Blog / Machine types
Machine types · Guide

OEE for CNC Machines: What to Count and Example

OL By OEE Lab Editorial |Updated 1 November 2026 |6 min read
Short answer

For a CNC machine, availability is run time over scheduled time (setup, waiting and tool breaks are stops). Performance compares the best proven cycle time with the actual one, so feed override below 100% is a speed loss. Quality is good parts over parts made.

Why CNC OEE needs its own rules

The OEE formula is the same for every machine: availability × performance × quality. What changes is what you count as a stop, what the ideal cycle time is and what is a good part.

On a CNC machine, three things cause trouble.

The spindle can turn while no good part is being made. Setup and first-article checks take a big share of the day. And an operator who runs two or three machines is often the reason one of them waits.

This page shows how to count each loss on a CNC machine, with a worked example. For software that reads CNC machines, see production monitoring software for CNC. For fault codes and repairs, see CNC machine troubleshooting.

The three factors for a CNC machine

FactorWhat to countOn a CNC machine
AvailabilityRun time ÷ scheduled timeScheduled time is the planned shift minus breaks. Setup, waiting and tool breaks are stops.
PerformanceIdeal cycle time × parts ÷ run timeThe ideal cycle is the best proven cycle for that part, at 100% feed and with good tools.
QualityGood parts ÷ total partsScrap, rework and the parts used to prove out the program count against it.

The three factors applied to a CNC machine.

OEE = availability × performance × quality
or, with three numbers only:
OEE = (good parts × ideal cycle time) ÷ scheduled time

The second form is the one in Overall Equipment Effectiveness (R. C. Hansen, 2001), who recommends it as the simplest check. It needs only scheduled time, good parts and the ideal cycle time. See how to calculate OEE for the full method.

Where each CNC loss goes

Event on the machineOEE lossNote
Setup, fixture change, program loadAvailability (stop)A planned changeover is still a loss. See SMED.
First-article inspection waitAvailability (stop)Time waiting for the quality check counts as a stop.
Waiting for material, program, tooling or craneAvailability (stop)Stops caused outside the machine.
Operator busy at another machineAvailability (stop)Common on multi-machine cells. Log it so it is visible.
Tool breakage or unplanned tool changeAvailability (stop)A planned tool change that is part of the cycle is not a stop.
Alarm, coolant, chip jamAvailability (stop)See CNC machine troubleshooting.
Feed override below 100%Performance (slow running)Running slower than the proven cycle.
Short pauses, single-block, door opened mid-cyclePerformance (minor stops)Many short stops add up. See micro-stops.
Scrapped part, rework, prove-out partsQualityCount in units, and time them at the ideal cycle.

Where common CNC events are counted. Decide your rules once and write them down.

Warm-up is a judgment call. If your plant plans a warm-up as part of the schedule, you can exclude it, but write that rule down and keep it the same for every machine. If you treat it as a stop, you will see it in the log. Do not move it from one bucket to another from month to month.

Setting the ideal cycle time

The ideal cycle is the best cycle time the part has run at with good quality, not the CAM estimate and not an average. Take the fastest cycle you have proven in production, with the tool set and feed you would run again.

If you do not know it, Hansen suggests the rate from the best 4 hours out of the last 400. Keep it fixed for the part.

When you improve the program and accredit a faster cycle, the faster cycle becomes the new ideal. Mark the change on your chart. See ideal cycle time.

Do not count tool-change time as a stop if it happens inside every cycle. It is part of the cycle. Count it as a stop only if it happens outside the cycle, for example a worn insert change that needs the operator.

Worked example: one machining center, one shift

Example: a machining center runs one part. The shift is 480 minutes with a 30-minute break, so scheduled time is 450 minutes. The ideal cycle time is 3.0 minutes a part.

ItemMinutes
Setup and first-article check60
Waiting for material20
Tool breakage15
Operator at another machine25
Total stops120
Run time (450 - 120)330

Example stop log. Hypothetical numbers.

In the 330 run minutes the machine made 100 parts. The ideal output for 330 minutes would be 330 ÷ 3 = 110 parts. Four parts were scrapped, so 96 were good.

FactorCalculationResult
Availability330 ÷ 45073.3%
Performance100 × 3.0 ÷ 33090.9%
Quality96 ÷ 10096.0%
OEE0.733 × 0.909 × 0.96064.0%
Check with 3 numbers96 × 3.0 ÷ 450 = 288 ÷ 45064.0%

Both methods agree. If they do not, one of the loss categories is missing.

