OEE explained: formula, examples and the machine data you need

OEE (Overall Equipment Effectiveness) is the percentage of planned production time that actually produces good parts at the ideal speed, calculated as Availability × Performance × Quality.

A line with an OEE of 100% would run for all of its planned time, at its fastest proven speed, and make no rejects. Real lines lose time to stops, run below their ideal rate and make some scrap. OEE folds those three losses into one number per machine, line and shift, so a production team can see where output goes and check whether a change helped. This guide gives the formulas, a worked shift, the data a machine has to supply and the points where OEE figures go wrong.

OEE tells you how much of the time you planned to produce was used well. It asks three questions about every shift: did the machine run, did it run at speed, and did it make good parts?

Each question matches one family of losses. Breakdowns, changeovers and waits for material reduce Availability. Short stops and reduced speed reduce Performance. Rejects and rework reduce Quality. Lean and TPM literature groups these into the six big losses, and ISO 22400-2 sets out key performance indicators for manufacturing operations management, the family OEE belongs to.

The base is planned production time, which is the shift minus breaks and planned stops. Time when nobody meant to run does not count against the machine. Time lost to missing orders or missing operators is a separate measure, TEEP, covered further down.

There is no single good OEE. The right target depends on the product, the process and where the line starts.

A figure of 85% is often quoted as the benchmark for discrete manufacturing. According to oee.com, it is built from 90% Availability, 95% Performance and 99% Quality, and it goes back to Seiichi Nakajima, who introduced TPM and OEE. The same page says most manufacturers sit closer to 60%, and that three factors of 90% each multiply to only 73%.

Treat these numbers as a reference point. A batch process with long changeovers and a high-speed packaging line do not share a ceiling. The comparison that helps is the same line against its own last quarter, with the same definitions.

OEE judges a line against the time you planned to run it. TEEP (Total Effective Equipment Performance) judges it against all the time in the period, nights and weekends included.

According to oee.com, TEEP equals OEE multiplied by Utilization, where Utilization is planned production time divided by all time. A line with a high OEE and a low TEEP is run well when it runs but sits idle for much of the calendar. That points to a sales or scheduling question, not a maintenance one.

OEE needs four signals from a machine: is it running, how many parts did it make, how many were rejected, and why did it stop. Everything else is calculation.

Running state usually comes from the machine PLC or from a contact on the line. The part counter comes from the PLC or from a photoelectric or proximity sensor at the end of the process. Rejects come from a reject counter, a vision system or manual entry. The stop reason is the hard one. The PLC knows that the machine stopped and sometimes which alarm tripped, but only the operator knows whether it was a changeover, a jam or a wait for material.

Three items of master data sit beside the signals: the ideal cycle time for each product, the shift calendar with planned breaks, and a table that maps PLC status codes to running, idle, stopped or alarm. Without them, the same raw data gives different OEE on different dashboards.

Timestamp every count at the source, not when the dashboard receives it. A network delay otherwise moves minutes from one shift to the next. The polling interval must also be shorter than the shortest stop you want to see, or micro-stops vanish into Performance without a trace.

Citia describes this path on its Factory OEE Solution page, where the CATALYST CT-OG gateway collects status and counts from PLCs, controllers and sensors. A delivered example is the machine connectivity and OEE project at PT Rich’s Product Manufacturing Indonesia in Serang, Banten. There PLCs, Advantech EdgeLink and the iFactory platform feed OEE, line comparison and alarm analysis.

For the protocols between machines and gateways, see OPC UA, Modbus and MQTT compared and the industrial IoT gateway guide.

Start with one line, not the whole plant. A single line with a known bottleneck gives you a baseline within a few weeks and shows which master data is missing.

Agree the definitions first. Write down the rules for planned production time, the ideal cycle time for each product and the list of stop reasons, and have production, maintenance and quality sign them. Then connect the signals, run the system next to a manual count for a week, and close any gap before you publish numbers.

Review the loss picture each week, not only the OEE figure. Two lines with the same OEE can lose output in very different places: one to changeovers, the other to short stops. The fix differs, and only the breakdown by loss shows it. A ranking of stop reasons by lost minutes is usually the first report worth reading.

A dashboard can show a precise OEE that is wrong. The error is almost never in the arithmetic. It is in the inputs.

The notes on the Rich’s project list the controls that matter. Production-per-hour values, shift schedules, status classification and the OEE formula must be agreed as official master data. Manual alarm and defect entry can be late, misclassified or altered, so edit and delete rights should be limited and logged. Manual alarms must be closed on time, or downtime is counted twice. Dashboard figures should be reconciled with production records, quality data and maintenance reports on a regular schedule.

If Performance ever exceeds 100%, the ideal cycle time was set too high, according to oee.com. And a higher OEE does not automatically mean a better business result. Link it to throughput, good output, downtime, production cost and market demand. Raising OEE on a product nobody orders improves nothing.