Key takeaways
- OEE = Availability × Performance × Quality: the single number for how productive your planned time really is.
- 85% is often called world-class for discrete manufacturing (traced to Nakajima, Introduction to TPM, 1988); published plant studies found average OEE of about 55% to 65% (sources: Ljungberg, 1998; Hedman et al., 2016).
- OEE losses sort into six big losses across availability, performance and quality.
- Hand-kept OEE is often overstated, because short micro-stops rarely make it onto the log. Calculating OEE from good units catches them.
Overall Equipment Effectiveness (OEE) is the standard measure of how productive a machine or line really is. In one number, it tells you how much of your planned production time is genuinely productive: making good parts at full speed with no stops. An OEE of 100% is perfect production: only good parts, as fast as the machine can run, with zero downtime.
It's powerful because it's honest. A line can look busy and still have an OEE of 55%, and that gap is money. Here's exactly how it works.
The OEE formula
OEE is the product of three factors:
OEE = Availability × Performance × Quality
- Availability = Run Time ÷ Planned Production Time.
Run Time = Planned Production Time - all stop time (breakdowns, changeovers, every stop). - Performance = (Ideal Cycle Time × Total Count) ÷ Run Time.
Captures speed loss and short stops, running slower than nameplate. - Quality = Good Count ÷ Total Count.
The share of parts that pass first time (rework counts as a defect).
Where every planned minute goes
OEE starts from the clock. Every minute of a machine's day belongs to exactly one bucket, and OEE is what is left once the losses are taken out of the time you planned to run. This is the time model R. C. Hansen lays out in Overall Equipment Effectiveness (Industrial Press, 2001), built on Nakajima's TPM formulas:
| Bucket | What goes in it | Counts against |
|---|---|---|
| Calendar time | Every minute of the clock: 1,440 a day, 525,600 a year | TEEP only |
| Excluded time | Time planned ahead as not for production: no orders, holidays, planned shutdowns, planned meetings, trials whose product will not be sold | Nothing (left out of OEE) |
| Planned stops | Changeovers, standard tests, planned material loading | Availability |
| Induced stops | Waiting for material, people or information, or for the machine upstream or downstream | Availability |
| Downtime | Unplanned: equipment failures, operator or procedure errors, material or process quality problems | Availability |
| Speed loss | Running, but below the ideal rate, including short stops nobody logs | Performance |
| Quality loss | Running at speed, making product that is not good first time | Quality |
| OEE | Good units × ideal cycle time: the time it would have taken to make only the good units at full speed | The result |
Two rules keep the model honest. There is no "other" bucket: a minute nobody can explain is still a loss, and it belongs somewhere. And the losses plus OEE always add up to 100% of planned time, which gives you a built-in check on your data (shown with the example below).
A worked example
Take an 8-hour shift (480 minutes) with 60 minutes of total stops, an ideal cycle time of 1 second per piece, 22,000 pieces produced and 500 rejects:
| Factor | Calculation | Result |
|---|---|---|
| Availability | 420 ÷ 480 | 87.5% |
| Performance | (1 × 22,000) ÷ 25,200s | 87.3% |
| Quality | 21,500 ÷ 22,000 | 97.7% |
| OEE | .875 × .873 × .977 | 74.7% |
Three ways to calculate OEE, and why they must agree
The same shift can be calculated three ways. If the answers disagree, one of them is missing a loss, and Hansen's rule is to trust the lowest result until you find it.
- From the three factors: 87.5% × 87.3% × 97.7% = 74.6%. Rounding each factor costs a tenth of a point here.
- From event records: add up the logged stops, the run time at each speed and the rejects, and rebuild the factors from them. Slower, but it tells you which loss to fix.
- From good units: good count × ideal cycle time ÷ planned production time = 21,500 × 1 s ÷ 28,800 s = 74.7%. It needs only three numbers, and a loss you forgot to log cannot hide from it.
Method 3 is the one to start with on a line that has no data system yet: planned time from the schedule, good units from the count that left the line, ideal cycle time from the machine spec.
