A line at 90% OEE can still be a weak commercial result if it is making the wrong or a low-priority product. OEE also excludes utilisation, the human factor, and energy and resource cost by design, which limits what the headline figure can tell you.
Changeover, CIP and allergen windows are large losses that get classified inconsistently as planned or unplanned, and the 'world-class' benchmark is the wrong target for a hygienic high-SKU line.
The metric's design limits are permanent, but poor input data is an engineering problem. Machine state, counts and rejects read directly from the equipment are timestamped when they happen; operator-entered figures are reconstructed from recall.
OEE is useful, but it is routinely over-trusted. On most Australian manufacturing dashboards the figure drives capital decisions, shift KPIs and improvement priorities, yet on a multi-SKU food and beverage line a single figure averages large structural losses into invisibility. Metromotion Controls builds production-data and OEE systems for manufacturers across Australia, and this guide sets out where OEE misleads and where the real limitation sits.
There are three distinct failure modes: what OEE excludes by design, what it distorts on food and beverage lines specifically, and where the input data quietly breaks. The first two are limits of the metric itself. The third is an engineering problem that can be solved.
This post supports our industrial data and IIoT service, where OEE, downtime reporting and historian data are built from validated plant signals rather than operator recall.
OEE is the product of availability, performance and quality. That makes it a clear single figure, but several things that matter to a business are excluded by construction.
Complementary metrics exist for some of these. TEEP captures schedule and utilisation loss; OAE and OLE bring in asset and labour effectiveness; statistical process control addresses variation. They supplement OEE rather than repair its blind spots.
OEE rests on a published international standard. ISO 22400-2:2014 defines OEE and its three constituent factors as manufacturing operations management key performance indicators, placing them at Level 3 of the IEC 62264 (ISA-95) functional hierarchy, the layer between plant-floor control and business systems. Anchoring an OEE programme to that definition gives a site one agreed formula rather than a house interpretation that quietly differs from the line next door.
The calculation itself is straightforward:
OEE = Availability × Performance × QualityEach factor is a ratio between 0 and 1, and the three multiply. The often-quoted 85% "world-class" figure decomposes as 90% availability, 95% performance and 99.9% quality, which industry guidance such as Vorne presents as a benchmark drawn from discrete manufacturing. The multiplication is the part that gets missed: because the factors compound, a line at 90% on each of the three sits at roughly 73% OEE, so apparently healthy individual numbers can still produce a modest headline figure.
OEE measures performance against planned production time. Total Effective Equipment Performance (TEEP) extends it with a fourth factor, Loading, the share of all calendar time the equipment is actually scheduled to run:
TEEP = Loading × OEE = Loading × Availability × Performance × QualityA site running one eight-hour shift across a five-day week is loading the asset for about a quarter of the 168 hours in a week, so its TEEP is capped near 24% even at flawless OEE. TEEP reframes the question from how well the line ran when it ran to how much of the asset's capacity is being used at all, which is often the more relevant question when a business is weighing a capital purchase against adding a shift.
The structural limits above apply everywhere. On Australian food and beverage lines, a second set of distortions is specific to how these plants run.
Changeover loss can be attacked, not just measured. The discipline is SMED (Single-Minute Exchange of Die), developed by Shigeo Shingo and documented by the Lean Enterprise Institute. Its core insight is to separate setup into internal tasks, which can only be done while the machine is stopped, and external tasks, which can be done while it is still running the previous job. Converting internal tasks to external ones, and streamlining what remains, drives changeovers toward the SMED target of single-digit minutes. On a high-SKU line, every minute taken out of a frequently repeated changeover is recovered availability that flows straight into OEE, so SMED work tends to move the metric more reliably than chasing a higher benchmark.
In hygienic processing, cleaning is not only an availability loss; it is an obligation that has to be evidenced. Australian food and dairy manufacturers run HACCP-based food safety programmes, and the regulator, customer and certification audits built on them expect records showing that cleaning and sanitation were carried out to specification. The same CIP events that depress availability are the events those records describe. Capturing CIP cycles in the control and historian layer lets one timestamped record serve both purposes: the cleaning loss becomes visible and consistent in the OEE calculation, and the cleaning evidence comes from the machine rather than a manually kept log.
