OEE calculation: the formula, a worked example and the traps to avoid

If you ask ten plants what their OEE is, you will get ten answers and at least three different definitions of the math behind them. Overall equipment effectiveness is the most widely used productivity metric in manufacturing and one of the easiest to misuse. A clean OEE calculation is the foundation of every credible continuous improvement program. A sloppy one will steer your investment decisions for years in the wrong direction.

This pillar guide walks through the OEE calculation the way a plant manager or continuous improvement lead needs it in 2026. We cover the OEE formula, each of the three factors (availability, performance and quality), the six big losses, a worked example with real numbers, the benchmarks that matter, the traps that quietly inflate the score and the modern way to capture the underlying data.

If you read one OEE calculation guide this year, this is the one we want it to be.

What OEE actually is

Overall equipment effectiveness is a single number, between zero and 100 percent, that describes how well a piece of equipment is converting planned production time into perfect production. World-class OEE sits around 85 percent. The average plant runs much closer to 60 percent. The 25 percentage points between those two numbers is the prize that continuous improvement, lean manufacturing and total productive maintenance (TPM) all chase.

OEE is the product of three factors. Availability captures how much of the planned production time the machine was actually running. Performance captures how fast it ran when it was running. Quality captures how many of the parts it produced were good the first time, with no rework.

The OEE formula is:

OEE = Availability × Performance × Quality

That is the entire calculation. Every other number you see on an OEE dashboard, including the six big losses, performance loss, speed losses, small stops, quality loss, asset utilization and equipment efficiency, is derived from those three factors and the underlying counts and timings.

The reason this single number matters is that it forces the conversation back to the four signals every plant has to track: how much time the machine is available, how fast it runs, how good the parts are and how that combination compares to perfect production. Anything that does not lift one of those four signals is not improving OEE.

The three factors, in plain English

Availability

Availability is the share of planned production time the machine was actually running.

Availability = Run Time ÷ Planned Production Time

Planned production time is the total scheduled time for the machine, minus planned downtime (planned stops like lunch breaks, scheduled changeovers built into the plan and planned maintenance windows). Run time is the planned production time minus all unplanned downtime, including breakdowns, equipment failures, unplanned stops and the small stops that operators often forget to log.

Availability hides more than it shows. Most plants underreport unplanned stops, especially small ones. A vision sensor at the end of the line that watches the actual flow of parts often turns up 15 to 30 percent more stop time than the operators have been recording, because small stops of 30 seconds or a minute almost never make it into the manual log.

Performance

Performance is the speed loss. It compares how fast the machine actually ran during its run time to how fast it could have run at the ideal cycle time.

Performance = (Ideal Cycle Time × Total Count) ÷ Run Time

The ideal cycle time is the fastest cycle the machine can achieve while still producing good parts. The total count is every part the machine produced during the run time, good or bad. Performance loss shows up as slow cycles and as small stops that are too short to be classified as downtime but slow the overall throughput.

Performance is where the six big losses concept earns its keep. Equipment that runs slower than design accumulates loss minute by minute that nobody notices until the end of the shift, when the throughput is short and nobody can explain why.

Quality

Quality is the share of total count that came out as good parts the first time. Rework counts as bad. Scrap counts as bad. Only what passes inspection on the first run counts as good.

Quality = Good Count ÷ Total Count

Quality is the factor most plants get wrong, because it depends on having a real measurement of good parts at the line, not a sample inspection three hours later in the lab. Camera-based inspection has changed this. A modern vision system that runs on a single iPhone with on-device AI can count good and bad parts in real time, which finally gives the OEE calculation a clean quality signal.

The OEE formula and the math

The full OEE formula is the product of the three factors:

OEE = (Run Time ÷ Planned Production Time) × ((Ideal Cycle Time × Total Count) ÷ Run Time) × (Good Count ÷ Total Count)

The math simplifies neatly. Two of the terms cancel, leaving:

OEE = (Ideal Cycle Time × Good Count) ÷ Planned Production Time

This second form of the OEE formula is what most modern OEE software calculates under the hood. It is a much cleaner way to think about the number, because it removes the bookkeeping around run time and total count and asks the only question that matters in the end: how many good parts did you make compared to how many you could have made if the machine ran perfectly the entire planned production time.