Check: the losses and OEE add up to 100%
| Share of planned time | Calculation | Result |
|---|---|---|
| Availability loss (logged stops) | 3,600 s ÷ 28,800 s | 12.5% |
| Speed loss and short stops | (25,200 s − 22,000 × 1 s) ÷ 28,800 s | 11.1% |
| Quality loss | 500 × 1 s ÷ 28,800 s | 1.7% |
| OEE | 21,500 × 1 s ÷ 28,800 s | 74.7% |
| Total | 100.0% |
If your own numbers do not add up to 100%, a count or a time is wrong. The usual suspects: a reject count that left out product put on hold, a run time taken from the schedule instead of the machine, or an ideal cycle time copied from an old spec sheet.
Skip the spreadsheet: our free calculator does this live and overlays the world-class benchmark.
OEE, TEEP and utilization are not the same number
| Metric | Divides by | Question it answers |
|---|---|---|
| OEE | Planned production time | How well do we run when we plan to run? |
| TEEP | All calendar time | How much of the machine's full potential do we use? The number for capacity and buy-or-not decisions. |
| Asset utilization | All calendar time | What share of the clock is the machine running? It ignores speed and quality, so it is not OEE. |
Run the example line one 8-hour shift a day, five days a week, and TEEP = 74.7% × (40 h ÷ 168 h) = 17.8%. The line runs well in the hours it is scheduled and stands idle for most of the week, which is a scheduling question before it is a maintenance one. Work it out for your own line with the TEEP calculator.
Which ideal cycle time, and what counts as a good unit
Most OEE arguments are really arguments about these two inputs. The rules below follow Hansen:
- Ideal cycle time is the fastest rate the machine, or the slowest machine on the line, is designed and proven to hold for that size or format. Not last month's average. Use one ideal rate per product family, so a hard product shows up as lower performance, which is exactly what you want to see when you price it.
- No ideal rate on file? Take the best 4 hours out of the last 400 hours of running, and replace it once the real figure is known.
- Performance above 100% means the ideal rate is set too slow. Fix the rate; OEE never passes 100%.
- Good units met spec the first time. Product on hold, product likely to be scrapped and anything reworked are not good units. If a batch is found bad after it left the line, move it to quality loss.
- Several lines or shifts: weight each OEE by its planned time, never take a plain average. Line A at 60% over 400 minutes and line B at 80% over 80 minutes average to 70%, but the honest figure is (400 × 60% + 80 × 80%) ÷ 480 = 63.3%.
What is a good OEE score?
The widely accepted benchmarks for discrete manufacturing:
| OEE | Rating | What it means |
|---|---|---|
| 100% | Perfect | Only good parts, full speed, no stops (theoretical). |
| 85% | World-class | 90% availability, 95% performance, 99% quality (traced to Nakajima, Introduction to TPM, 1988). |
| 60% | Typical | Between the 55% and 65% averages found in published plant studies (sources: Ljungberg, 1998; Hedman et al., 2016); a clear signal of room to improve. |
| 40% | Low | Large, fast upside. |
One caution: benchmarks vary by industry and process. A continuous process line and a high-changeover job shop are not comparable. Compare yourself to your own trend first, then to the references above.
Hansen grades OEE on key processes in three bands: below 65% unacceptable, 65% to 75% passable only if the quarterly trend is improving, and 75% to 85% pretty good. His world-class targets, built on Nakajima's TPM figures, depend on the type of process (source: Hansen, Overall Equipment Effectiveness, 2001):
| Type of process | World-class OEE |
|---|---|
| Batch and discrete processes | 85% or more |
| Dedicated discrete, and non-dedicated continuous processes | 90% or more |
| Dedicated continuous processes | 95% or more |
The gap between your OEE and the target for your type of process is the first measure of what Hansen calls the hidden factory: output you already pay for and do not get.