OEE is frequently used to rank lines or shifts against each other, and the comparison is usually invalid. Different OEM machines report state, counts and rejects differently, so two lines can compute OEE on subtly different definitions of the same thing. Different shifts can apply different planned-downtime rules, so the same event lands as planned on one shift and unplanned on another. And comparing a single-SKU line to a high-changeover line is comparing two different operating realities, so the figures are not comparable even though they share the same units. Making the numbers comparable means first standardising how each machine reports state and counts, which is harder on multi-OEM lines than it looks.
The design limits of OEE are permanent. The input-quality limit is not, and it is where an integrator, rather than a software vendor, adds value.
Operator-entered downtime is a reconstruction of an event rather than a measurement of it. A reason entered hours later depends on recall and shift pressure, and it can be edited afterwards. Machine state, by contrast, transitions in the controller in real time: when an interlock trips and a motor stops, the PLC registers it immediately, and when a reject sensor activates, the count increments as the reject leaves the line. Both are timestamped at the moment they happen and far harder to dispute than a figure entered later from memory. That signal layer is read and standardised in the PLC, SCADA and HMI layer before it ever reaches a dashboard.
A typical packaging line has six to twelve machines from different OEMs, each with its own controller and its own way of reporting state and counts. Some expose a clean PackML state model a line controller can read directly; others expose a basic running contact, or a totaliser that has to be polled and differenced.
The mechanism that makes cross-OEM state consistent is PackML, the OMAC machine-state model published by ISA as technical report ISA-TR88.00.02. PackML defines a standard 17-state model, covering states such as Idle, Execute, Held, Suspended, Aborted and Stopped, together with a standardised set of data structures called PackTags. When each machine implements PackML, a line controller reads the same state names and the same count and reject tags from every machine regardless of OEM, which is exactly what a defensible availability, performance and quality calculation needs. Where a machine does not natively expose PackML, the integrator's task is to map its native states and counters onto that model so the line computes one consistent OEE rather than averaging several incompatible definitions.
Standardising those signals is not a configuration exercise. It requires understanding how each machine actually works internally, not just connecting to its network interface, which is the core of multi-OEM systems integration. In high-volume FMCG plants especially, that engineering layer determines whether an OEE programme produces numbers a business can act on.
OEE is a useful metric with real and permanent limits. It measures efficiency rather than effectiveness, it excludes utilisation, people, energy and cost, and on multi-SKU food and beverage lines it distorts around changeover and cleaning and invites comparisons that do not hold. None of that is fixed by better software or a higher target.
What can be fixed is the signal layer. The work that makes OEE trustworthy sits between the equipment and the dashboard, where some signals are read directly and others are filled in from memory, and where multi-OEM lines have to be standardised properly rather than approximated. For a business that uses OEE to make decisions, that is where defensibility is won or lost.
Tommy Kim writes for Metromotion Controls, a Melbourne control systems integrator delivering PLC, SCADA, controls integration and commissioning for food, beverage, dairy and FMCG manufacturers across Australia.
OEE, downtime, historian and production reporting systems built from reliable plant signals.
Line integration, OEE data and controls upgrades for high-volume manufacturing sites.
Automation, traceability, CIP, SCADA and production data for Australian food and beverage plants.
How beverage automation reduces variability: recipe and batch control, versioned recipes you can trace back in time, work orders as a single live source of truth, and a historian that records exactly how every batch was made.
Key point
A single live source of truth
Published 3 June 2026
How CIP automation cuts trade waste and chemical cost with endpoint-driven phases, protects chemical tanks, verifies coverage and records audit evidence.
Key point
Endpoint-driven phases cut cost without weakening the clean
Published 9 May 2026
Practical guide to food and beverage automation in Australia: ISA-88 batch control, CIP, pasteurisation, hygienic design, traceability and OEE.
Key point
Process and packaging automation are different problems
Published 13 Apr 2026
Map out scope, delivery approach and what to have ready before the first conversation. Answer a few questions and Metromotion Controls returns a tailored scoping brief on screen.