Manufacturing 13 min read

OEE is one formula. Your production environment decides everything else.

A bottling hall, a machine shop, a kiln and a cleanroom all "monitor production" — and share almost nothing about what to measure first, which assets fail, or where the losses hide. The environment map, sector by sector.

By Frank Guo · Technology & Product Leadership, addanode

TL;DR — Availability × performance × quality is universal; where the points go missing is not. Discrete lines (bottling, packaging, assembly) lose OEE to short stops and changeovers, and count-based monitoring finds it fast. Job shops (CNC, fabrication) lose it to utilisation nobody measures — spindle-on time versus paid hours. Continuous plants (cement, paper, extrusion) barely stop at all, so condition monitoring of the few critical rotating assets is the programme. Batch environments (chemicals, food cooking, pharma) lose time between steps, in cleaning and in waiting, which machine sensors alone never see. Pick the instrumentation to match the loss physics of your environment — the table below maps ten of them — and start with the one number your environment hides best.

First, know which production type you are

Almost every monitoring mistake starts by importing another environment's playbook. The four production types lose time differently, so they instrument differently:

Discrete line Job shop Continuous process Batch process
ExamplesBottling, packaging, assemblyCNC machining, fabrication, tool & dieCement, paper, glass, extrusionChemicals, brewing, pharma, cooking
Where time diesShort stops, changeovers, jamsIdle spindles, setup, waiting for workRate losses and rare, huge stoppagesBetween steps: cleaning, transfers, QC holds
OEE fitTextbook — count in, count outUse utilisation first, OEE laterOEE on rate vs design capacityOEE per batch step; schedule adherence
First signal to wireProduct counts + run stateSpindle/arc-on current per machineThroughput rate + drive currentsStep start/stop times + temperatures
Condition-monitoring weightModerate — many small assetsLow–moderate — spindles, compressorsDominant — few assets, huge stakesAgitators, pumps, utilities

Food & beverage — bottling, dairy, bakery, brewery, meat

South Africa's largest manufacturing sector runs on discrete lines with a process core: a bottling or canning line loses its OEE to micro-stops at the filler and labeller; a dairy adds CIP (clean-in-place) windows that quietly swallow availability; a bakery lives and dies on oven and proofer stability; breweries are batch upstream and discrete downstream; and meat and poultry plants layer cold-chain compliance on top of line efficiency. The environmental constraints are real: washdown chemicals and hose-down pressure demand IP-rated sensors, and the compliance layer — cold rooms, blast freezers, effluent to the municipal sewer — is monitored by the same platform that counts the line (see our cold chain and trade effluent pages). The classic first win: count-based OEE at the line constraint, which in most plants is the filler.

Packaging, printing & converting

Corrugators, flexo and litho presses, laminators, bag-making and label lines share one economic shape: the machine earns only at speed, and changeover plus web breaks are where the day disappears. Monitoring that matters here: run-rate versus rated speed per job, changeover clocking with reason codes, and waste counts at rewind. Because a converting plant's product changes hourly, the win is per-job OEE — which shift, which substrate, which job structure kills the rate — rather than a plant-wide average that hides everything. Our andon guide and shop-floor data collection cover the operator side that makes reason codes survive contact with reality.

Plastics — injection moulding, extrusion, blow moulding

An injection moulding hall is cycle-time arithmetic: every second added to a 30-second cycle across 20 machines is a machine's worth of capacity lost, so the monitoring core is cycle-time per machine per mould against standard, plus reject counts at the press. Extrusion behaves as continuous — melt pressure, barrel-zone temperatures and line speed are the signals. Blow moulding sits between. Plastics also carries a silent cost line worth instrumenting: energy per kilogram processed, since barrels, chillers and dryers make plastics one of the most electricity-intense discrete environments per rand of output — the same per-line energy measurement covered in our OEE solution.

