OEE in manufacturing: what it tells you — and what it hides

OEE in manufacturing (Overall Equipment Effectiveness) measures how much of your planned production time is truly productive. It multiplies three factors — Availability, Performance, and Quality — into one percentage. A score of 100% means only good parts, at full speed, with zero stops. Most plants measuring honestly for the first time land well below 85% — and that is the point: the number shows you where the loss is.

What is OEE in manufacturing?

OEE in manufacturing measures how much of a machine's scheduled production time actually turns into good parts, at full speed, with no stops — expressed as one percentage.

OEE = Availability x Performance x Quality

The OEE formula is straightforward: multiply Availability, Performance, and Quality — each expressed as a percentage — to get the share of planned time that produced good output.

Each factor captures a different kind of loss:

• Availability — what percentage of planned time the machine was actually running. Breakdowns and unplanned stops eat into this.
• Performance — when the machine was running, was it running at full speed? Speed losses and minor stoppages bring this below 100%.
• Quality — of everything produced while it ran, how many pieces were good? Defects, rework, and startup scrap live here.

Multiply the three and you get the fraction of planned time that actually produced good output. A machine available 90% of the time, running at 85% of its ideal speed, and producing 97% good parts: 0.90 x 0.85 x 0.97 = 74.2% OEE.

The framework was defined and popularised by Seiichi Nakajima of the Japan Institute of Plant Maintenance (JIPM) as part of the TPM framework in the 1980s. See our full guide to Total Productive Maintenance — OEE is the headline metric that TPM is built around.

How do you calculate OEE?

The OEE calculation needs five inputs per machine per time period:

1. Planned production time — the time the machine was scheduled to run (shift time minus planned breaks and scheduled stops such as changeovers)
2. Downtime — unplanned stops: breakdowns, material waits, operator waits
3. Ideal cycle time — the rated output per minute at maximum speed
4. Actual output — pieces actually produced during the running time
5. Good output — pieces that passed quality, first time, without rework

Here is a complete worked example — one machine, one eight-hour shift:

The maths forces a useful decision. A 73.1% OEE is not a damning verdict in isolation. But look at where the loss sits: 60 minutes of the 480-minute shift went missing, half of it to a breakdown that stopped the line entirely. Availability is the lever — and availability is the factor maintenance can move.

What's a good OEE score?

Nakajima's classic benchmark for "world class" in discrete manufacturing is 85%. It is quoted on almost every OEE resource online. The honest framing is more useful:

• 40–60% is typical for plants measuring OEE for the first time and recording honestly
• 65–75% is a reasonable working target after six months of running the numbers and attacking losses
• 85% is achievable in a high-volume, dedicated-line environment with a mature TPM program — it is not the starting line

The 85% figure was set for discrete manufacturing (machined parts, assembly). Process industries (chemicals, food, textiles) and job-shops running 30 different part numbers per week will see different norms. Applying Nakajima's benchmark to a changeover-heavy job-shop and concluding the plant is failing is a category error.

More useful than chasing a number: use OEE as a before-and-after tool. Baseline it on your highest-priority line, attack the biggest loss (usually availability), measure again in 90 days. The direction matters far more than the absolute value.

The most consistent pattern I have seen in plants with frequent changeovers is this: setup time gets logged as downtime — pulling Availability down — and by end-of-shift the floor is compensating. Operators run the machine at a higher feed rate, or run it in air with no material loaded, letting the cycle counter tick up and inflating the Performance figure. The headline OEE looks reasonable. The losses are real — they have just moved categories. This is why the ideal cycle time and the actual conditions on the floor need to be reviewed together, not reported separately.

OEE vs TEEP: what's the difference?

OEE and TEEP (Total Effective Equipment Performance) use the same three factors — Availability, Performance, Quality — against a different denominator.

A line running two shifts, five days a week, can post a strong OEE and a weak TEEP at the same time — the gap is the third shift and the weekend, hours OEE never counts against you because they were never scheduled. TEEP is the number for "should we add a shift" or "do we need a second machine" — a capacity question, not a maintenance one. OEE stays the right tool for "how well are we running right now."

The three losses — and which one maintenance owns

TPM theory names six categories of loss, mapped onto the three OEE factors:

Availability is the factor maintenance can move the most. Breakdowns and unplanned stops are what a CMMS is built to reduce — through preventive maintenance that keeps components from failing, faster diagnosis when they do, and a work history per machine that surfaces repeat failures before they become line stops.

Performance and Quality improvements usually require engineering changes, process redesign, or operator discipline that runs well outside maintenance's scope. This is why availability improvements often get credited to maintenance in factory reports — not because maintenance owns OEE, but because the availability factor is where maintenance actually has leverage.

A worked example of this leverage: the 87.5% availability in the table above was dragged down by a 45-minute breakdown. Fix the root cause — say, a PM that catches the bearing condition that caused the stop — and availability moves from 87.5% to something like 93.75%. Run the numbers: 0.9375 x 0.857 x 0.975 = 78.3%. A five-point OEE improvement from one maintenance action on one machine.