The six big losses
OEE is most useful when you trace each lost point to a cause. The "six big losses" map onto the three factors:
| Factor | Loss | Examples |
|---|---|---|
| Availability | Breakdowns | Equipment failure, unplanned stops |
| Setup & adjustments | Changeovers, warm-up, tooling | |
| Performance | Idling & minor stops | Jams, misfeeds, sensor blocks: micro-stops |
| Reduced speed | Running below nameplate rate | |
| Quality | Process defects | Scrap and rework in steady state |
| Reduced yield | Defects during startup/changeover |
How to improve OEE
- Measure automatically. Manual logs miss the small losses. Start from machine data, not clipboards.
- Attack the biggest factor first. If availability is lowest, work on unplanned downtime; if performance is lowest, you have a micro-stop or speed problem; if quality is lowest, target scrap and rework.
- Find the true cause. "Performance loss" isn't actionable; "the labeller jams on wet cartons" is.
- Fix and verify. Detect → diagnose → fix → verify the same failure stops recurring.
For the full method, see the step-by-step guide to how to improve OEE. Starting from scratch? Read how to implement an OEE program.
Why your reported OEE is probably too high
Here's the trap. OEE only subtracts the downtime you record, and stops of a minute or two rarely make it onto a paper log. Their lost time vanishes from availability and shows up, if at all, as a vague performance gap. The result: a hand-kept OEE can look healthier than the line really runs.
Small stops add up faster than they look. In a case Hansen describes, output fell about 10% with no single big event behind it: the time to restart after each tiny stop had crept from 0.8 to 1.1 minutes, and only a chart of restart time per stop showed it. Record how often stops happen, not only how long they last. More on these micro-stops.
A quick test: if OEE calculated from good units (method 3 above) comes out lower than your reported OEE, the difference is loss your logs are not catching. Put a price on the whole gap with the hidden factory calculator.
OEE on your machines
OEE works the same on any machine. To find what is pulling a specific machine down, start from its guide, for example the conveyor, injection moulding, CNC machine and filling machine, or browse the troubleshooting directory. How the numbers are collected matters as much as the formula; see OEE data collection methods.
Fabrico reads OEE directly from the machine and shows video of a stop streamed on request next to its machine data, where cameras are fitted.
Fabrico calculates OEE from machine data instead of a shift sheet: stops, speed and counts live from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, with the causes tagged by operators at the line.
Frequently asked questions
What is OEE in simple terms?
The percentage of your planned production time that's truly productive: good parts, at full speed, with no stops.
Is OEE the same as utilization or TEEP?
No. OEE measures planned production time only. TEEP (Total Effective Equipment Performance) extends OEE to all calendar time, so it also reflects whether you're scheduling the asset enough.
Can OEE be over 100%?
No. If your math produces over 100%, an input is wrong, usually the ideal cycle time is set too slow.
How often should I measure OEE?
Continuously, per machine. Shift- or day-level averages hide the patterns; the value is in seeing when and where losses happen.
How do you calculate OEE from good units?
Multiply the good count by the ideal cycle time and divide by planned production time. 21,500 good pieces × 1 second ÷ 28,800 seconds = 74.7%. It gives the same result as availability × performance × quality and needs only three numbers.
How do you average OEE across lines or shifts?
Weight each OEE by its planned production time. A line at 60% over 400 minutes and one at 80% over 80 minutes give (400 × 60% + 80 × 80%) ÷ 480 = 63.3%, not the plain average of 70%.
Is a changeover downtime in OEE?
It is a planned stop inside planned production time, so it counts against availability. Only time planned ahead as not for production, such as no orders, holidays or a planned shutdown, is left out of OEE; that time shows up in TEEP.
What OEE should a continuous process reach?
R. C. Hansen (Overall Equipment Effectiveness, 2001), building on Nakajima's TPM targets, sets world-class at 95% or more for dedicated continuous processes, 90% or more for non-dedicated continuous and dedicated discrete processes, and 85% or more for batch and discrete processes.
Related guides: OEE for Injection Molding Machines: Worked Example · OEE for Continuous Processes: How to Measure It · Fully Productive Time: Definition and Waterfall · History of OEE: From Nakajima to Hansen and ISO · Unplanned Downtime Examples by Type and Machine · OEE values reported in published studies