OEE: formula, worked example and how to measure it without paper
OEE (overall equipment effectiveness) is a metric that tells you what share of a machine’s planned production time turned into good parts made at ideal speed. You calculate it by multiplying availability, performance and quality. A score of 70% means 30% of the planned time went on stops, slower running or scrap, and the three parts show you which.
In short
- OEE measures one machine or line over planned production time. Unplanned nights and weekends don’t lower the score; the whole calendar is measured by a separate metric, TEEP.
- The formula is availability × performance × quality. A faster route is good parts × ideal cycle time ÷ planned production time, and both give the same answer.
- Most errors come from the inputs, not the formula: an inflated cycle time, changeovers taken out of planned time, and downtime written up from memory the next day.
- The 85% “world class” figure has its roots in the Japanese car industry of the 1970s. Steady improvement on the same machine is a better target than someone else’s number.
- OEE is not for rating people or comparing different machines. Once people are judged by it, the data soon stops being true.
Contents
- What is OEE and where does it come from?
- The OEE formula: availability × performance × quality
- A worked OEE example, step by step
- Is 85% “world class”? What should OEE be?
- Six mistakes that make OEE lie
- How to measure OEE automatically, without paper
- When does OEE mislead?
- Where to start measuring OEE in your plant
- Frequently asked questions
What is OEE and where does it come from?
OEE measures how well a machine or line uses the time in which it was meant to produce. It doesn’t measure people, it doesn’t measure the whole factory, and it doesn’t tell you whether you’re making the right product. It tells you how much one piece of equipment loses and why. An OEE of 100% means that for the whole planned time the machine made only good parts and never slowed down.
The term first appeared in Seiichi Nakajima’s 1982 book “TPM Tenkai” as part of TPM (total productive maintenance), an approach in which operators share responsibility for the condition of their machines. That explains the focus on losses: OEE exists to show where a machine gives time away, so the team can remove those losses one by one.
OEE is measured over planned production time. If the plant runs one shift, nights and weekends don’t lower OEE. Use of the whole calendar is measured by another metric, TEEP, covered below.
The OEE formula: availability × performance × quality
OEE is the product of three components, each answering a different question. Multiplied together they give one number, but when you look for causes you always look at the components separately.
| Component | How to calculate it | Question |
|---|---|---|
| Availability | run time ÷ planned production time | How much of the planned time did the machine actually run? |
| Performance | (parts made × ideal cycle time) ÷ run time | How fast did it run compared with ideal speed? |
| Quality | good parts ÷ all parts | What share of parts was good first time? |
The same formulas appear in the Wikipedia entry on OEE. There is also a shortcut that is handy for checking your arithmetic: OEE = good parts × ideal cycle time ÷ planned production time. Run time cancels out and you are left with a simple question: how many minutes of “good production” did you get out of the minutes you planned?
A few terms to define once and stick to:
- Planned production time is shift time minus breaks in which the machine is deliberately not meant to run (a meal break, no orders planned in advance).
- Run time is planned time minus all stops: breakdowns, changeovers, missing material, waiting for inspection.
- Ideal cycle time is the shortest time per part the machine can achieve on that product: from the nameplate, from the manufacturer, or the best result measured in practice. A standard with a buffer built in won’t do.
OEE vs TEEP: a different time base
OEE measures losses during planned production time. TEEP measures the same thing against the whole calendar: 24 hours a day, 7 days a week. According to Wikipedia, TEEP is OEE multiplied by the loading factor, meaning the share of the calendar you plan for production at all.
| Metric | Time base | Question it answers |
|---|---|---|
| OEE | planned production time | How much do I lose when the machine was meant to run? |
| TEEP | the whole calendar | How much capacity does this machine have left before I buy another? |
When you compare results from different sources, always check which time base was used. The same machine can score very differently depending on whether the base was the plan or the calendar.
TEEP helps with investment decisions. If OEE is 70% and the machine runs one shift, check whether a second shift or better OEE would be cheaper before you buy a second machine.
A worked OEE example, step by step
The example below is made up to show the method. One machine, one 8-hour shift (480 minutes), including a 30-minute planned break. Ideal cycle time is 1 minute per part. During the shift there were two stops: a 35-minute changeover and a 25-minute feeder breakdown. The machine made 330 parts, 15 of them scrap.
- Planned production time: 480 − 30 = 450 minutes.
- Run time: 450 − 35 − 25 = 390 minutes.
- Availability: 390 ÷ 450 = 86.7%.
- Performance: 330 parts × 1 minute ÷ 390 minutes = 84.6%.
- Quality: 315 good ÷ 330 total = 95.5%.
- OEE: 86.7% × 84.6% × 95.5% = 70.0%.
- Shortcut check: 315 good parts × 1 minute ÷ 450 minutes = 70.0%. The results match.
Percentages suit a report, but minutes are better for decisions. The same example then looks like this:
| Where the minutes went | Minutes | Share of 450 |
|---|---|---|
| Stops | 60 | 13.3% |
| Slower running | 60 | 13.3% |
| Scrap (15 parts) | 15 | 3.3% |
| Good production | 315 | 70.0% |
Stops are the changeover and the breakdown. Slower running is the gap between 390 minutes of run time and the 330 minutes the parts made are “worth”. Scrap is 15 parts at 1 minute each. Now you can see something the 70% alone doesn’t tell you. Slower running costs as much as the stops, yet nobody reports it, because the machine is “running” the whole time. Behind those 60 minutes there are usually short stoppages of a few seconds each and running at reduced speed. These losses don’t appear in any downtime report, but they show up in the performance component.
