OEE Benchmarks by Industry: What 60% Actually Means
Someone in your organization will eventually ask: "Our OEE is 62%. Is that good?" The honest answer is that the question, as asked, has no answer. A 62% OEE on a job shop running forty changeovers a week and a 62% OEE on a single-SKU filling line are not the same situation, not remotely. This post explains why industry benchmark tables mislead, what directional ranges actually look like, and how to build a benchmark that means something: your own line, four weeks, trending against itself.
Why Published Benchmarks Mislead
OEE is only comparable within the same line, the same product mix, and the same measurement convention. Change any of the three and you are comparing different numbers that happen to share a name.
Same line, different mix. A line that runs one product all week has different losses than the same line running six products with changeovers between them. The changeover minutes land in availability, so product mix alone can move OEE by twenty points without anything on the floor getting worse.
Different conventions. One plant counts changeovers as downtime; another treats them as planned stops and removes them from planned production time entirely. One plant runs three shifts, another two, so the same machine accumulates different maintenance windows. Neither convention is wrong, but the resulting OEE figures are not comparable, and the benchmark table does not say which convention produced each row.
Self-selection. Plants that publish their OEE are plants that measure it well, which usually means automated data capture, which usually correlates with newer equipment and tighter processes. The average in a published table is the average of the best reporters, not the average of the industry.
And the famous number itself — 85% is "world class" — is a textbook composite, not a survey result. It comes from multiplying three assumed-best-in-class factors: 90% availability × 95% performance × 99.9% quality ≈ 85%. No industry study produced it. It is an illustration of how the multiplication punishes: three individually good factors multiply into one mediocre-looking number. Treat it as arithmetic, not as evidence about what real factories achieve.
Directional Ranges by Context
With that caveat, ranges are still useful for orientation. Everything below is a rule of thumb, not a standard. The reason each context skews the way it does matters more than the number itself.
Job shops and high-mix discrete manufacturing
Typically the lowest OEE of any context, often well below 60% on an honestly measured line. The reason is structural: many changeovers, many unplanned interruptions (missing material, drawing questions, priority changes), and ideal cycle times that are hard to pin down across varied parts. Low OEE here is not necessarily a problem — it is the cost of flexibility. The number to watch is not the OEE level but the trend: if availability erodes month over month, changeover discipline is slipping.
Packaging lines
Packaging machinery is fast, which punishes performance. At hundreds of units per minute, a two-second jam is invisible to an operator but destroys the performance factor, because the machine is designed to run near its ideal speed whenever it runs. Packaging OEE is typically dominated by micro-stops, and lines that measure only major downtime report a much rosier picture than their performance factor tells. Rule of thumb: packaging lines that measure honestly often sit in the 50–70% band, with performance the weakest factor.
Food and beverage filling lines
Filling is repetitive and fast, so performance is usually strong and quality losses (fill-weight rejects, seal failures) are visible and bounded. Availability is the swing factor: cleaning and format changeovers eat scheduled time. F&B lines with frequent SKU changeovers tend to run below lines dedicated to one format, for the obvious reason. Honest measurements commonly land in the 55–75% range depending on changeover frequency.
Pharmaceutical production
Pharma availability is eaten alive by cleaning and changeover: cleaning-in-place cycles, line clearance, and documentation between batches are mandatory and slow. Quality requirements also cap performance — the machine may not be permitted to run at its mechanical maximum. The result is that pharma OEE is often low on paper while the plant is running exactly as validated. A 45–60% OEE on a pharma line can be a compliant, well-run line; the same number on a bottling line would be a scandal. Judging a pharma line against a generic benchmark table is a category error.
CNC machining
Machining centers run long cycles on defined parts, so performance is usually the most stable factor — spindle load is what it is. Availability depends heavily on tooling: tool breaks and tool changes dominate downtime, and unattended shifts (lights-out running) change what "planned production time" even means. Machining OEE tends to be mid-range and, because the process is stable, responds well to small systematic fixes like tool-life management.
If there is one takeaway from the ranges: the reason a line skews matters more than where it lands. A low number with a structural cause (changeovers, validation) is normal. A low number with no structural cause is a to-do list.