Metal fabrication & machine shops

The job-shop trap: a CNC shop that quotes on machine-hours usually cannot say what fraction of paid hours the spindles actually cut. Spindle utilisation — measured from spindle load or current, no CNC integration required — is routinely a shock the first time it is graphed, and it reprices every quote the shop issues. Around it: setup-time capture per job, arc-on time for welding bays, press-stroke counting in fabrication, and compressor and dust-extraction health as the shared utilities everything depends on. This is the environment where our no-PLC monitoring approach earns its keep — many SA machine shops run machines spanning four decades, and current clamps do not care how old the machine is.

Automotive components

SA's component plants — stamping, welding assemblies, plastic trim, harnesses — inherit their monitoring from their customer contracts: OEE per line as a reportable number, stroke counts and die-protection on presses, weld-cell cycle times, and traceability the OEMs audit. The distinctive pressure is takt discipline: a component plant misses a shift's takt and pays for it in airfreight. Monitoring here is less about discovery and more about defensible numbers — the same "whoever holds the count holds the argument" logic that runs through our MES vs OEE analysis.

Cement, brick, glass & aggregates

Continuous heavy process: a kiln, a glass furnace or a crusher-mill train stops rarely, catastrophically and expensively — so the monitoring programme is condition monitoring. Vibration and temperature on the kiln drive, mill bearings, crusher shafts and ID fans; motor current signatures on everything large; and dust-siege protection for the instruments themselves. The OEE lens still applies, but as rate versus design capacity: a mill running 8% under rated throughput for a quarter is a stoppage nobody logged. Our bearing fault detection guide covers the failure physics; condition monitoring the delivery.

Agro-processing — sugar, grain & feed mills

Campaign industries: a sugar mill earns its year in a crushing season measured in weeks, which converts every hour of unplanned downtime into revenue that never comes back. Grain and feed mills run longer seasons with the same shape — hammer mills, pellet presses, elevators and conveyors, all rotating, all dusty, all monitorable by vibration and current. The campaign economics justify condition monitoring that would look premium in a year-round plant, because the repair window is the off-season and the failure window is never. Conveyor health — the artery of every mill — has its own guide: conveyor belt monitoring.

Pharmaceutical & cleanroom manufacturing

Pharma inverts the priority: the product is often small, but the environment is the compliance object — GMP environmental monitoring of cleanroom differential pressure, temperature, humidity and particle excursions, with an audit trail a regulator will read. Line OEE matters (packing lines especially, where serialisation slows everything), but the non-negotiable layer is continuous environmental records with alarm response times. It is the same telemetry discipline as our water-quality compliance work pointed at air instead of water: measure continuously, alarm on excursion, keep the record that survives an audit.

Chemicals & batch processing

Batch plants lose time between the steps the recipe names: transfers, cleaning validations, QC holds, waiting for a shared reactor. Step-level timestamping — when did each batch actually start and finish each stage, against standard — surfaces a schedule-adherence picture that machine-level sensors cannot see. On the asset side, agitator drives, transfer pumps, and the utilities (steam, chilled water, compressed air) carry the condition-monitoring weight; on the compliance side, effluent and emissions records mirror the trade effluent pattern.

Textiles, clothing & footwear (CMT)

The CMT environment is many small machines and many hands: monitoring is less about spindle physics and more about flow — pieces per hour per line against target, WIP piling between operations, and downtime reasons that are mostly upstream (waiting for cutting, waiting for trims). Simple count-based line monitoring plus an andon escalation covers most of the value; the compressed-air system and boiler are usually the only assets worth condition-monitoring individually.