  • OEE (good parts)64.0%
  • Setup and first article13.3%
  • Operator at another machine5.6%
  • Waiting for material4.4%
  • Tool breakage3.3%
  • Slow running6.7%
  • Scrap2.7%
Where the 450 scheduled minutes went. The shares add to 100%: 64.0 + 13.3 + 5.6 + 4.4 + 3.3 + 6.7 + 2.7 = 100.0.

Setup is the biggest single loss at 13.3%. A shorter setup is worth more here than a faster cutting program. That is the kind of ranking OEE is for.

Scheduled time (450 min)100%
After availability (120 min of stops)73.3%
After performance (110 parts possible, 100 made)66.7%
After quality (96 good of 100): OEE64.0%
The three factors applied one after the other: 100% × 0.733 × 0.909 × 0.960. Availability takes the biggest bite (26.7 points), performance 6.6, quality 2.7. Example numbers.

Now compare this with spindle time. If the spindle turned for 300 minutes, "utilization" is 300 ÷ 450 = 66.7%. It sounds close to the OEE, but it is not the same number and it does not tell you what to fix. Utilization ignores speed and quality. See utilization vs OEE.

Mixed parts and short runs

Job shops run many parts a day. Give each part number its own ideal cycle time, work out the theoretical run time of the good parts for each, and add them. Then divide by the total scheduled time. Do not average the OEE of each part. Weight by time.

OEE = Σ (good parts × ideal cycle time of each part) ÷ scheduled time

Example: 60 good parts of A at 3.0 min and 40 good parts of B at 5.0 min in 450 minutes. Theoretical time = 60 × 3 + 40 × 5 = 380 min. OEE = 380 ÷ 450 = 84.4%.

How to collect the data

You need four signals: the machine state (running, stopped, alarm), the cycle start and end, the part count and the stop cause. States and counts can often come from the controller or a PLC signal. On older machines, an external sensor on the stack light or a part counter can give the same facts. See PLC signals for OEE and retrofit sensors for OEE.

The cause is the part that needs a person. Ask the operator to tag it at the machine, with a short list of codes. For the options, see OEE data collection methods.

What is a good target?

Hansen sets world-class bars by process type, building on Nakajima's TPM figures: 85% or more for discrete and batch processes, 90% or more for dedicated discrete, and 95% or more for dedicated continuous. A job shop is a discrete or batch process, so 85% is the bar to look at.

Use it as a direction. Set your own first target from your own baseline and your own mix. See how to set OEE targets and is 85% OEE realistic.

See it live in Fabrico

CNC OEE needs the machine state, the cycle and the part count. Fabrico reads stops, speed and counts live from PLCs or from IoT sensors on older machines, and operators tag the cause at the machine, so setup, waiting and tool breaks show up as separate losses. See machine connectivity for OEE.

Book a Fabrico demo

What people ask next

Is spindle-on time the same as OEE?

No. Spindle time says the machine was turning. It does not say whether the part was good, whether the cycle ran at its best time, or what the machine waited for. OEE counts all three. See utilization vs OEE.

Does setup time count against OEE?

Yes. A planned changeover is still time the machine could have made parts. Hansen classes it as a planned stop that is part of normal operation, and it reduces availability. Shortening it is one of the best levers. See how to measure changeover time.

How do I treat a tool change?

If the tool change is inside the program and happens every cycle, it is part of the cycle time. If it is an unplanned change, like a broken tool, or a wear change that needs a person, it is a stop.

How do I handle an operator running several machines?

Log "waiting for operator" as its own stop reason. It shows how much time is lost to the cell design and not to the machine, and it points to different fixes, such as a better layout or an extra person at peak times.

How do I get cycle and part counts from an older CNC?

Many older machines still have a signal for the cycle or a stack light. A sensor on that signal and a part counter give you state, cycle time and count. See retrofit sensors for OEE.

Related guides: Tool Change Downtime: Measure It and Cut It · OEE & Downtime in Machinery & Fabrication · OEE for Job Shops: High Mix and Short Runs · OEE Calculation for a Machine Shop, by Example

← More OEE Lab guides · Open the OEE calculator →

OEE essentials: OEE calculator · What is OEE · Six big losses · Availability · TEEP · Cost of downtime · OEE benchmark · Free OEE tracker · Free OEE software · How to track downtime · How to implement OEE · Light MES vs full MES