OEE benchmark by industry 2026: real numbers from 412 plants

OEE has a folklore problem. Every consultant deck still quotes the same 85 percent World Class number from a 1990s textbook and the same 60 percent industry average from a paper nobody can find. The numbers became wisdom without anyone updating the underlying data.

This piece is an attempt to fix that with current numbers. Over the past nine months we collected OEE data from 412 manufacturing plants across 14 industries, primarily in Europe and North America, supplemented by anonymised submissions from operators in the Enao community. The dataset is not a representative sample of global manufacturing. It is a useful working benchmark, and the medians shift the picture meaningfully from the consulting folklore.

The headline: median OEE in 2026 is 64 percent. The top decile is at 81 percent. The bottom decile is at 42 percent. The 85 percent World Class number is reached by 4 percent of plants in the dataset. The 60 percent industry-average legend is closer to the median than to the average, but the spread between industries is wide enough that any single benchmark hides more than it reveals.

Where the data came from

412 plants submitted data between September 2025 and May 2026. The methodology was deliberately conservative: only plants with continuous PLC-based capture (no manually-logged OEE) were included, and only data covering at least 12 consecutive weeks per plant. The split across industries:

• Automotive (parts and assembly): 71 plants
• Food and beverage: 58 plants
• Consumer packaged goods (non-food): 47 plants
• Pharmaceutical (solid and liquid dose): 39 plants
• Electronics and semiconductors: 34 plants
• Metals (casting, stamping, machining): 31 plants
• Plastics and polymers: 28 plants
• Building materials (ceramics, glass, cement): 26 plants
• Paper and packaging: 22 plants
• Textiles and apparel: 18 plants
• Chemical (specialty and fine): 16 plants
• Pet food and animal feed: 9 plants
• Aerospace components: 8 plants
• Medical devices: 5 plants

The medians and ranges that follow are computed per-industry and pooled where the sample size in an industry is below 10 plants.

The headline numbers

Median OEE by industry, with 25th and 75th percentile ranges.

Automotive parts and assembly: median 71 percent, range 58 to 79 percent. The highest-median industry in the dataset, driven by mature lean culture and decades of investment in capture. Top-decile plants reach 84 percent.

Food and beverage: median 62 percent, range 49 to 73 percent. The wide spread reflects the gap between high-volume beverage canning and low-volume specialty food. Top-decile reaches 80 percent.

Consumer packaged goods (non-food): median 67 percent, range 55 to 76 percent. Less spread than food because the production patterns are more standardised.

Pharmaceutical (solid and liquid dose): median 58 percent, range 47 to 71 percent. The validation overhead and changeover frequency hold the median down. Top-decile reaches 79 percent.

Electronics and semiconductors: median 78 percent, range 68 to 85 percent. The highest-median group, driven by capital intensity that forces availability-first plant design. Top-decile reaches 89 percent (the only industry where a meaningful number of plants exceed the World Class 85 percent threshold).

Metals (casting, stamping, machining): median 63 percent, range 51 to 74 percent. The setup-heavy nature of small-batch metal work pulls the lower quartile down.

Plastics and polymers: median 68 percent, range 56 to 77 percent. Continuous extrusion lines pull the median up; injection moulding pulls it down.

Building materials: median 71 percent, range 60 to 79 percent. Continuous kilns and presses sustain high availability when the process is stable.

Paper and packaging: median 73 percent, range 64 to 81 percent. Continuous high-speed lines, low changeover share.

Textiles and apparel: median 54 percent, range 42 to 66 percent. The lowest-median industry in the dataset; high variety, high manual content, weak capture infrastructure.

Chemical (specialty and fine): median 67 percent, range 55 to 78 percent. Pooled with other low-sample industries for the range.

Pet food, aerospace components, medical devices: pooled median 62 percent, range 50 to 73 percent.

The cross-industry median of 64 percent is the right working number for someone benchmarking a new plant. The textbook 60 percent is close but on the low side.

How the OEE calculation actually works

A working benchmark only helps if the OEE calculation behind it is consistent. The OEE formula multiplies three terms: availability times performance times quality. The shape of each term matters more than the textbook definition.

Availability is running time divided by planned production time. Planned production time excludes planned stops (scheduled maintenance, planned downtime, the lunch break). It includes everything else, including the changeover and the cleaning. Setup and adjustment time at the start of a run is in. Breakdowns are in. Minor stops are in.

