An OEE of 60% means 40% of your equipment-hours produce nothing sellable. For a Belgian food packaging plant running three shifts, that translates into roughly 60 hours per week of phantom losses: no clear breakdown, no logged stoppage, no obvious defects. Capacity that evaporates between micro-stops, slow changeovers, reduced speeds and rejects. Before signing the order for a new €1.2 M line "to expand capacity", most production directors discover the capacity was already hidden inside the assets they already owned. The tool to surface it is called OEE, and this is the practical guide you need to start using it.

What OEE actually measures (and what it doesn't)

OEE stands for Overall Equipment Effectiveness. It is not a fad KPI: Seiichi Nakajima formalised it in the 1960s as part of Toyota's TPM (Total Productive Maintenance) framework, and it remains the de facto standard in discrete and process manufacturing.

OEE breaks efficiency down into three factors, each expressed as a percentage between 0 and 100:

Availability (A): percentage of planned time the machine is actually producing. Penalises breakdowns, changeovers and unplanned stops.
Performance (P): how fast the line runs versus its theoretical design speed. Penalises minor stops and reduced speed.
Quality (Q): percentage of good units over total units produced. Penalises rejects and rework.

The canonical formula is:

OEE = Availability × Performance × Quality

Watch out: it is a product, not an average. Three components at 90% do not give you 90%; they give 0.9 × 0.9 × 0.9 = 72.9%. That is why an "impressive-looking" OEE of 85% (Nakajima's world-class benchmark) still means 15% of capacity is gone. The real industrial average sits around 60%. A "good" target for a mature line is 75-80%. Well-tuned bottling lines touch 90%; typical CNC machining 70-80%; complex food packaging 65-75%.

What OEE does not measure: profit per unit, customer satisfaction, ergonomics, energy consumption, or whether the product you build has a market. It is a productive-efficiency indicator, not a business-health one.

How to calculate it: a worked example

Take a beverage filling line in Belgium, 8-hour shift = 480 minutes.

Step 1 — Availability

Total planned time: 480 min
Planned downtime (legal breaks, scheduled CIP cleaning): 60 min
Planned operating time: 480 − 60 = 420 min
Unplanned stops during the shift (labeller breakdown 22 min, conveyor jam 8 min, unscheduled changeover 20 min): 50 min
Actual run time: 420 − 50 = 370 min

Availability = 370 / 420 = 0.881 = 88.1%

Step 2 — Performance

Theoretical design speed: 1,000 bottles/min
Theoretical maximum output during run time: 1,000 × 370 = 370,000 bottles
Actual counted output: 320,000 bottles

Performance = 320,000 / 370,000 = 0.865 = 86.5%

The 50,000-bottle gap is made of micro-stops under 5 minutes and reduced-speed moments that almost nobody logs by hand.

Step 3 — Quality

Total produced: 320,000 bottles
Rejects (mislabelled, underfilled, contaminated): 5,000
Good first time: 315,000

Quality = 315,000 / 320,000 = 0.984 = 98.4%

Step 4 — OEE

OEE = 0.881 × 0.865 × 0.984 = 0.750 = 75.0%

A 75% is reasonable: above the industrial average, below world-class. Key reading: the biggest improvement headroom sits in Performance (those micro-stops) and then Availability (the labeller breakdown). Quality is already near ceiling.

The Six Big Losses (where OEE goes to die)

Classic TPM identifies six losses that destroy OEE. Knowing them turns the aggregate number into an actionable list.

Availability losses:

Equipment failures — Bearing failure, electrical fault, defective sensor. Typical impact: 5-15% of planned time. Example: a servo-motor failure on the case packer halts the entire downstream line for 90 minutes.
Setup and adjustments — Changeovers, calibration, first-good-piece. Typical impact: 5-20%, especially in high-mix plants. A line with 8 daily changeovers of 40 minutes loses over 5 hours per day in setup alone.

Performance losses:

Minor stops and idling — Small jams, sensors needing reset, quick manual adjustments. Each lasts seconds or a few minutes, but they add up. Typical impact: 5-15%. The most underestimated loss because almost nobody logs it.
Reduced speed — The machine runs below its nameplate rate, due to wear, conservative operator setting, or product requiring lower cadence. Typical impact: 5-10%.

