A plant manager once told me his line was running at 95% efficiency. It sounded impressive until I asked what it was measuring.
Output over planned output, he said. Of the units the line was scheduled to make, it made 95% of them. Which is a reasonable thing to measure. It is also not OEE, and the gap between the two is where most improvement programmes quietly fail.
OEE looks like arithmetic and behaves like a diagnosis. Done properly it tells you where your capacity actually went. Done lazily it becomes a number that gets reported upwards and argued about downwards, and nobody learns anything.
The short version. OEE is availability times performance times quality. Each of the three points at a different family of loss, so the ratios are more useful than the total. A 60% result with one dominant problem beats a 70% result with no idea where to start.
What OEE actually measures
Three ratios, multiplied together. Take them one at a time.
Availability is the time equipment could actually run, divided by the time it was planned to be available. A line scheduled for 480 minutes that lost 60 to breakdowns and changeovers has availability of 420 divided by 480, or 87.5%.
Performance is the output you got, divided by the output you should have got while running. If the line ran 420 minutes at a rated speed of one unit per minute, it owed you 420 units. Producing 360 makes performance 85.7%.
Notice that performance only counts running time. This is why availability comes first: without it, you would be measuring speed against time the machine never had.
Quality is good units divided by total units. If 360 units came off and 342 were right first time, quality is 95%.
Multiply them: 0.875 times 0.857 times 0.95 gives 71.2%. That is the OEE.
The multiplication matters. A line cannot hide a bad area behind a good one, because a brilliant quality result cannot rescue a machine that spent its morning broken.
The six big losses, and which number each one hits
The reason OEE is worth the bother is that each ratio maps onto specific, fixable losses. That is the mental model which turns a number into action.
- Breakdowns and setup and adjustment reduce availability. These are the visible ones: the machine is stopped and somebody knows why.
- Idling and minor stops and reduced speed reduce performance. These are the quiet ones. The machine is moving, so nobody escalates it, but a jam every twelve minutes costs more than the breakdown did.
- Defects and rework and startup rejects reduce quality. Startup rejects are easy to forget because the run settles after twenty minutes, but twenty minutes of scrap every shift compounds.
If you only track the total, you cannot tell which of the six is eating your week. Track the three ratios and you can. That is the whole argument for OEE over a plain efficiency figure.
Why a 60% OEE is often perfectly healthy
Somewhere in the 1980s, world-class OEE got defined as 85%. It has been quoted at people ever since, usually by somebody who has not checked whether their process has the same shape as a car plant running three shifts on a bottleneck asset.
A single-shift operation with frequent product changes will not hit 85% availability, because changeover is real work the schedule demands. A plant with seasonal demand should not run flat out in the trough. A job shop making forty different parts a week has a different loss profile entirely.
The honest question is not are we at 85%. It is which loss is largest, and what would halving it be worth.
A line at 55% with one dominant changeover problem is in far better shape than a line at 70% with losses scattered evenly and no idea where to start. The first has a project. The second has a chart.
What OEE leaves out
OEE measures equipment, not flow. A machine can show a respectable OEE while producing exactly the wrong thing, in batches larger than the next step can consume, straight into a warehouse nobody asked for.
If your constraint sits downstream, optimising upstream OEE makes the problem worse by burying the constraint under inventory.
It also says nothing about the people running the asset, the quality experienced by the customer, or whether the output was needed at all. Treat it as a measure of equipment potential realised, not as a score for how good the process is. Our glossary has plain-English definitions for the surrounding terms if you want to line them up against each other.
How to start measuring it this week
Do not start with software. Start with a definition and a sheet of paper.
- Fix the planned time. Write down what the line is genuinely scheduled to run, and keep it constant for a month. If planned time moves whenever the number looks bad, you are measuring mood, not equipment.
- Agree the rated speed. Use the manufacturer figure or your best demonstrated output, and be honest about which. Then leave it alone.
- Log stops in three columns: start, end, reason. Ten seconds per event. This gets you availability immediately and gives you the Pareto that tells you where to look.
- Report the three ratios separately, weekly, on one page. Not the total. The ratios start arguments about reality, and those arguments are the point.
Two weeks of that gives you a baseline more trustworthy than a year of estimated efficiency. If you want the structure ready-made, the free tools section has calculators, and the template library has the sheets to log against.
Watch out. The most common failure is OEE used as a target rather than a diagnostic. The moment a bonus depends on the number, performance gets reclassified as a slower rated speed, minor stops get logged as planned maintenance, and quality escapes get reworked off the record.
The measure survives. The improvement does not.
Keeping the number honest
Keep OEE where it belongs: a weekly conversation about which loss to attack next, held by the people closest to the asset. The number is only ever as good as the honesty in the log behind it.
If you are building measurement discipline for the first time, the Yellow Belt material walks through baseline and data collection properly, and the certification guide explains which belt level covers this ground. Everything on this topic sits alongside our other practical improvement guides.