Either form gives the same answer. The first form is more useful when you want to see where the loss came from. The second form is more useful when you want a single trustworthy OEE score.

The six big losses

Total productive maintenance gives us the standard taxonomy for what kills OEE, called the six big losses. Every credible OEE calculation maps loss back to these six categories.

The first two losses hit availability. Equipment failures, also called breakdowns, are the long unplanned stops that everyone notices. Setup and adjustment, also called changeovers, is the time spent switching products or recalibrating after a change.

The next two losses hit performance. Idling and minor stops, also called small stops, are the short interruptions, usually under five minutes, that operators rarely log. Reduced speed, also called slow cycles, is the difference between the ideal cycle time and the actual cycle time at the observed throughput.

The last two losses hit quality. Process defects, also called quality loss, are the bad parts produced during steady-state running. Reduced yield, also called startup rejects, are the bad parts produced at startup before the machine reaches steady state.

Once you have the OEE score, the six big losses tell you where the gap to perfect production actually sits. A plant at 60 percent OEE with 35 percent of the loss in small stops and 25 percent in changeovers needs a very different improvement plan than a plant at 60 percent OEE with 30 percent of the loss in quality.

A worked example

Take a single shift on a single line with a packaging machine.

• Shift length: 480 minutes (8 hours)
• Planned downtime (breaks, planned changeovers): 60 minutes
• Planned production time: 480 - 60 = 420 minutes
• Unplanned downtime (breakdowns, small stops): 47 minutes
• Run time: 420 - 47 = 373 minutes
• Ideal cycle time: 1.5 seconds per part
• Total count produced during run time: 12,200 parts
• Good count (parts that passed quality inspection on first run): 11,650 parts

Now compute each factor.

Availability = 373 ÷ 420 = 88.8 percent

Performance = (1.5 × 12,200) ÷ (373 × 60) = 18,300 ÷ 22,380 = 81.8 percent

Quality = 11,650 ÷ 12,200 = 95.5 percent

OEE = 0.888 × 0.818 × 0.955 = 0.694, or 69.4 percent

That single OEE score tells you the line is running at 69 percent of its perfect production potential during the shift. The breakdown tells you where to look. Quality is in good shape at 95.5 percent. Availability is decent at 88.8 percent. Performance is the weak link at 81.8 percent, which means the line is either running slower than ideal cycle time or losing throughput to small stops the operators have not logged.

The improvement plan writes itself. Spend a week studying performance loss on this line. Map the small stops and the slow cycles. Most of the time, three or four root causes account for the bulk of the loss. Close those out and the OEE score moves before the quarter ends.

This is what a clean OEE calculation actually does. It does not just give you a number. It tells you exactly where to spend the next 20 hours of engineering time.

What "good" OEE looks like

Industry benchmarks for OEE land in roughly four bands. World-class OEE for discrete manufacturing sits around 85 percent, which means 90 percent availability, 95 percent performance and 99 percent quality multiplied together. Very few plants live here, and the ones that do treat OEE as a board-level metric.

A good plant runs in the 60 to 80 percent band. This is where most well-run discrete manufacturing operations sit and the realistic short-term target for a plant starting a continuous improvement program from a lower base. A typical plant runs in the 40 to 60 percent band, which is the manufacturing productivity average across discrete manufacturing globally. A struggling plant runs below 40 percent, producing less than half of what the equipment is capable of, with loss usually split between poor downtime tracking, untracked small stops and quality issues that never make it back to the line.

These industry standards are useful for orientation, not for blame. The most useful benchmarking a plant can do is internal: line A versus line B on the same product, shift A versus shift B on the same line, or this month versus last month on the same shift. Internal benchmarking changes behavior. External benchmarking mostly produces slide decks.

