Field Guide · Manufacturing · Interactive

OEE, Explained by Carving Up a Shift

Overall Equipment Effectiveness looks like one tidy percentage. It’s really a story about where a machine’s minutes went. So take one machine, one 480-minute shift, and carve the losses out yourself — every slider below moves the same bar, and OEE recomputes live.

1 Before OEE starts

Planned production time

OEE doesn’t measure the whole shift. First you set aside the downtime you chose — breaks, meetings, periods with no scheduled demand. Those aren’t failures, so they leave the denominator entirely. What remains is planned production time: the minutes you genuinely intended to run. Every OEE loss is measured against that.

One shift · grey hatch = out of scope for OEE430 min planned
0120240360480 min
Fully productive 259 minDefects 15 minStartup rejects 8 minReduced speed 38 minMinor stops 35 minChangeovers 45 minBreakdowns 30 minNo demand 20 minBreaks 30 min
Excluded from OEE — but TEEP remembers it. More below.
Planned production time
430 min
480 − 50 min planned downtime
Defining planned time honestly is the first place OEE is won or lost. Pad it and every number downstream flatters you.
2 The first cut

Availability — was it running at all?

Now the losses begin. Breakdowns and changeovers stop the machine during time you planned to run — these are availability losses, the first two of the six big losses. Drag them up and watch red minutes carve out of the bar. On many lines the changeover slice is the fattest, most controllable target — which is exactly what SMED, the changeover-reduction method, exists to shrink.

Red = availability lossesAvailability 82.6%
0120240360480 min
Fully productive 259 minDefects 15 minStartup rejects 8 minReduced speed 38 minMinor stops 35 minChangeovers 45 minBreakdowns 30 minNo demand 20 minBreaks 30 min
Setup + adjustment, per shift. The SMED lever.
Availability
82.6%
355 min run ÷ 430 min planned
3 The subtle cut

Performance — was it running at full speed?

The machine is up — but is it producing at the rate it was designed for? Two losses hide here: minor stops (ten-second jams and misfeeds nobody logs) and reduced speed — an actual cycle time longer than the ideal. Say this machine should finish a unit every 60 seconds; drag the actual cycle slower and a striped amber erosion spreads through the remaining bar. Cycle time doing the quiet damage here is the same clock from the cycle/takt/lead time guide.

Amber = performance losses (stripes = speed)Performance 79.5%
0120240360480 min
Fully productive 259 minDefects 15 minStartup rejects 8 minReduced speed 38 minMinor stops 35 minChangeovers 45 minBreakdowns 30 minNo demand 20 minBreaks 30 min
Small, unlogged, and usually the biggest surprise of a first study.
Performance
79.5%
282 min at ideal rate ÷ 355 min run · 282 units made
4 The final slice

Quality — was it making good parts?

Last cut: some of what you made wasn’t right the first time. Defects during steady running and startup rejects while the process settles after each changeover both count — the machine spent real minutes making them, and those minutes are gone whether the part is scrapped or reworked. What survives all three cuts is the blue segment: fully productive time.

Violet = quality losses · blue = fully productiveQuality 91.9%
0120240360480 min
Fully productive 259 minDefects 15 minStartup rejects 8 minReduced speed 38 minMinor stops 35 minChangeovers 45 minBreakdowns 30 minNo demand 20 minBreaks 30 min
Warm-up scrap after starts and changeovers.
Quality
91.9%
259 good ÷ 282 made
Now multiply

The whole shift, one waterfall

Here’s everything you just carved, side by side: the shift bar and the multiplication it implies. OEE = Availability × Performance × Quality — each factor is a fraction of what the previous one left behind, so the waterfall can only step down. Try the presets: the ~60% “typical” and ~85% “world-class” figures are widely-cited rules of thumb, not measurements — useful as orientation, nothing more.

Your shiftOEE 60.3%
0120240360480 min
Fully productive 259 minDefects 15 minStartup rejects 8 minReduced speed 38 minMinor stops 35 minChangeovers 45 minBreakdowns 30 minNo demand 20 minBreaks 30 min
60.3%= 82.6% A × 79.5% P × 91.9% Q
259 min of 430 min planned minutes did what the customer pays for.
The multiplication, drawn
Planned production time430 min× Availability 82.6%−75 min× Performance 79.5%−73 min× Quality 91.9%−23 min= OEE 60.3% (259 min of 430 min)
Multiplication is brutal: three factors at a healthy-sounding 90% each compound to 0.9 × 0.9 × 0.9 ≈ 73% OEE. That’s why a “pretty good everywhere” line still loses a quarter of its planned time — and why averaging the three factors (which says 90%) misleads you by seventeen points.
The map behind the colors

The six big losses

Every colored segment you dragged belongs to one of the six big losses — the classic TPM catalogue of where equipment time dies. The point of OEE is not the percentage; it’s that each factor points at two specific, fixable losses.