Quality losses:

Process defects — Out-of-spec product during normal operation. Typical impact: 1-5%.
Startup rejects / reduced yield — First units after a startup or changeover that fail spec. Typical impact: 1-3%, larger when changeovers are frequent.

When you decompose your total loss across these six buckets, you typically discover that two of them explain more than 60% of the problem. That is where you concentrate effort.

How to start measuring OEE without spending €100k on software

You do not need an enterprise MES to begin. Three viable tiers:

Tier 1 — Manual (€0-1,000): Paper sheet or Excel at the line head. The operator logs shift start, stoppages (cause and duration), good units and rejects. One pilot line over 4-8 weeks gives you an honest baseline. Enough to surface the two or three dominant losses.

Tier 2 — Tablet + PLC (€5,000-30,000): Tablet on the line with software like Ignition, Wonderware, or cloud-MES solutions. Reads PLC counters automatically and only asks the operator to code each stoppage cause. Removes the manual-logging bias and gives real-time data.

Tier 3 — Enterprise MES + IIoT (€50,000+): Full platform with IoT sensors, ERP integration and advanced analytics. Justified only when you already know what to measure and need to scale across 10+ lines.

Practical recommendation: start at Tier 1 on your most critical line for 6 weeks. You will learn more about the process in that time than in six months of vendor brochures. Then decide what to automate.

Critical warning: do not sample only "good hours". Real OEE is measured 24/7 across full planned time — including night shifts, Monday mornings and late Friday afternoons. Cherry-picked data lies beautifully.

Concrete actions to lift OEE by 10-20 points

A well-executed OEE programme delivers +10-20 points in 6-12 months. Not by magic, by prioritisation:

For Availability:

SMED (Single-Minute Exchange of Die): Shigeo Shingo's methodology for reducing changeovers. Reductions from 4 hours to 30 minutes are not unusual on lines that have never worked SMED. Filming a changeover, separating internal tasks (machine stopped) from external ones (machine still running), and standardising usually delivers the first 50% improvement for free.
Preventive maintenance on the 5 most critical assets: bearings, drives, sensors and motors with an inspection and replacement plan based on operating hours, not "when it fails".

For Performance:

Hunt micro-stops <5 min: most plants ignore them. Put a tablet on the line for 2 weeks and count every stoppage >30 seconds. The result usually explains 10-20% of invisible loss.
Speed standardisation: one single set-point per SKU, not "whatever the afternoon-shift operator deems prudent".
Operator training: an experienced operator reacts to a minor alarm in 20 seconds; a novice takes 4 minutes. The gap: 5-8% of OEE.

For Quality:

In-line SPC (Statistical Process Control): control charts on critical parameters (weight, temperature, dimension) that warn before the batch drifts out of tolerance.
Root cause analysis on the 3 most frequent defect modes. The 80/20 rule works here too: three causes typically explain 80% of rejects.

Combined with selective industrial automation at bottlenecks, the compound effect is real.

OEE traps to avoid

Cherry-picking data: measuring only good shifts or favourite lines. A credible OEE covers 100% of planned time.
Reclassifying unplanned stops as planned: if a recurring breakdown is rolled into the plan as "extended cleaning", OEE rises on paper while the plant gets worse in reality.
"Optimising" by cutting preventive maintenance: a 2-point gain at 3 months, a breakdown spike at 9 months and an 8-10 point drop. Classic trap.
Comparing OEE across different equipment types: a rotary bottler and a thermoformer are not comparable. Every asset has its own benchmark.
Failing to connect OEE to money: if a 10-point lift does not translate into €/month recovered, leadership will lose interest. Always express it in cost-hour-equipment and additional sellable units.

Where to start

OEE is not a dashboard, it is a discipline. Three steps this week: pick your most critical line, define availability/performance/quality operationally, and start logging for 4 weeks. Once the baseline is in, attack the two dominant Big Losses — almost always changeovers and minor stops. Expect +10-20 points in 6-12 months if the team commits.

If you want a free OEE assessment on one of your lines with a Six Big Losses breakdown, contact us and we will schedule a no-strings technical visit.

OEE: how to measure and improve your machine efficiency