The traps that inflate OEE

Five traps come up over and over in OEE deployments.

The first is overstating planned downtime. The math hides anything you call "planned" in the denominator, so a plant that treats long, optional changeovers as planned will see an artificially high OEE. If a changeover could have been shorter with better setup, it is not fully planned.

The second is missing small stops. Stops of 30 seconds to a few minutes are the largest source of loss in most plants and never make it into the manual log. The only reliable capture is automated stop detection from a PLC, a vision sensor or a camera at the line.

The third is the wrong ideal cycle time. Some plants use the original manufacturer spec, some the fastest observed cycle, some a current best-practice number. The choice changes the OEE score by several percentage points. Pick one definition, document it and stick with it across lines.

The fourth is uncategorized stops. If 30 percent of your downtime ends up tagged "other," the OEE calculation is technically correct but operationally useless. Root cause analysis cannot run on a category that means nothing.

The fifth is sampling-based quality. If the quality factor comes from a sample taken every two hours in the lab, the calculation is missing the rework and the in-process defects that happen between samples. The trap closes once you have continuous quality measurement at the line.

Where modern OEE software changes the calculation

The classical OEE calculation assumes you have PLC integration, an HMI on every line and an MES that ties it all together. That assumption was reasonable in 2010 and unaffordable for most plants in 2026.

The change of the last three years is that a credible OEE calculation can now run on a single iPhone, a lamp and a mount, sitting at the end of the line. On-device AI counts good parts and bad parts. The camera detects stop and start events from changes in motion. Total count, good count and stop time come out of the same data stream, which means availability, performance and quality can all be computed without wiring into the PLC.

This matters because most plants never get to a clean OEE calculation, not for lack of will but because the per-line cost of traditional production monitoring software is too high to justify across 30 or 40 lines. A camera-first approach gets per-line hardware under €1,000, which means the OEE calculation can start on a single line in a week and expand from there. The cleanest deployment we have seen recently mounted a refurbished iPhone over the end of a packaging line and surfaced OEE on a dashboard the supervisor checks every hour. Three months later the score was up 11 percentage points, almost entirely from small stops the camera caught and the manual log had been missing for two years.

How to set up your OEE calculation in week one

If you are starting from scratch, here is the shortest practical path.

Define planned production time with care. Subtract only the planned downtime that is truly unavoidable, and write down the definition so nobody can quietly inflate it later. Pick one line and instrument it for total count, good count and stop time. The cheapest path in 2026 is a camera at the end of the line, but a clean PLC integration works equally well if you have it. Document the ideal cycle time and review it once a quarter.

Set up reason codes for the six big losses on this line and configure the system to prompt the operator the moment a stop crosses a five-minute threshold. Calculate OEE at the shift level for the first month. Once the data is trustworthy, move to hourly or live updates. Run the first improvement project against the largest of the six big losses and verify the OEE score moves before you start the second.

That sequence gets you from no OEE calculation to a live OEE score the line supervisor trusts inside four weeks, on a single line, with under €1,000 of hardware. From there you can scale across the plant without anyone needing to approve a six-figure capital project.

OEE is the start, not the end

OEE is the most useful single number a plant can track for equipment efficiency and overall equipment effectiveness. The OEE calculation becomes reliable once you have clean planned production time, real stop tracking and an accurate good count.

The score is not the point. The OEE score exists to direct your improvement work toward the largest loss, which is what the six big losses framework is for. A plant that runs an honest OEE calculation, watches the six big losses every week and closes one root cause per month will hit world-class OEE within three years. The math works that way.

Get OEE running on your line

Enao Vision runs the OEE calculation on a single iPhone, a lamp and a cable. The hardware to get started costs less than €1,000 per line, and most teams have good count, bad count and downtime tracking running inside a week. Start a free trial and we will help you get the first line live.

Join the community

We run a free Slack community for shopfloor builders, continuous improvement leads and operations people who want to compare OEE numbers in the open. Members swap reason-code playbooks, six-big-losses breakdowns and deployment lessons every week. Join the community.