Which loss lands where?
Every loss on a machine lowers exactly one OEE component. This map helps you decide what to collect and who should own each component.
| Loss | Lowers | How to see it |
|---|---|---|
| Breakdowns | availability | a stop logged with a cause |
| Changeovers and adjustments | availability | time from the last good part to the first good part on the new product |
| Short stops (seconds, minutes) | performance | part counter from the machine, rarely from paper |
| Running at reduced speed | performance | part count compared with ideal cycle time |
| Scrap and rework | quality | scrap logged when it happens, not at the end of the shift |
| Start-up losses | quality | a separate scrap cause, “start-up” |
Short stops matter most for performance. An operator won’t report every 20-second jam, so this component is only reliable once the part count comes from the machine rather than by hand.
Is 85% “world class”? What should OEE be?
There is no single good OEE value. You often hear that 85% is world class, but even Wikipedia notes that treating 85% as a target is in many cases incorrect.
The figure has a specific origin. According to oee.com, it comes from Japan in the 1970s and from the car industry, meaning long runs of the same product. A plant that makes short runs and changes over several times a shift has lower availability by definition. It can be very well run and still sit at 55%.
Multiplication also pulls the score down: three components at 90% give about 73%, not 90%. A more useful target is steady improvement on the same machine, measured the same way every week, rather than comparing yourself with someone else’s number.
Six mistakes that make OEE lie
The formula is simple, so the mistakes nearly always sit in the inputs and in what someone decided “not to count”. These change the result the most.
- Ideal cycle time taken from a padded standard. If the standard is 1.2 minutes and the machine can do 1 minute, performance comes out higher and speed losses vanish from the report.
- Changeovers taken out of planned time. Availability rises on paper and the plant loses its reason to shorten changeovers, even though they are often the biggest loss.
- Downtime written up from memory. A sheet filled in at the end of the shift has rounded times and the cause “other”. After a month you don’t know what to fix.
- Scrap counted only at the end, or rework counted as good. A reworked part is a part made twice. Quality counts only parts that were good first time.
- Averaging OEE across machines. A simple average across a line where one machine is the bottleneck blurs the picture. Wikipedia points out that OEE is not suited to comparing machines or adding up to plant level.
- Using OEE to rate staff. When a crew is judged on the score, downtime causes start to “improve” and the score rises with no change on the shop floor. OEE measures the machine, not people.
How to measure OEE automatically, without paper
The best set-up splits the work: the machine reports what it knows by itself, and a person adds only what the machine can’t know. The machine knows run time and part count. A person knows why it stopped and whether a part is good.
In practice it looks like this:
- Run signal and part counter from the machine. From the controller, a sensor at the outfeed or a simple pulse counter. Availability and performance are then calculated without anyone’s input.
- Downtime cause from a tablet at the machine. When the machine stops for longer than a set threshold (say 2 minutes), the tablet asks for a cause from a short list. Start and end times are recorded automatically.
- Scrap reported straight away. A “scrap” button with a cause at the workstation, or data from quality inspection.
- A shop-floor board and an end-of-shift report. The supervisor sees a stop while it’s happening, not the next day.
- A weekly review. The five biggest losses in minutes and what was done about each.
Reefa, my cleaning company, isn’t a factory and I don’t calculate OEE there. But the same principle runs in the bot our crews use to log shifts. A worker presses start and finish, and the time and location are recorded automatically. Lateness is flagged by the system, with alerts at 20, 40 and 60 minutes. A shift someone forgot to close is closed after 14 hours using the median length of that worker’s recent shifts at that site. The bot also carries a warning for OEE, about data coverage. In June 2026 roughly half of planned visits were logged through the bot. A metric built on half the data looks just as trustworthy as one built on all of it, and it isn’t. So before you trust your OEE, check what share of shifts and machines actually reach the system.
If you already collect downtime and wonder which system should organise it, I cover the differences between an MES, an ERP module and your own production board in what is an MES system.
When does OEE mislead?
OEE is good for finding losses on a specific machine. It works poorly when the question is about something else.
- The machine isn’t the bottleneck. Raising OEE on a machine that waits for the one before it won’t add a single part at the end of the line.
- You’re building stock nobody will buy. High OEE on an unwanted product means a well-used machine and badly used money.
- Manual production and assembly. When people, not the machine, set the pace, measure workstation output and time per order instead.
- Small batches with frequent changes. OEE will be low by definition here. Measure it, but set the goal as shorter changeovers, not the overall score.
- Data from half the shifts. A metric based on incomplete data looks precise and leads to bad decisions.
Where to start measuring OEE in your plant
Start with one machine that is the bottleneck and 30 days of logging stops with a cause. The first result isn’t meant to be a nice percentage. It should be a list of the five biggest losses in minutes.
Before you calculate the first score, agree three things in writing: what counts as planned production time, what the ideal cycle time is for each product, and a short list of downtime causes (six at most). If nobody in the plant has written down how work flows at the workstation, start with process mapping in a company. To decide whether to build your own board or buy a ready-made program, read custom system or off-the-shelf tool.
To work out what retyping paper sheets into a spreadsheet costs you today, use the automation savings calculator. If you need a production board that collects data automatically from machines and tablets, scope and prices are on the business systems page: a build with me starts from PLN 60,000 excl. VAT. And if you’d rather talk about your data first, book a 30-minute call and bring even one week of sheets from one machine.