Robot Workcell KPIs: Benchmarks for OEE, Availability, Pick Rate and Changeover
OEE and its three factors, pick rate and payload headroom, path efficiency and changeover: how to measure each robot cell KPI, what moves it, and which figures have a cited source.
Use these benchmarks as reference points for spotting a gap worth investigating in a robot cell, not as targets. Your own trend per cell says more than any single published figure.
Every figure shows its basis. A superscript number links to a public source in the Sources list, and the sentence gives that source's scope, such as a cross-industry study or a manufacturer test cycle. A figure marked no cited source is kept exactly as originally published because it has not been matched to a public source that states it, so treat it as illustrative.
Ranges are as originally published unless a source is cited. The cited OEE figures come from cross-industry studies and the total productive maintenance (TPM) world-class factors, not from robot cells, and support the floor of each world-class band rather than its ceiling. Where the original used a word other than typical or strong, that word is kept in the cell. Not stated means no figure was given for that column.
Method: How the figures in this guide are sourced
A figure with a superscript number is stated by the cited source, listed under Sources with its publisher and access date. Where a source measures something broader than robot cells, such as cross-industry OEE, facilities maintenance or a manufacturer's test cycle, the sentence says so. A figure marked no cited source is kept exactly as originally published because it has not been matched to a public source that states it. It is illustrative, not measured industry data. Worked examples are arithmetic illustrations, not benchmarks.
Robot cell OEE
Robot cell OEE (overall equipment effectiveness) is the headline KPI, and the benchmark spread is wide. Typical brownfield cells run 55 to 70 percent OEE1, 2, 3, a well-run cell lands 75 to 85 percent no cited source, and world-class high-volume automation reaches 85 to 92 percent, where 85 percent is the classic world-class floor4.
The cited figures are cross-industry, not robot-specific. Across 23 Swedish companies, average OEE was 65 percent and the median 70 percent1; a second Chalmers study found machine OEE averaged 51.5 percent3; and Vorne, an OEE software maker, says most manufacturers run closer to 60 percent2. The 85 percent world-class mark traces to Nakajima's total productive maintenance (TPM) work4.
Because OEE multiplies availability, performance and quality, a cell at 90 percent on each factor still yields only 73 percent OEE2. The Lean Enterprise Institute's own example, 90 percent availability, 95 percent performance and 99 percent quality, gives 84.6 percent8.
A low reading is rarely bad everywhere. It is usually one weak factor dragging two healthy ones down, so always decompose it.
How to measure
Multiply availability, performance and quality, each as a fraction.
Report the three factors beside the composite every time.
What moves it
The single weakest factor, found by decomposing the number.
Trade-offs between factors, since pushing arm speed can raise scrap and faults.
Availability is where most cells bleed, and the targets are stricter than on manual lines. World-class robot availability sits at 95 percent or better, matching the equipment availability benchmark of above 95 percent in the Department of Energy's facilities operations and maintenance guide5.
That DOE benchmark, attributed to NASA (2000), counts hours available to run at capacity; it is not an OEE factor5. The TPM world-class availability factor is above 90 percent4. Typical cells run 85 to 92 percent, and anything under 85 percent signals chronic faults, air leaks or gripper misfeeds no cited source.
A cell losing 6 percent to unplanned stops is losing about 27 minutes per 7.5 hour shift (6 percent of 450 minutes). With 15 minute repairs, pushing mean time between failures (MTBF) from 4 hours to 12 raises availability by about 3.8 points; with repairs of 12 to 20 minutes, the gain is roughly 3 to 5 points.
How to measure
Divide run time by planned production time.
Separate planned stops from unplanned ones.
Track MTBF and repair time for the cell.
What moves it
Mean time between failures, the main lever.
Air leaks and gripper misfeeds, the usual chronic faults.
Repair time, which sets how much each failure costs.
Performance efficiency compares actual output with the cell's own ideal cycle, and small losses compound. Benchmark performance is 92 to 98 percent for mature cells; the TPM world-class performance factor of above 95 percent sits inside that band4.
Under 90 percent usually means micro-stops, reduced speed after a fault, or a taught path slower than design no cited source. Minor stops are easy to miss: Vorne notes that most companies do not accurately track idling and minor stops9.
If ideal cycle is 3.4 s but actual averages 3.7 s, performance is 92 percent (3.4 / 3.7). On a cell rated near 1,000 per hour, that 0.3 s gap costs about 85 parts per hour: 3,600 / 3.4 is about 1,059 parts an hour against 973 at 3.7 s.
How to measure
Divide ideal cycle time by actual average cycle time.
Log micro-stops and reduced-speed running separately.
What moves it
Dwell times, audited station by station.
Re-taught arcs and approach paths, rather than a higher rated arm speed.
Quality rate for robot cells should be high, since repeatability is the whole point. Target first-pass yield of 99.5 percent or better; the TPM world-class quality rate is higher still, above 99.9 percent4.
Typical cells run 98.5 to 99.5 percent, and below 98 percent the placement, vision or upstream part presentation is out of control no cited source. The quality lever is rarely the robot itself; it is fixture wear, incoming part variation or gripper slip.