Performance is the product's ideal cycle time multiplied by the total count, divided by the running time. Performance captures speed losses (reduced speed during a run) and minor stops that did not get coded as availability losses. It is the term that the median plant most often gets wrong, because the ideal cycle time on the spec sheet rarely matches the actual best-observed cycle time on the line. A plant running at the spec rate is typically running at 95 to 100 percent performance. A plant running at the best-observed rate is often at 88 to 93 percent.

Quality is good count divided by total count. It captures defects and the rework that follows them. The trap is to count rework as good once it has been reprocessed. The right method counts rework as a loss against quality and against the operating cost of the line.

Multiply the three and the OEE score lands. A line at 90 percent availability, 92 percent performance, and 95 percent quality scores 78.7 percent OEE. The same line, reported as 95 percent availability (because the cleaning got moved to planned downtime), 98 percent performance (because the spec rate was used as the ideal cycle time), and 96 percent quality (because the rework was counted as good) scores 89.4 percent OEE. Same line, ten points apart. This is why the benchmark is fragile if the calculation discipline is not pinned down at the same time.

A few related metrics worth naming. Overall equipment effectiveness is the proper full name of OEE and is sometimes written out in audits and regulatory filings (especially in pharmaceutical manufacturing). TEEP (Total Effective Equipment Performance) is OEE multiplied by the share of calendar time that the line is planned to run. A line at 75 percent OEE running two shifts a day, five days a week has a TEEP of about 27 percent. Manufacturing productivity is the broader frame that puts OEE in context with labor productivity and yield.

The six big losses

The OEE conversation gets cleaner once everyone names the six losses the same way. The framework comes from TPM (Total Productive Maintenance) and is the same in lean manufacturing.

The two availability losses: breakdowns (unplanned equipment failures) and setup and adjustment (changeover plus the warm-up time before the line is in spec).

The two performance losses: minor stops (the under-five-minute pauses that PLC logs under-report) and reduced speed (the line is running but below the ideal cycle time, often quietly).

The two quality losses: startup defects (the first batch after a changeover that did not meet spec) and production defects (everything later, including the parts that go to rework).

A median plant attacking OEE should know which of the six losses is the biggest by hours. The Pareto chart by big-loss category is the single most useful weekly report the operations team can produce. Discrete manufacturing typically has all six in roughly even shares. Continuous-process industries are dominated by breakdowns and reduced speed. Pharmaceutical manufacturing is dominated by setup and adjustment because of the cleaning between products.

Maintenance posture and the OEE link

The maintenance team owns the breakdowns share of the OEE loss. The posture they run determines the trajectory of that share over time.

Reactive maintenance fixes the line when it breaks. Preventive maintenance schedules service on a calendar. Predictive maintenance uses sensor signals to call the repair before the failure happens. Total productive maintenance is the cultural overlay that puts the operator at the centre of basic maintenance work. A modern CMMS (computerised maintenance management system) is the system of record that makes any of these postures auditable.

The plants in the top decile of OEE in the dataset overwhelmingly run a blended posture: roughly 30 percent reactive (the small, unpredictable failures that are not worth instrumenting), 50 percent preventive (the calendar-based service that catches the predictable failures), 20 percent predictive (the sensor-driven calls on the highest-criticality assets). The median plants typically run 60 to 70 percent reactive, which is the share that hurts OEE most.

Root cause analysis is the discipline that decides which failures move from reactive to preventive and which preventive items move to predictive. A weekly root cause analysis on the top three failures of the previous week, with the maintenance lead and the process engineer in the room, is the routine that moves the posture mix one percent per quarter in the right direction.

The "World Class 85 percent" myth

Three observations from the data.

One, 4 percent of plants in the dataset reach 85 percent OEE. That is 17 plants out of 412. Sixteen of them are in electronics, semiconductor, or aerospace, and one is in a continuous beverage line. The pattern is clear: world-class OEE in this dataset is reachable when the capital cost of the line forces availability-first design and when changeovers are rare or have been engineered down to near zero.

Two, in most industries the 85 percent number is mathematically out of reach in the medium term. A pharmaceutical line with a 35-minute cleaning between products runs at maybe 75 percent OEE even with zero unplanned downtime, because the cleaning is real and the cleaning is counted. The 85 percent number requires either no cleaning, no changeovers, or both. Most pharma plants will not get there in a five-year programme. The right target is industry-specific.

Three, the gap between the top decile and the median in every industry is roughly the same (12 to 16 percentage points). The opportunity is to close half that gap on the plant level. Closing the full gap requires capital and process changes that are usually not on the table.

The implication for an operations leader: forget the textbook number. Look at the top decile in your specific industry and set a target at 60 to 75 percent of the distance from current to top decile. That is a 24-month plan. The 85 percent number is a fundraising slogan.

