OEE Done Properly: Why Your Number Is Probably Wrong
The short answer
OEE is Availability times Performance times Quality, each as a fraction, so a line that is available 90 per cent of planned time, runs at 95 per cent of rated speed and produces 99 per cent good parts has an OEE of 0.90 x 0.95 x 0.99 = 84.6 per cent. Around 85 per cent is the figure usually called world class for a discrete manufacturing line. If your plant reports 95 per cent, the definition is wrong somewhere, not the plant.
The three factors
Availability = run time divided by planned production time. Planned production time excludes time the plant deliberately did not intend to run: no shift, no orders, planned maintenance. It includes everything that stopped you unintentionally, including changeovers.
Performance = actual output divided by what the line should have produced in that run time at its ideal cycle time. This is where small stops and slow running show up.
Quality = good parts divided by total parts produced. First pass, before rework.
The four ways the number gets inflated
Excluding stoppages from planned time. A breakdown reclassified as planned maintenance disappears from Availability entirely. This is the most common and the most damaging, because the metric stops seeing the thing it exists to see.
Using an achievable rate instead of the ideal cycle time. If the machine is rated at 60 parts a minute and you set the standard at 45 because "we never get 60", Performance is measured against 45 and the missing 15 vanishes. The ideal cycle time is the design rate, permanently.
Counting reworked parts as good. They consumed capacity twice and they were not right first time.
Ignoring micro-stops. A stop of eight seconds happening forty times a shift is over five minutes of lost run time and it is invisible on a clipboard. This is precisely what automated data capture is for.
What it is for, and what it is not for
OEE is a diagnostic for one machine or line over time. Its value is in the decomposition: an OEE of 62 per cent tells you nothing, but 62 per cent made of 0.71 availability, 0.92 performance and 0.95 quality points straight at downtime.
It is a bad tool for comparing two different lines, and a worse tool for comparing plants. Different products, different changeover frequencies and different definitions make the comparison meaningless, and once OEE is used to rank people the definitions start moving.
Taking it from the controller
The reliable version is measured where the parts are made:
- Run state from the PLC, not from an operator's start button.
- Part count from the same counter that indexes the machine.
- Reject count from the reject station or the vision system, not from a tally sheet.
- Stop reasons from a fault word in the PLC mapped to a reason code, so downtime is attributed automatically rather than from memory at the end of a shift.
The last one is the difference between a dashboard people believe and one they argue with. If an operator has to choose a reason from a dropdown an hour later, the largest downtime category in every plant becomes "Other".
The Power BI OEE dashboards course builds the reporting layer, SQL for PLC and SCADA engineers covers getting the data out, and Node-RED for industrial IoT covers moving it from the machine.
Frequently asked questions
Is 85 per cent a target? It is a benchmark for discrete manufacturing, not a target for your line. A process plant, a job shop and a high-changeover packaging line all have different realistic ceilings.
Should changeover count against Availability? Yes, under the standard definition. Excluding it hides the single biggest improvement opportunity in most plants.
What about TEEP? Total Effective Equipment Performance measures against all 168 hours in a week rather than planned time. It answers a different question: how much of the asset you own are you actually using.
