How to Calculate CNC Machine Efficiency

• 1
  Key takeaways
• 2
  What efficiency means on a CNC machine
• 3
  Collect the raw data before you calculate anything
• 4
  Calculate availability, performance and quality
• 5
  Which of the three losses to attack first
• 6
  When the OEE number misleads you
• 7
  Step by step: run the calculation this week
• 8
  Efficiency methods compared

Efficiency on a CNC machine is not one number. It is the ratio between what a machine could produce in a given window of time and what it actually produced, counted in good parts. A mill that runs 20 hours a day but scrapped one in ten parts is less efficient than a mill that runs 14 hours and ships everything. That is why almost every machine shop and every customer audit falls back on OEE.

OEE stands for Overall Equipment Effectiveness. It combines three losses into a single percentage: time lost to stops, speed lost to slow cycles, and output lost to defects. When an engineer asks how to calculate CNC machine efficiency, OEE is the answer they are usually looking for, because it names the loss instead of hiding it inside one vague percentage.

You can also track simpler numbers. Utilization is run time divided by available time, and it ignores speed and quality. Yield is good parts divided by total parts, and it ignores time. Both are useful on a daily board, but neither tells you whether the machine earned money. Use them as inputs, not as the headline.

One warning before you start. Efficiency figures are only comparable when the part number, the material and the fixture are the same. Comparing a 6061 aluminum bracket to a 17-4PH stainless housing tells you nothing about the machines. Compare like with like, or compare each machine against its own history.

Write down four numbers for the shift or the week you want to measure. First, planned production time: total shift minutes minus planned breaks, scheduled maintenance and any deliberate idle time with no work available. Second, actual run time from the machine log or the CNC monitoring feed. Third, total parts produced. Fourth, good parts after final inspection.

Get these from the control, not from memory. Most modern controllers timestamp program start, program end and alarm events, so the run time is already there. If the machine has no logging, a simple paper sheet per shift works, but the operator has to record stops as they happen. Reconstructed logs are always optimistic.

Decide up front what counts as planned downtime. A tool change inside a program is a stop. A changeover between two jobs is planned downtime only if the schedule included it. This single decision moves OEE by 5–10 points, so write the rule down and apply it to every machine the same way.

Set the measurement window to at least one full week of normal production. A single good day after a setup is not data. If a machine runs three shifts, capture all three, because the night shift often has a different stop pattern and a different scrap rate.

Availability is run time divided by planned production time. If the machine was planned for 450 minutes and alarms, chip clearing and tool changes consumed 50 minutes, availability is 400 / 450 = 88.9%. This is the easiest of the three to inflate, because unplanned stops are easy to forget, so pull the alarm log rather than asking the operator.

Performance is the total parts produced multiplied by the ideal cycle time, divided by run time. If the CAM program says 1.4 minutes per part and 250 parts came off in 378 minutes of run time, performance is (250 × 1.4) / 378 = 92.6%. Take the ideal cycle time from the verified program at the feeds and speeds you actually run, not from a supplier catalog or a theoretical maximum.

Quality is good parts divided by total parts. If 250 parts were made and 8 were scrapped at final inspection, quality is 242 / 250 = 96.8%. Count rework as a defect, because the second pass consumes machine time that the schedule never budgeted for.

Multiply the three: 0.889 × 0.926 × 0.968 = 79.7%. That number is your OEE for the period. Anything above 85% is world class, 60–70% is typical for general job-shop work, and below 50% usually points to a scheduling problem rather than a machine problem.

Fix the smallest number first, not the largest loss in minutes. If quality is 92% and performance is 96%, the defects are costing you more per point than the speed loss, and they also carry inspection and rework time that never appears in the OEE formula.

For availability, look at the stop list in order of total minutes. Changeovers, tool changes and chip evacuation usually top it. Moving a job to a 5-axis machine that machines five faces in one setup removes most of the re-clamping time, and that shows up directly as availability. Reducing setup time is the fastest lever in most shops.

For performance, check whether the program is running at the posted feeds and speeds. Operators often dial back feed override to protect a thin wall or a long tool, and the cycle time quietly grows. If the override sits at 70% all day, the program is wrong, not the operator.

For quality, group defects by cause before you change anything. Tool wear, chatter on deep pockets and burrs on cross-holes have different fixes. On a 10,000-part run, a 1% scrap reduction is worth more than a 1% speed increase, because scrap also consumes the time of the parts that were made correctly.

OEE punishes low-volume, high-mix work. A shop running one-off prototypes will never reach 85%, because setup and programming dominate the clock. Judge prototype work by quote-to-ship time and first-part accuracy instead, not by OEE.

The formula also treats all parts as equal. One good part on a 4,000 mm gantry job and one good part on a Ø20 mm fitting both count as 1 in the quality term, even though one represents twenty times the machine time. Weight the parts by cycle time when you compare across very different jobs.

A high OEE is not automatically good. If the schedule is full of parts the shop should have declined, a 90% OEE just means you are efficiently making the wrong things. Read OEE alongside on-time delivery and first-pass yield before you draw conclusions.

Finally, do not compare your OEE with a published number from another industry. A die-casting cell, a grinding cell and a 5-axis milling cell have different loss structures. Compare your machines against your own baseline, month over month.