The environment map on one table

Environment Dominant loss First thing to instrument Condition-monitoring focus Environmental constraint
Bottling / canningMicro-stops at filler & labellerCounts + run state at the constraintFiller drives, conveyors, compressorsWashdown — IP-rated everything
Dairy / breweryCIP windows, batch waitsStep timestamps + line countsPumps, agitators, refrigerationHygiene zoning, CIP chemicals
Bakery / meatThermal stability, cold-chain breaksOven / cold-room temperaturesRefrigeration plant, ovensCold, wet, hygiene audits
Packaging / printingChangeovers, web breaksRate vs rated speed, per jobPress drives, dryersSolvent zones on some presses
Injection mouldingCycle-time creep, rejectsCycle time per machine per mouldHydraulics, chillers, dryersEnergy intensity — meter it
CNC / machine shopIdle spindles, setupSpindle-on current per machineSpindles, compressed airMixed machine ages — go PLC-free
Automotive componentsTakt misses, die damageStroke counts + cycle vs taktPresses, weld cells, robotsCustomer-auditable records
Cement / glass / aggregatesRate below design, rare big stopsThroughput rate + drive currentsKiln drives, mills, crushers, fansDust, heat, vibration sieges
Sugar / grain / feed millsCampaign-season downtimeVibration on mill trainHammer mills, presses, elevators, conveyorsDust (explosive), seasonal windows
Pharma / cleanroomExcursions, serialisation dragEnvironmental telemetry + audit trailHVAC, compressors, purified-waterGMP validation of the monitoring itself
Chemicals / batchBetween-step waits, QC holdsStep start/stop timestampsAgitators, pumps, utilitiesHazardous-area ratings
Textiles / CMTFlow imbalance, upstream waitsPieces/hour per line vs targetCompressor, boilerMany small machines — count flow, not motors

Three rules that survive every environment

  • Instrument the constraint first. Every environment above has one asset or step that sets the plant's pace — the filler, the spindle bank, the kiln, the shared reactor. Points recovered there are the only points that reach the bank.
  • Match the sensor to the loss, not to the catalogue. Count-based signals find discrete losses; current and vibration find rotating-asset decay; timestamps find batch waits. Buying vibration sensors for a changeover problem is the most common way to spend money and learn nothing.
  • The old-machine excuse does not survive contact with a current clamp. Every environment on this page can be monitored without touching a PLC — counts, currents, temperatures and timestamps sit outside the control system, which is why machine age is not the obstacle it is assumed to be. That argument in full: OEE on an old line without a PLC.
FAQ

OEE by environment — common questions

Does OEE work the same way in every factory?

The formula is universal; the measurement is not. Discrete lines measure counts and run state; job shops should measure utilisation before OEE; continuous plants measure rate against design capacity; batch plants measure step times and schedule adherence. Importing a bottling hall's OEE playbook into a CNC shop produces numbers nobody trusts — the environment decides the instrumentation.

What should a food or beverage factory monitor first?

Counts and run state at the line constraint — usually the filler — because micro-stops there set the whole line's OEE. In parallel, the compliance layer: cold rooms and blast freezers on continuous temperature recording, and trade effluent if you discharge to a municipal sewer. Washdown environments need IP-rated sensors from day one; retrofitting weather-sealed hardware costs more than specifying it.

How do you monitor CNC machine utilisation?

Spindle load or motor current, read by a clamp that needs no CNC integration, distinguishes cutting from idle from off. Graphed against paid hours, first-time utilisation numbers are routinely far below what the shop assumed, and they reprice every machine-hour quote. Setup capture and job attribution come next; full OEE only after utilisation is believed.

Why is condition monitoring more important in cement and milling than in assembly?

Because the loss physics inverts: a continuous plant has few critical rotating assets, rarely stops, and a single unplanned failure — kiln drive, mill bearing, crusher shaft — costs more than a year of small line losses in a discrete plant. Vibration and current monitoring on that short list of assets is the whole game, and in campaign industries like sugar the season makes the failure window unaffordable.

What is GMP environmental monitoring in pharma manufacturing?

Continuous recording of the cleanroom parameters GMP regimes care about — differential pressure, temperature, humidity, and particle counts where required — with alarms on excursion and an audit trail a regulator can read. The monitoring system itself falls under validation, which is the real difference from ordinary factory telemetry: the record has to prove not only what happened but that the recorder can be trusted.

Can old machines in these environments be monitored without PLC integration?

Yes — that is the practical point of the whole map. Product counts, motor currents, temperatures, vibration and step timestamps are all measurable outside the control system with clamp-on and bolt-on sensors, so a 1985 press, a mixed-age CNC floor or an undocumented mill train can report OEE and asset health without touching its controls. Machine age changes the sensor choice, not the feasibility.

Find your row. Instrument the constraint.

Tell us which environment on this page is yours. We will tell you what to measure first — and what it would take.