For the metrics that sit inside the availability factor — MTTR, MTBF, and downtime per asset — see our maintenance KPIs guide.

How to improve OEE in manufacturing

Three moves, in the order they actually pay off:

1. Fix Availability first. Breakdowns are the biggest, fastest-moving lever, and the one maintenance owns directly — a PM that catches a failing component before it stops the line, work-order history that surfaces which asset keeps repeating the same failure, faster diagnosis when something does go down. The worked example above showed one fixed breakdown moving OEE five points on its own.

2. Then attack planned stops. Changeovers and setups count against Availability too, and they're usually the second-biggest recoverable chunk after breakdowns. Rapid-changeover methods like SMED often move the number faster than anything aimed at Performance or Quality.

3. Only then touch Performance and Quality — and check the ideal cycle time first. Both sit mostly outside maintenance's control. And any Performance fix is wasted effort if the ideal cycle time baseline is wrong — confirm it against the machine's actual rated speed before investing there.

Work the list in that order. Plants that jump to a Performance or Quality initiative while Availability is still leaking tend to spend effort polishing a number that one bad shift of downtime will erase.

What OEE hides

OEE is useful precisely because it is simple. That same simplicity hides things a plant manager should see.

1. Whether you should be producing at all. A 95% OEE making a product nobody ordered is still waste. OEE measures productive use of planned time — it says nothing about whether demand justified running the line in the first place. Schedule adherence and utilisation (actual vs calendar time) are different questions.

2. Which loss to attack first. A 73% score tells you something is wrong; it does not tell you whether to fix the breakdown that dragged availability down or the scrap that pulled quality. You need the three components separately, and then you need the loss breakdown within each one. The headline OEE is the conclusion, not the analysis.

3. How the number was calculated. Two plants reporting "75% OEE" may be measuring very differently — different definitions of planned time, different cycle-time baselines, different ways of counting or excluding changeovers. OEE numbers from different factories are almost never directly comparable, which is a polite way of saying the benchmarking conversations at industry events are largely noise.

4. That the ideal cycle time is honest. If the ideal cycle time used to calculate Performance is set below the machine's actual rated speed — because someone wanted a comfortable target, or because nobody remembered to update it after a process change — the Performance component is overstated and the overall OEE looks better than it is. Gaming this is extremely common, and it usually happens quietly.

My advice to any plant starting with OEE for the first time: do not try to read all three components every single day from the beginning — you will get inconsistent numbers and draw the wrong conclusions. Start with Availability. The available time is the denominator that expands your Performance calculation; the more you keep the machine running, the more output you produce, and a higher production volume naturally lifts your Quality yield as well. Most plants' single biggest OEE leak is a machine that simply is not running when it should be. Make the machine available first, and the other two factors tend to follow.

How to start measuring OEE without a dashboard

You do not need software to start. Three numbers per machine per shift:

1. Downtime log — a sheet per shift recording every unplanned stop: start time, end time, reason. The discipline is that every stop gets logged at the time it happens, not reconstructed at shift end. Shift-end reconstruction invariably smooths the data into a usable-looking lie.

2. Actual output count — usually already tracked for production reporting. If not, a tally counter works.

3. Reject count — pieces scrapped or sent for rework. Also usually tracked.

From those three numbers and the shift plan, you can compute all three OEE factors and the headline number. Start on one machine — your highest-priority line by breakdown hours or by the cost of a stop. After four weeks of consistent data, the biggest loss becomes visible. The decision about where to invest (a PM schedule, a changeover redesign, a quality fix) stops being a management debate and starts being arithmetic.

One honest note on ideal cycle time: the first time a plant runs this calculation, the Performance component often comes out improbably high — above 95%, sometimes above 100%. That usually means the ideal cycle time is set too low (too conservative a target speed). Reset it to the machine's actual manufacturer-rated output capacity, not whatever the line is accustomed to running at. The Performance number will drop, and the OEE will look worse. That is not failure — that is what the number is supposed to do.

Where a CMMS enters the picture: MachDatum does not compute your OEE. OEE needs Performance data (actual vs ideal speed) and Quality data (reject counts), which come from the machine or the production system — not from maintenance records. What MachDatum does compute, automatically from work-order data, is the availability component: MTTR per machine, MTBF per machine, downtime hours by asset and by team, updated with every closed work order. If your OEE is being dragged down by the availability factor — and for most plants running mainly reactive maintenance, it is — that is where to start, and that is what a CMMS gives you visibility on.

Where MachDatum fits: reducing availability losses is exactly what a CMMS is built for. PM auto-dispatch keeps scheduled services from slipping. Every breakdown becomes a work order with a root-cause field and a due date. MTTR and MTBF per machine update automatically from the work-order history — no spreadsheet. We're onboarding our first group of manufacturing teams right now — see the analytics view at machdatum.com/cmms.