from the denominator, which inflates the headline figure by 5-10 percentage points against a true calculation. Once a plant systemises changeover with a proper Job Change Tool mapped to the nine-stage Job Change Lifecycle, that gap becomes visible fast, because changeover time gets tracked as its own KPI instead of buried inside a monthly average nobody questions.

Gulf plants and European plants read this forecast differently, and a deliverable that ignores that is lazy. A Kingdom-based flint or amber line competing for the growth segment the Saudi Industrial Development Fund has flagged under Vision 2030 localisation targets is landing gas at under $2/MMBtu, against €25-35/MWh in the EU, with no ETS-equivalent carbon cost at all. That's a real structural advantage on landed cost into Europe and Africa. It means a GCC operator's uplift forecast should weight capacity and reject-rate wins harder than energy-efficiency capex, because the payback math doesn't clear as fast as it does for a European furnace facing EU ETS Phase IV's declining free allocation under the European Commission's 2026-2030 benchmark reduction.

But the forecast still has to be honest about magnitude. A well-scoped hot-end audit and 90-day corrective programme typically buys +4 to +8 OEE points, and cuts cross-shift changeover variance that runs 30-60% in an unsystemised plant down toward single digits. That's not because the ceiling is higher than that. It's because most plants have never systemised the changeover in the first place.

A 30-60-90 plan with named owners, not just deadlines

The last piece is the 30-60-90 plan, and it's where most engagements quietly die. A punch list without owners is just an inventory. The plan needs a name against every action: the hot-end superintendent owns recipe lock and doesn't let an operator move a set point without sign-off, QA owns moving the coating booth spray check from weekly to per shift, the mould shop owns writing down the mould-fit knowledge that right now lives in one senior mechanic's head.

Shift handover is where a lot of this quietly fails without anyone noticing. On most lines I've audited, the 0600 handover misses the night shift's swabbing data more often than it should, easily seven times out of 10. Nobody writes it down because nobody's ever been made to own it. Put a name against it in the 30-day column and it stops happening within a fortnight.

Look, the data says one thing and the shift log says another, and a 30-60-90 plan is the document that forces the two to agree.

If the last deck you got back had a findings section and nothing else, that's worth fixing before the next one lands on your desk. Our hot end audit is built the way this article describes it: a variance map, a punch list ranked by EBITDA, a KPI forecast that survives board questions, and a 30-60-90 plan the floor can action on day one. The case studies show what that's looked like across furnace ages, machine generations and three continents.