To calculate OEE (Overall Equipment Effectiveness), multiply three percentages: Availability × Performance × Quality. Availability is run time divided by planned production time, Performance is actual output against the ideal rate, and Quality is good parts divided by total parts. This guide shows each step with a worked shift example and explains what a good OEE score really is.
What is OEE?
OEE is a single percentage that tells you how much of your planned production time was truly productive. A score of 100% means the machine made only good parts, as fast as it is designed to run, with no stops, for the whole time it was scheduled to run.
No real machine does that. The value of OEE is that it breaks the gap between 100% and your actual score into three kinds of loss, so you can see whether your problem is machines stopping, machines running slowly, or machines making scrap. It works for a CNC cell, a packing line, an injection moulding press or a continuous process line, as long as you can count output and time.
The OEE formula
The standard formula is:
OEE = Availability × Performance × Quality
Each factor is a ratio between 0 and 1 (or 0% and 100%):
- Availability = Run Time ÷ Planned Production Time
- Performance = (Ideal Cycle Time × Total Count) ÷ Run Time
- Quality = Good Count ÷ Total Count
If you multiply those three together, most terms cancel out, which gives a useful shortcut:
OEE = (Good Count × Ideal Cycle Time) ÷ Planned Production Time
In plain words: how much time would it have taken to make only the good parts at full speed, compared with the time you had? The shortcut is a good way to check your arithmetic, but you still need the three factors to know where the losses are.
How to calculate each OEE factor
Availability: are the machines running when they should be?
Start with the shift length and take out planned stops such as meal and tea breaks, or time when there is no production scheduled. What is left is planned production time. Then take out every unplanned or avoidable stop during that time, such as breakdowns, waiting for material, and changeovers. What is left is run time.
Availability = Run Time ÷ Planned Production Time
Performance: are they running at full speed?
Performance compares what the machine actually produced during run time with what it could have produced at its ideal cycle time, the fastest time in which it can make one part under good conditions. Slow cycles and short stops that nobody logs (a jam cleared in 40 seconds, a sensor blocked for a minute) all show up here.
Performance = (Ideal Cycle Time × Total Count) ÷ Run Time
You can also write it as Total Count ÷ (Run Time × Ideal Run Rate), where the ideal run rate is parts per minute at full speed.
Quality: are the parts good first time?
Quality is the share of parts that met specification the first time through, without rework. Parts that are reworked and later passed still count as defects for OEE, because the time spent making them the first time was lost.
Quality = Good Count ÷ Total Count
OEE calculation example, step by step
Here is an illustrative example for one machine over one shift. The numbers are made up to show the method; they are not from any real plant.
| Input | Value |
|---|---|
| Shift length | 8 hours = 480 minutes |
| Planned stops (one 30-minute meal break, two 15-minute tea breaks) | 60 minutes |
| Unplanned breakdown | 27 minutes |
| Changeover between two products | 21 minutes |
| Ideal cycle time | 2 seconds per part (30 parts per minute) |
| Total count | 9,500 parts |
| Rejected parts | 285 parts |
| Good count | 9,215 parts |
Step 1: planned production time
480 minutes − 60 minutes of planned breaks = 420 minutes.
Step 2: run time and Availability
Downtime = 27 minutes breakdown + 21 minutes changeover = 48 minutes.
Run time = 420 − 48 = 372 minutes.
Availability = 372 ÷ 420 = 0.8857, or 88.6%.
Step 3: Performance
At 30 parts per minute, 372 minutes of run time could have produced 372 × 30 = 11,160 parts.
Performance = 9,500 ÷ 11,160 = 0.8513, or 85.1%.
(Using the other form: 2 seconds × 9,500 = 19,000 seconds of ideal time, divided by 372 × 60 = 22,320 seconds of run time, also gives 0.8513.)
Step 4: Quality
Quality = 9,215 ÷ 9,500 = 97.0%.
Step 5: OEE
OEE = 0.8857 × 0.8513 × 0.9700 = 0.7313, or 73.1%.
Check with the shortcut: 9,215 good parts × 2 seconds = 18,430 seconds. Planned production time is 420 × 60 = 25,200 seconds. 18,430 ÷ 25,200 = 0.7313. The two methods agree.
Note that the three factors look healthy on their own. None is below 85%. Yet the combined score is 73.1%, because the losses multiply. This is why OEE is harder to move than any single metric, and why improving the weakest factor, here Performance, usually gives the biggest gain.
The six big losses and how they map to OEE
The losses behind OEE are traditionally grouped into six categories, often called the six big losses. Each one reduces exactly one of the three factors.
| Loss | OEE factor | Typical examples |
|---|---|---|
| Equipment failure (breakdowns) | Availability | Motor failure, hydraulic leak, tool breakage, electrical trips |
| Setup and adjustments | Availability | Changeovers, die or mould changes, warm-up, material shortage, waiting for an operator |
| Idling and minor stops | Performance | Jams, blocked sensors, misfeeds, short stops cleared in a minute or two |
| Reduced speed | Performance | Running below rated speed, worn tooling, operator caution, poor material |
| Process defects | Quality | Scrap and rework during stable production |
| Reduced yield (startup losses) | Quality | Rejects after a changeover or start-up until the process settles |
Minor stops are the easiest to miss. Operators rarely log a 90-second jam, so they vanish into Performance loss. If Performance is low but nobody can explain why, short stops are the first place to look.