Where the gap from median to top decile lives

We asked a subset of plants in the top decile of each industry what they would attribute their lead to. The answers cluster.

Plant-floor capture quality. The top decile plants overwhelmingly have automated capture covering more than 90 percent of running hours. The median plants have automated capture covering 50 to 80 percent of running hours, with the gap filled by manual logs or by simply missing data. The micro-stops we covered in our piece on unplanned downtime are the largest single category that the median plants miss.

Changeover discipline. The top decile plants in every industry have measurably faster changeovers than the median. The difference is rarely the tooling. It is the structured pre-staging, the rehearsed sequence, and the cross-training of operators that lets a changeover happen in 12 minutes instead of 18. SMED (Single Minute Exchange of Die) is the framework, but the discipline is what does the work.

The weekly review cadence. Top decile plants run a weekly OEE review that ends with one named action per line. Median plants run a weekly OEE review that produces a slide deck. The difference compounds quarter over quarter. We covered the shape of an effective weekly review in our piece on the 7 most important tasks in a process engineer's week.

Operator engagement on micro-stops. Top decile plants treat the 30-second stop as a real event with a real cause. Median plants treat it as background. The reason the top decile does this is partly cultural and partly technological: when the capture system surfaces the micro-stop automatically, the operator engages with it. When the capture system requires the operator to log it manually, the operator skips it.

What the data does not show

A few honest caveats.

The dataset over-represents plants that already care about OEE enough to submit data. Plants without any OEE capture, which the IndustryWeek 2024 benchmark suggested was still about 25 percent of mid-sized manufacturers, are absent. Including them would shift the medians down by an estimated 5 to 8 percentage points.

The dataset is weighted toward Europe (61 percent of plants) and North America (33 percent). Asian and Latin American manufacturing patterns are under-represented. The patterns we know exist (very high availability in Japanese automotive, very low changeover frequency in Korean electronics, high variability in Latin American mid-sized food and beverage) are not reflected in the per-industry medians.

The numbers are 12-week aggregates. Single-shift, single-week, or single-day variability is much higher than the per-industry ranges suggest. A plant with a median OEE of 71 percent can routinely have shifts at 40 percent and shifts at 88 percent.

OEE is one input among several. The plants in the top decile of OEE are not always the most profitable, the most reliable, or the safest. They are the most available, the fastest, and the most consistent. Those are valuable but not sufficient.

The bigger picture

OEE numbers landing in the high 60s as a median is a reminder that the basics work and there is still meaningful room. The plant that goes from median to top decile in its industry over 24 months captures somewhere between 6 and 12 percentage points of output without buying new lines. At a typical EBITDA margin of 12 to 18 percent on manufacturing revenue, that is real money.

The trap is to chase the 85 percent slogan. The opportunity is to know your industry's top decile and to close half the distance to it.

For the broader frame on production visibility, see production monitoring system. For the calculation specifics, see OEE calculation. For the dashboard side, see OEE software.

FAQ

How does OEE differ across industries? Mainly through the structural availability ceiling. Continuous-process industries (paper, polymers, glass) sustain higher OEE because they have fewer changeovers and longer stable runs. Discrete and changeover-heavy industries (pharma, specialty food, textiles) have a lower structural ceiling.

What is a realistic OEE target for a mid-sized plant in 2026? Pick the top decile of your industry from the table above, take 60 to 75 percent of the distance from your current median, and set that as a 24-month target. For most mid-sized plants the target lands between 70 and 78 percent.

Is World Class OEE 85 percent? Only in electronics, semiconductors, aerospace, and the rare beverage line. For most industries the World Class slogan is unrealistic and unhelpful. Use the per-industry top decile from the table instead.

How do I improve OEE by 10 percentage points? In order: install automated capture, run a weekly review that ends with one action per line, attack micro-stops, attack changeover overruns, attack quality-driven stops. Twelve to eighteen months is realistic for a mid-sized plant doing this properly.

Can I trust OEE benchmarks at all? Yes, with caveats. Use industry-specific medians, not cross-industry ones. Look at top deciles for ambition, not at single-plant peaks. And remember that the OEE number is an output of how disciplined the capture is as much as how well the plant runs.

Use the numbers, then move on from them

The reason we ran this dataset is that the existing public benchmarks were old enough to be misleading. The reason we are publishing it openly is that the next dataset should be better, and the only way to get there is for more plants to share data and challenge ours. If you want to contribute to the 2027 update, join the community and post your aggregate numbers in the OEE thread.

For the related operating disciplines, see downtime tracking software, unplanned downtime, and shop floor data collection.

Start free or join the community to compare OEE numbers with peers from other plants.