Benchmark the Factors, Not the Product
Here is the trick that makes benchmarks actionable: decompose. Two lines can both report 70% OEE with completely different problems, because the factors multiply. The product hides the disease; the factors reveal it.
| Same OEE | Availability | Performance | Quality | The actual disease |
|---|---|---|---|---|
| 70% | 74% | 95% | 99.5% | Maintenance and changeover problem. The machine is not running enough of the time. |
| 70% | 95% | 74% | 99.5% | Process problem: micro-stops, wrong ideal cycle time, or speed losses. The machine runs, but slowly and jerkily. |
| 70% | 95% | 95% | 77.6% | Quality problem: scrap and rework. The machine produces fast and makes junk. |
| 70% | 85% | 85% | 96.8% | Spread thin: moderate losses everywhere, usually meaning no one has attacked any single loss category yet. |
Check the arithmetic on row one: 0.74 × 0.95 × 0.995 ≈ 0.70. Identical OEE, entirely different Monday morning meeting. This is why "our OEE is 62%, is that good?" is unanswerable and "our availability is 74% against our own 90% baseline, driven by changeovers" is a project charter.
When benchmarking, compare each factor against rules of thumb for the factor, and compare the product only against your own history:
- Availability below 85% on a dedicated line usually points at unplanned downtime or changeover method.
- Performance below 90% on a repetitive line usually points at micro-stops nobody logs, or an ideal cycle time that is wrong in either direction.
- Quality below 99% on a stable process usually points at a specific defect mode, not general chaos — go find it.
If you want to run your own numbers through the three factors, the live OEE calculator does the multiplication and shows each factor separately, so you can see which disease you have before deciding what to fix.
How to Build a Benchmark That Means Something
Skip the industry table. Build your own in four steps.
1. Pick one line. The one with the most economic weight or the most complaints. One line, not the whole plant; averaging across lines destroys the signal.
2. Fix the convention in writing. What counts as planned stop vs. downtime (changeovers included or not — decide and document), what the ideal cycle time is and where it came from, what counts as scrap. Half of OEE arguments are convention arguments in disguise.
3. Measure honestly for four weeks. Every shift, every stop, including the ugly ones. Manual logging will strain here — this is exactly where it decays — but four weeks of imperfect real data beats a year of cherry-picked numbers. If you need the mechanics first, work through how to calculate OEE with a worked example and the common input mistakes.
4. Trend against yourself. After four weeks you have a baseline. That baseline is your benchmark. Improvement targets are "beat last month on availability," not "reach 85% because a textbook said so." A line that moves 58% → 63% in three months with the loss reasons visible has accomplished more than a line that reports a constant, unexplained 82%.
This is also the right order of operations before any software purchase: measure manually first, find out what breaks, then decide what system you need. We make that argument fully in why you should measure OEE before buying a MES.
Frequently Asked Questions
Is 85% OEE realistic?
On a dedicated, high-volume, repetitive line with automated data capture and disciplined changeovers: occasionally, yes — for a week or a specific product. As a plant-wide steady state, no, not for most mid-market manufacturers, and the 85% figure itself is a composite (90% × 95% × 99.9%), not a survey finding. Set targets against your own trend, not against the textbook number. If your line is at 55%, a realistic 12-month target is 65–70% with named causes closed, not a leap to world class.
What OEE should a job shop expect?
Lower than a repetitive manufacturer, and that is fine. Job shops pay for flexibility in changeovers and interruptions, and those minutes land in availability. A job shop that measures honestly may see 40–60% and still be well run. The useful metric is the trend and the factor decomposition: if availability is eroding, look at changeover method and scheduling, not at the OEE headline.
Can I compare OEE between two of my own plants?
Only at the factor level, and only after checking that both plants use the same measurement convention. If one plant counts changeovers as downtime and the other as planned stops, the OEE figures are not comparable at all. Compare availability to availability, performance to performance, quality to quality — and audit the conventions before drawing any conclusion.
Benchmark Against Yourself First
Voltrus MES rolls up availability, performance, and quality per station and per line every hour, with consistent conventions and no manual logging. Four weeks of honest data, then trend against yourself.
See Voltrus MES