A cell dropping from 99.5 to 98 percent yield loses 1.5 percent of output, near 14 parts per hour at 963 good per hour. That quietly erases the gains from a hard-won availability improvement.
How to measure
Divide first-pass good parts by parts produced.
Log rejects by cause: placement, vision or part presentation.
Pick rate and its utilization tell you whether the robot is fast enough or just busy. High-speed delta-robot picking runs 80 to 150 picks per minute; ABB's IRB 360 FlexPicker data sheet (2013) says its 8 kg version handles up to 100 cycles per minute6.
Articulated pick-and-place typically runs 20 to 45 per minute no cited source. For reference, ABB's IRB 1600 specification lists an approximate packer cycle of 0.95 s at 7 kg and 1.01 s at 10 kg, about 60 cycles per minute on that manufacturer test cycle10; application cycles with longer moves run slower.
Raw speed matters less than payload and reach headroom. Keep payload utilization at 60 to 80 percent of rated payload; above 80 percent, arms derate acceleration and cycle time climbs 5 to 15 percent no cited source.
Keep reach margin at 10 to 15 percent of rated reach so the arm avoids near-singularity slowdowns no cited source. Cells failing these two checks often show a pick rate 10 to 20 percent below the robot's spec sheet no cited source.
How to measure
Count picks per minute over a full shift, not a burst.
Divide tool plus part mass by rated payload.
Compare the farthest pick point with rated reach.
What moves it
Payload headroom, which protects acceleration.
Reach margin, which keeps the arm away from singularities.
Robot selection matched to the pick rate the line needs.
Path efficiency is an underused KPI with fast payback. World-class taught paths hold 85 to 95 percent efficiency, meaning actual travel is within 5 to 15 percent of straight-line; typical un-optimized cells sit at 70 to 82 percent no cited source.
No single target fits every cell. A machine-tending cell that waits on a spindle scores lower than a conveyor picker by design, so compare a cell with its own history first.
Improving from 75 to 90 percent path efficiency cuts travel by about 16.7 percent (1 minus 75 / 90). On a motion-bound cycle where motion is half to three quarters of the time, that can cut 8 to 12 percent off cycle time with zero capital.
This is usually the cheapest KPI to move, since it costs teach-pendant time rather than new hardware.
How to measure
Divide straight-line distance by actual travel for each move.
Rank moves by wasted distance.
What moves it
Re-teaching the two or three moves that waste the most distance.
Verifying each re-taught path in measured output, not just on the pendant.
Changeover performance decides how automation handles mix. Benchmark robot changeover is under 10 minutes for SMED-style cells, the single-digit goal of single-minute exchange of die7; typical mixed cells run 15 to 40 minutes no cited source.
A cell doing 6 changeovers a shift at 25 minutes forfeits 150 minutes, a third of a 450-minute shift. Cutting changeover from 25 to 8 minutes on that cell recovers roughly 100 minutes per shift (6 x (25 minus 8) = 102), which often beats any speed tuning the robot itself can deliver.
How to measure
Time each changeover from the last good part to the first good part.
Multiply changeovers per shift by their average duration.
Set targets as a linked scorecard, not isolated numbers, because the KPIs trade against each other. Pushing arm speed to lift performance can raise scrap and cut availability through more faults.
For example, a cell balanced at 95 percent availability, 95 percent performance and 99.5 percent quality reaches about 90 percent OEE (0.95 x 0.95 x 0.995 = 89.8 percent).
Write each KPI's definition down and keep it fixed.
Review the scorecard weekly.
Flag any factor more than 3 points below target no cited source.
Attack the single largest loss first.
Recompute after each change, so improvements are verified in output, not just claimed on a dashboard.
World-Class OEE: Set Targets To Drive Improvement. Vorne Industries (OEE.com). Sections 'What Is World-Class OEE?', 'TPM and World-Class OEE', 'Some Perspective on World-Class OEE'. Accessed September 30, 2026.
Anatomization of the overall equipment effectiveness (OEE) for various machines in a tool and die shop. Journal of Intelligent Manufacturing and Special Equipment, Vol. 3 No. 1, pp. 97-105 (Emerald, CC BY 4.0), DOI 10.1108/JIMSE-01-2022-0004. published 2022. p. 101 (PDF p. 5): sentence "Seiichi Nakajima included the above four “world-class” numbers." and Table 1 "The world-class OEE factor"; p. 103 (PDF p. 7): "worldwide threshold of >85% set by the Japan Institute of Plant Maintenance (JIPM)". Accessed September 30, 2026.
IRB 360 FlexPicker data sheet (ROB0082EN_F). ABB Robotics. published May 2013. p. 1 product description; p. 2 'Performance' table, typical cycle times for 1130 mm variants. Accessed September 30, 2026.
OEE Six Big Losses. Vorne Industries (OEE.com). Sections 'Equipment Failure' (threshold rule of thumb) and 'Idling and Minor Stops'. Accessed September 30, 2026.