What is a good OEE score?
The figure you will see quoted most often is 85% as world-class OEE, sometimes broken down as roughly 90% Availability, 95% Performance and 99.9% Quality. Treat this as a rule of thumb, not a standard. It dates from early work on Total Productive Maintenance and has been repeated so often that it is now taken as fact, but what is achievable depends heavily on the industry, the product mix and how many changeovers you run.
In practice:
- Many plants find their OEE is much lower than they expected the first time they measure it honestly, often somewhere well below 85%. That is normal and is exactly why the measurement is useful.
- A high-mix job shop with frequent changeovers will usually score lower than a dedicated line making one product all day. Comparing the two directly is not fair.
The more useful question is not "what is a good OEE score?" but "is our OEE improving, and which loss is the biggest?" Set your baseline, pick the largest loss and track the trend week by week.
Common mistakes when calculating OEE
Treating avoidable stops as planned downtime
If changeovers, cleaning, waiting for material or planned maintenance that overruns are all put under "planned stops", they disappear from Availability and the score looks better than it is. Keep planned stops to things you have deliberately decided not to use the machine for, such as breaks and unscheduled shifts. Changeovers and setups belong in downtime, because they are losses you can reduce.
Using the wrong ideal cycle time
The ideal cycle time should be the fastest the machine can reliably run the part, usually the nameplate rating or the best sustained rate you have seen. If you use a standard or average cycle time, which already includes some slowdown, Performance will look high and can even go above 100%. If your OEE or Performance ever exceeds 100%, the ideal cycle time is wrong. On mixed-product lines, use the ideal cycle time for each product, not one figure for the whole line.
Measuring over too short a window
An hour's OEE can be 95% or 20% depending on whether a changeover happened in it. Use a whole shift as the minimum unit for decisions, and look at weeks and months for trends.
Counting reworked parts as good
Quality should use first-pass good parts. Reworked parts cost time twice and should be counted as defects in OEE, even if they are shipped.
Averaging OEE across machines
A simple average of machine OEE percentages gives a small, rarely used machine the same weight as your bottleneck. For a line, focus on the constraint machine. For a plant view, weight by planned production time.
How to measure OEE automatically from machine data
Most plants start OEE on paper or in a spreadsheet: operators write down stops and reasons, supervisors enter counts at the end of the shift, and someone builds a weekly report. It works as a start, but it has well-known weaknesses. Short stops go unrecorded, times are rounded, counts are entered late, and the report arrives days after anyone could have acted on it.
The alternative is to take the inputs directly from the machines:
- Run status. A running or stopped signal from the PLC or a gateway gives exact run time and every stop, including the short ones.
- Part counts. A counter from the PLC, or a sensor at the machine output, gives total count without manual entry.
- Reject counts. Rejects from an inspection station, or rejects entered by the operator, give the quality figure.
- Schedule and ideal cycle time. These come from your shift pattern and product master, and need to be kept accurate.
- Stop reasons. Machine data tells you when a stop happened; an operator still usually adds why. A short list of reason codes, picked on a screen at the machine, is much easier to keep up than a paper log.
Most PLCs and gateways can publish these signals over standard industrial protocols. If you are deciding how to connect them, our guide to OPC-UA vs MQTT vs Modbus explains the trade-offs. Once run time and stops are captured automatically, the Availability losses become visible machine by machine, which is also where work to reduce unplanned downtime should start.
This is the kind of data MIE, the ThinklytixAI Manufacturing Intelligence Engine, is built to collect. It takes the readings you already have from your PLC, gateway, historian or edge device over HTTPS, MQTT, OPC-UA or Modbus, without new sensors, and only accepts data from machines and tags you have registered. You can ask it plain-English questions about that data, and it escalates alerts on Slack, WhatsApp or webhook until someone acknowledges them.
For reporting, MIE custom dashboards built by describing them in plain English, such as "output vs target by day, downtime by machine", are next on the MIE roadmap, so a view of the inputs behind OEE can be requested rather than built by hand in a spreadsheet.
Summary
- OEE = Availability × Performance × Quality, or equivalently (Good Count × Ideal Cycle Time) ÷ Planned Production Time.
- In the example, 88.6% × 85.1% × 97.0% gives an OEE of 73.1%.
- 85% is widely quoted as world-class, but it is a rule of thumb. Your own trend matters more.
- Get the definitions right first: what counts as planned, which ideal cycle time you use, and how long a window you measure over.
- Capturing run status and counts directly from machines removes most of the errors that make spreadsheet OEE unreliable.