Availability losses
1 · BreakdownsEquipment failures — the machine is down when you meant to run.
2 · ChangeoversSetup and adjustment between products. The classic SMED target.
Performance losses
3 · Minor stopsJams, misfeeds, sensor trips — under a few minutes, rarely logged.
4 · Reduced speedRunning below the ideal cycle time, often after a problem was “fixed.”
Quality losses
5 · DefectsUnits that need scrap or rework — the time they took is lost.
6 · Startup rejectsWarm-up scrap after changeovers and starts, until the process settles.
Read before you benchmark

The classic OEE traps

Comparing OEE across different machines
A packaging line running one SKU all week and a machining cell doing six changeovers a day live in different worlds — their OEEs aren’t comparable, and ranking them punishes whoever does the harder job. OEE is a trend line for one machine against itself. Compare this week to last week, not Line 2 to Line 5.
Chasing OEE on non-bottlenecks
Push OEE on a machine that isn’t the constraint and you don’t get more output — you get inventory piling up in front of the machine that is. The bottleneck sets the pace of the whole line, the same way the tallest bar caps a Yamazumi chart in line balancing. Measure OEE where the constraint is; elsewhere, high OEE can literally mean overproduction.
Gaming the ideal cycle time
Performance is measured against the ideal cycle time — so quietly “re-rating” the machine from 60s to 70s buys ten points of OEE without improving anything. Slide the actual cycle in section 3 and imagine moving the ideal instead: same shift, better number, zero more parts. Use the design rate the equipment can genuinely sustain, and write it down where nobody can nudge it.
Trusting the machine’s own counter
Control systems log the big breakdowns and round away the ten-second jams — and those minor stops are routinely the largest single loss a first honest study finds. The fix is unglamorous: someone stands at the machine with a clipboard for a few shifts and writes down every stop, however small.
The stricter cousin

OEE vs TEEP

Remember the grey minutes you excluded in section 1? TEEP (Total Effective Equipment Performance) puts them back. Where OEE asks “of the time we planned to run, how much was fully productive?”, TEEP asks the same against all the time there is. OEE is the operator’s number — it forgives decisions the operator didn’t make. TEEP is the capacity-planner’s number — it shows how much machine you actually own before buying another one.

OEE (÷ planned)60.3%
TEEP (÷ whole shift)54.0%
Same shift, same sliders — the gap between the two is exactly your planned downtime. Add “no demand” minutes in section 1 and watch only the grey bar fall.
Your first measurement

How to start measuring — honestly

Don’t start with software. Start with paper: pick one machine — ideally the constraint — and have someone watch it for a few shifts, logging every stop with a time and a reason, counting every unit and every reject, and timing the real cycle with a stopwatch. Then compute the three factors exactly the way this page does. The first honest number is usually lower than anyone expected, and that’s the point — a flattering OEE improves nothing, while an honest one hands you a ranked list of lost minutes to go fix.

When you’re ready to run the numbers on your own line, the free OEE calculator does this arithmetic for real shift data.

Open the OEE calculator →
Good to know

Frequently asked

What is OEE and how is it calculated?
OEE — Overall Equipment Effectiveness — measures how much of a machine’s planned production time was fully productive: running, at full speed, making good parts. It’s calculated as Availability × Performance × Quality. Availability = run time ÷ planned production time (losses: breakdowns and changeovers). Performance = time at ideal rate ÷ run time (losses: minor stops and reduced speed). Quality = good units ÷ total units made (losses: defects and startup rejects). Planned downtime — breaks, no demand — is excluded before you start.
What is a good OEE score?
The commonly-cited rules of thumb put a typical manufacturing plant around 60% and “world-class” around 85% — but treat those as orientation, not targets, because OEE isn’t comparable across different machines or product mixes. A line doing six changeovers a day will honestly score below one running a single SKU all week. The useful benchmark is your own machine’s trend: an honest 55% that climbs beats a flattering 80% that hides losses.
What are the six big losses in OEE?
Two per factor. Availability losses: breakdowns (equipment failure) and changeovers (setup and adjustment). Performance losses: idling and minor stops (brief jams and misfeeds, usually unlogged) and reduced speed (running below the ideal cycle time). Quality losses: process defects during steady running and startup rejects while the process settles after starts and changeovers. The value of OEE is that each disappointing factor points at exactly two places to go look.
What is the difference between OEE and TEEP?
OEE measures fully productive time against planned production time — breaks and no-demand periods are excluded, so it forgives scheduling decisions. TEEP (Total Effective Equipment Performance) measures the same fully productive time against all available time, planned or not. OEE is the operator’s number and drives loss-hunting on the floor; TEEP is the capacity planner’s number and shows how much latent capacity you own before buying another machine. TEEP is always less than or equal to OEE.
Why is OEE multiplied instead of averaged?
Because the losses compound in sequence: performance losses can only eat time the machine was actually running, and quality losses can only eat time it was running at speed. Three factors at 90% each multiply to about 73% OEE, not 90% — a quarter of planned time gone even though every factor looks healthy. Averaging would hide that compounding and overstate the line by double digits, which is exactly the kind of flattery OEE exists to prevent.
How do you start measuring OEE?
Start with paper, one machine — ideally the bottleneck. For a few shifts, have someone log every stop with a time and reason, count total and rejected units, and time the actual cycle with a stopwatch. Define planned production time honestly (subtract breaks and no-demand, nothing else), use the true design cycle time as the ideal, and compute the three factors. Automating the collection comes later; the discipline of an honest first number comes first.
MS
Matthew Savas

Founder of Kaizumi, an AI-powered Lean training platform. More about Matthew →

Updated July 2026 · The machine and all slider figures are illustrative, created for teaching. The ~60% typical and ~85% world-class figures are widely-cited rules of thumb, not measured benchmarks. Definitions follow standard OEE/TPM usage: OEE = Availability × Performance × Quality against planned production time; TEEP measures the same against all available time.