The Five Job Change KPIs That Predict Pack-Rate Recovery

Twelve minutes. That is the number your shift handover report should lead with, and almost certainly doesn't.

Data from IS-machine changeovers across European and North American container glass plants is specific on this point: time-to-first-valid-pack under 12 minutes consistently correlates with two-hour pack rates at or above 92% of steady state. Push that number past 22 minutes and you'll land at 78% or below. That is a 14-point spread sitting entirely inside the window most container glass operations treat as unavoidable settling time (Glass International, Forming Technology Supplement, 2023).

Five KPIs drive post-changeover pack-rate recovery. Most plants formally track one of them. The other four live in shift logs, in experienced operators' handwritten notes, or nowhere at all. That isn't a data problem. It's an accountability problem.

Changeover time is a lagging indicator, not a management target

Every hot-end superintendent can quote their changeover time. The cross-shift variance story is real and worth closing. On many lines that haven't systemised the handover, a 30-60% spread on identical SKUs is normal, and that variance costs real money. But changeover time tells you when the job change ended, not how well the plant will recover for the next four hours. Treating it as the primary changeover KPI is like judging a pit stop by how fast the tyres went on without checking whether the pressures were set correctly.

In 2017 I was running a four-line soda-lime plant and we logged our best-ever changeover time on a 330ml beer bottle to 750ml wine bottle swap. Forty-one minutes, clean execution, the shift supervisor was proud of it. Two hours later, pack rate was sitting at 74% and the hot-end camera was throwing bird-swing rejects at a rate I hadn't seen since commissioning. The changeover time was excellent. The recovery was a disaster, because nobody had set a target for anything that happened after the first gob came down.

Changeover time belongs on your dashboard. It just can't be the only number on it.

The leading pair: TFVP and first-ware quality

TFVP (time-to-first-valid-pack, measured from blank-mould-closed to first container passing hot-end camera inspection) is the single most predictive leading indicator of recovery trajectory. The 12-minute threshold isn't a rough guide. It's a hard split in the data, and what it is measuring underneath is whether your forming team got the gob weight right, quickly.

The target is ±0.3g from nominal within 20 minutes of first gob on new moulds. Every sustained 1g overweight condition adds approximately +120 PPM of blowout defects per section and delays lehr temperature re-equilibration by 8-12 minutes. That re-equilibration lag is the mechanism that turns a gob-weight miss at T+5 minutes into a pack-rate problem still visible at T+90 minutes. It compounds, and by the time the shift report catches it, the cause is already two hours old.

First-ware quality, the second leading KPI, needs to track check defects separately in the first 30 minutes. Check defects peak between T+10 and T+20 minutes post-mould swap because cold metal is contacting hot glass parison. Pre-heating blank and blow moulds to 180-220°C before fitting is the intervention that controls this. Plants that skip it, or apply it inconsistently, see check PPM run at 1,200-2,500 in the early window against a steady-state baseline under 150 PPM. Averaged into the shift total, that peak disappears and looks like normal variation. Tracked as a discrete first-ware metric, it tells you whether mould prep is being executed or just ticked off a checklist.

If your post-changeover KPI report doesn't distinguish the first 30 minutes from the rest of the shift, you are averaging away every diagnostic signal the job change is producing.

Section variance is where recovery goes to hide

Fleet-level OEE is a comfortable number to report. A single section running at 68% efficiency inside a fleet average of 88% generates disproportionate check and bird-swing defects and never triggers a plant-level alarm. The KPI target is all individual sections at or above 88% within 60 minutes of first valid pack. That isn't a soft aspiration. It's the threshold at which fleet recovery becomes predictable.

Bird-swing defects are the tell on NNPB lines during the post-changeover window. The defect signals settle-blow timing error. Settle-blow pressure should sit at 0.5-0.8 bar with counter-blow at 1.0-1.4 bar for lightweight wine containers, and a ±0.05 bar deviation at job start correlates with a +4-7% bird-swing reject rate on the affected section. That's one section. Not visible in your fleet OEE. Responsible for a meaningful share of your post-changeover scrap.

