OEE calculation: a worked example for a robot cell

OEE is one number that tells you how much of your available machine time actually produced good parts. It is simple arithmetic. Almost every plant that quotes an OEE figure is calculating it wrongly.

This guide covers availability, performance and quality, runs a full worked example on a robot cell, maps the six big losses to each factor, and explains why 85% is a benchmark rather than a target you should chase.

The three numbers inside OEE

Overall Equipment Effectiveness multiplies three ratios. Each one answers a different question, and each is measured only over the planned production time.

• Availability = run time ÷ planned production time. Did the machine run when it was supposed to?
• Performance = (ideal cycle time × total count) ÷ run time. When it ran, did it run at its designed speed?
• Quality = good count ÷ total count. Of what it made, how much was sellable?

Planned production time excludes breaks, meals, planned maintenance and shifts with no demand. If you include planned stops, you are measuring the calendar, not the machine, and the number becomes meaningless in both directions.

A worked example on a robot cell

Take a single-robot machining cell on a two-shift pattern. One shift is 480 minutes, with a 30-minute planned break.

Inputs, one shift

Shift length                            480 min
Planned break                            30 min
Planned production time                 450 min

Downtime: changeover 20, breakdown 25    45 min
Run time                                405 min

Ideal cycle time                         42 s
Total parts produced                      520
Rejected parts                             12
Good parts                                508

Step 1 - Availability
  405 / 450 = 0.900                    90.0%

Step 2 - Performance
  ideal time = 520 parts * 42 s
             = 21,840 s = 364.0 min
  364.0 / 405 = 0.8988                 89.9%

Step 3 - Quality
  508 / 520 = 0.9769                   97.7%

OEE = 0.900 * 0.8988 * 0.9769
    = 0.7904                           79.0%

Cross-check, the short way:
  good parts * ideal cycle / planned time
  = 508 * 42 / 27,000 s
  = 21,336 / 27,000 = 0.790            79.0%

If the two methods disagree, one of the
counts is wrong.

The cross-check is the most useful line in that block. It catches double counting, wrong shift length and counts that include parts made during the break.

The six big losses

Every point of lost OEE belongs to one of six categories. Naming the loss correctly is what turns OEE from a report into an improvement plan.

What 85% actually means

85% is a long-standing benchmark for discrete manufacturing, from the TPM literature. It is a useful comparison point, not a universal target. A cell making one part with a stable process can sit above 90%. A job shop with six changeovers a day may be excellent at 65%.

Two rules keep the number honest. First, never compare OEE across machines with different ideal cycle times unless the ideal times are validated. Second, never use OEE as a league table between operators. The moment it becomes a score, the counts stop being true.

OEE for a cell, not just the robot

A robot’s own utilisation is a poor metric, because the robot is rarely the constraint. The cell is. If the robot is waiting for a part, its utilisation looks terrible while the machining centre that should be feeding it is the real problem.

So define the unit of measurement before you define the target. Measure the constraint: the machine or station that limits output. Robot waiting time is then a symptom with a name — starved or blocked — rather than a separate KPI to be optimised.

• Starved. The robot has no part because the upstream machine is slow or stopped.
• Blocked. The robot cannot unload because the downstream conveyor or buffer is full.
• Robot-caused stop. The robot itself failed, which is genuine availability loss for the cell.

Measuring without an IIoT stack

You do not need a platform to start. You need a counter, a clock and a reason code. Plenty of Indian plants run a 90% accurate manual OEE for a month and learn more than they would from a year of dashboards.

1. Agree planned production time in writing, shift by shift.
2. Record total count from the machine’s own counter, not from dispatch.
3. Record reject count at inspection, not at rework.
4. Log every stop longer than two minutes with a reason code.
5. Validate the ideal cycle time from a video, not from the supplier brochure.
6. Review weekly, on the floor, with the people who run the cell.

When the manual number stabilises, automate the data capture. The IIoT monitoring guide covers the options and costs, and the maintenance checklist keeps the availability factor from drifting.

Where the first five points usually come from

Improve in the order of the multipliers, not in the order of enthusiasm. Availability gains are usually the cheapest because they need discipline, not capital.

Value of a 5-point OEE gain (indicative)

Current OEE                                 79%
Improved OEE                              84%
Planned time per shift                   450 min
Ideal cycle                                 42 s

Good parts now
  450 * 60 * 0.79 / 42 = 507.8  -> 508

Good parts at 84%
  450 * 60 * 0.84 / 42 = 540.0

Extra parts per shift                 32
Shifts per day                          2
Working days per year                 300
Extra parts per year               19,200
Contribution per part              INR 150

Annual contribution            INR 28.8 lakh

A 5-point gain is worth more than most
capital projects, and it usually comes
from changeover discipline and minor
stops - not from a faster robot.

Common mistakes

• Including planned stops. Breaks and planned maintenance belong outside the calculation.
• Using the supplier’s cycle time. The brochure figure is usually the best case at 100% speed with no tool load.
• Counting reworked parts as good. Rework consumes capacity, so it belongs in quality loss.
• Ignoring minor stops. They hide inside performance and make it look like the machine is simply slow.
• Measuring only the bottleneck when it moves. Recalculate the constraint after every improvement.
• Reporting a single plant-wide number. An average across machines with different constraints describes nothing.

Correct cycle-time arithmetic is covered in the industrial cycle time guide, and the link from cycle time to money is in the ROI guide.

What a robot changes, and what it does not

A robot usually improves quality consistency and reduces the human variation in cycle time, which lifts performance and quality together. It does not automatically improve availability. In fact, a poorly integrated cell adds new failure modes: gripper wear, programming errors, safety trips and handshake faults with the machine tool.

That is why the integration and commissioning scope matters. A cell that is programmed for fault recovery, with clear alarms and a trained operator, protects availability. A cell that needs the integrator for every alarm destroys it. The integration cost guide explains what that scope should include.

Sources and further reading

• OEE.com — definitions, formulas and the six big losses.
• SME — lean manufacturing and total productive maintenance resources.
• A3 Association for Advancing Automation — robot cell performance and integration guidance.
• International Federation of Robotics — automation adoption and productivity data.