And this is where fleet-level dashboards consistently mislead. NNPB now accounts for an estimated 45% of European wine-bottle production and is considerably more sensitive to gob-weight deviation than blow-and-blow. A ±1g gob error in NNPB produces wall-thickness variation of ±0.15mm, versus ±0.08mm in a blow-and-blow equivalent (Glass Technology Services / Sheffield Hallam University, 2022). Plants running NNPB without section-level disaggregation during the transition phase are working with information that is structurally incapable of finding the problem.

The lehr is the other place section variance hides. Zone 1 entry temperature needs re-profiling to within ±5°C of the weight-adjusted nominal (560-580°C for standard soda-lime containers), and that re-profiling is a timed, owned KPI, not a task assigned to nobody in particular. A 10°C under-temperature condition at lehr entry generates residual stress failures that surface at the cold-end polariscope and get attributed to hot-end forming. Mould redesign conversations start. The actual cause, lehr re-stabilisation lag after a container-weight change, goes unfixed. I've seen this cycle run for three years in the same plant (and on an ageing Emhart 10-section running 1990s analogue-to-servo retrofit controls, distinguishing lehr lag from mould-origin defect is not fast, but the misattribution is identical on modern kit).

Scrap cost per event, not per month

Most container glass plants calculate job-change scrap as a monthly aggregate. By the time it appears on a management report, the plant has run 15 more changeovers and the causal link between a specific job change and its scrap tonnage is gone. Tracking scrap cost as a per-event KPI, assigned in real time to each changeover record, changes what the floor pays attention to and makes the post-mortem actionable rather than historical.

The financial logic sharpens considerably at current US market conditions. Container glass operating rates dropped from 91% in 2021 to 82-84% in 2024, as the Glass Packaging Institute reported in its Q1 2025 market outlook. At lower utilisation, fixed furnace costs are spread across fewer packs per campaign, and every hour of extended post-changeover recovery carries proportionally higher cost impact. For plants absorbing broader SKU portfolios after North American capacity rationalisation, an estimated 8-12% increase in per-machine job-change frequency at facilities that absorbed Ardagh's Milford volume means that arithmetic is compounding.

In Europe, carbon accounting adds a direct monetary figure to every slow recovery. Under EU ETS Phase IV, the hollow-glass product benchmark is 0.441 t CO₂ per tonne of liquid glass, set under Commission Implementing Regulation EU 2021/447. A plant producing 300 t/day that burns an additional 2% fuel across a four-hour recovery window generates approximately 1.1 excess EUA tonnes per changeover event. At 2024-2025 European carbon prices of €60-75 per tonne, that is €66-82 per event before accounting for the scrap glass itself. For California operations under the CARB Cap-and-Trade programme, allowance prices tracked USD 32-40 per tonne CO₂e through 2024-2025, creating the same real monetary penalty for combustion overshoot. This cost exists whether it appears in your changeover KPI or not.

A vendor-neutral view of this KPI set is uncomfortable for some plants. OEM-affiliated consultancies typically benchmark success against published steady-state throughput figures and don't model the shape of the recovery curve at all. Generic Lean-Six Sigma boutiques instrument overall OEE and apply DMAIC to steady-state defect distributions, missing section-level disaggregation during the transition phase entirely. Neither category typically integrates carbon-accounting data into the changeover KPI dashboard, leaving the full financial cost of slow recovery invisible at plant management level.

The 0600 handover is where these KPIs go to die on most lines I audit. Night-shift swabbing data doesn't transfer. Section-level efficiency numbers from the last changeover aren't in the briefing. The incoming hot-end superintendent is starting cold, every time. That isn't a technology problem. It's a process problem that requires a system.

The five KPIs described here sit inside Job Change Tool KPI tracking as event-level data, trended over time, with section-by-section visibility built into the 9-stage lifecycle. If you're working with a container glass consultant on changeover improvement, ask specifically how they track the transition-phase recovery curve. The answer will tell you quickly whether they've run a forming line or just read about one. For plants that want to make this data visible without waiting for an ERP project to deliver it, our digitalisation and reporting work is often where the engagement starts.