Is 85% OEE really world-class?

The 85% figure comes from Seiichi Nakajima's original TPM formulation (about 90% Availability × 95% Performance × 99% Quality) and has been repeated since the 1980s. It is a legitimate reference point but not an audited standard, no standards body certifies OEE benchmarks, and appropriate targets differ sharply between continuous process lines and changeover-heavy discrete lines.

What is the average OEE in manufacturing?

Around 60% is commonly cited as typical, but there is no rigorous industry-wide measurement, published averages come from vendor surveys and consultant estimates using inconsistent calculation methods. Treat any "industry average OEE" as folklore with a grain of truth, and benchmark against your own line's history instead.

Why is my OEE lower after installing automatic monitoring?

Because automatic monitoring sees losses that paper logs miss, especially micro-stops under a couple of minutes and small speed losses. The line didn't get worse; the measurement got honest. A drop in reported OEE when measurement improves is common and is actually a healthy sign that you can now see the real losses.

Can you compare OEE between two production lines?

Only with heavy caveats. Product mix, changeover frequency, ideal cycle time policy, planned-time conventions, and measurement method all differ between lines and can move the score by double digits each. A cross-line OEE league table mostly ranks those differences. Compare each line to its own history, and use output against plan for cross-line questions.

Is a higher OEE always better?

No. OEE can rise for bad reasons: a trimmed denominator, deferred maintenance that mortgages next quarter, or a non-bottleneck machine running flat-out to build inventory nobody needs. Read OEE alongside units shipped, maintenance completion, and where the machine sits relative to the constraint.

What Is a Good OEE Score? An Honest Answer

Q: What is a good OEE score?

A good OEE score is one that is honestly measured, better than the same line's own baseline, and trending up. The commonly cited reference points are 85% for world-class discrete manufacturing, around 60% as typical, and around 40% for lines with no improvement program, but these are decades-old rules of thumb, not audited benchmarks.

Where the 85% world-class figure really comes from, why cross-line comparisons mislead, and the benchmark that actually means something.

A good OEE score is one that is measured honestly, trending up, and better than the same line's own history. The commonly cited scale calls 85% world-class, 60% typical, and 40% common for unimproved lines, but those figures are decades-old rules of thumb, not audited benchmarks, and comparing scores across different lines is usually meaningless.

"What's a good OEE?" is the first question every plant asks after calculating the number, and most answers on the internet are more confident than the evidence allows. This post gives the honest version: where the famous benchmarks come from, what they are good for, why cross-line comparison misleads, and what to benchmark against instead.

Where does the 85% "world-class" figure come from?

The 85% figure traces to Seiichi Nakajima, who developed Total Productive Maintenance at the Japan Institute of Plant Maintenance and published the framework in the 1980s. His formulation: roughly 90% Availability × 95% Performance × 99% Quality ≈ 85% OEE as the mark of an excellently run discrete-manufacturing operation. Nearly every "world-class OEE" claim you will read descends from that one source.

Two things follow. First, the figure deserves respect, it came from the person who invented the metric, based on plants he studied. Second, it deserves context: it is commonly cited not continuously audited. There is no standards body that certifies OEE benchmarks, no government survey that collects OEE the way the Federal Reserve collects capacity utilization (75.7% for U.S. manufacturing in May 2026, for macro flavor, a related but different metric). Published OEE "industry averages" are vendor surveys and consultant estimates with unknown measurement methods. Treat them as folklore with a grain of truth.

The commonly cited OEE reference bands. Useful for orientation; not an audited standard, and not a fair comparison across unlike lines.

Why does comparing OEE across lines mislead?

Because OEE is exquisitely sensitive to inputs and context that differ line to line. Two lines with identical management quality can honestly score 25 points apart:

Product mix and changeovers. A line running one SKU for a week will out-score a line running eight changeovers a day, no matter who manages better. Changeover-heavy lines carry a structural Availability penalty.
Ideal cycle time policy. One line uses nameplate speed, another uses "budgeted rate." The second line's Performance is inflated before anyone walks the floor. Same plant, incomparable numbers, the calculation choices in how OEE is computed swing results by double digits.
What counts as planned time. Whether changeovers, planned maintenance, or sanitation count against Availability varies by plant convention. Every convention is defensible; mixing them is not.
Measurement method. A line with automatic stop detection sees every micro-stop; a line on paper logs misses most sub-two-minute stops entirely. Ironically, the better-measured line often shows the worse score. Better data lowering the number is a known effect, and a healthy sign.
Process type. Continuous process lines routinely and legitimately run far higher OEE than discrete lines with heavy changeovers; the commonly cited world-class threshold differs by industry for exactly this reason.

The practical conclusion: an OEE league table across unlike lines mostly ranks measurement honesty and product mix, not management. If leadership wants one number across the plant, give them output against plan, and keep OEE as each line's own improvement instrument.

How do expectations differ by process type?

They differ a lot, and in predictable directions. Continuous and near-continuous processes, paper, chemicals, bottling at steady state, can legitimately run OEE in the high 80s and 90s, because the process runs one product for long stretches at a fixed rate and stops are rare, large events. Discrete manufacturing with frequent changeovers, packaging with daily SKU swaps, machining job shops, co-packing, structurally cannot match that, because every changeover is an Availability hit the continuous plant never takes.

Batch processes sit in between, and their scores depend heavily on convention: whether cleaning and sanitation count as planned exclusions or Availability losses can move the number ten points by itself. Food and beverage plants with allergen changeovers and clean-in-place cycles routinely make this choice differently from one another, which is another reason their scores don't compare.

So when someone quotes a single world-class threshold across all of manufacturing, the honest response is: world-class at what? A changeover-heavy line at 68% with fast, disciplined setups may be operating closer to its physics than a single-SKU line coasting at 82%. The score's ceiling is set by the process design; management controls the distance to that ceiling. Improvement work, and credit, should be judged on distance closed, not the raw number.

What should you benchmark against instead?

Benchmark each line against itself, decomposed into factors, over time. That comparison holds product, measurement method, and conventions constant, so a change in the number means a change in reality. The working sequence:

Freeze the calculation. Fix the time base, ideal cycle times, and exclusion rules in writing. No benchmark survives a moving definition.
Baseline four to six weeks. Capture normal variation before judging anything, single-shift OEE bounces around for reasons that have nothing to do with improvement.
Decompose before you target. A 55% OEE from 70% Availability is a downtime problem; the same 55% from 70% Performance is a speed and micro-stop problem. The factor tells you where to work; the six big losses tell you what to attack within it.
Set the next target relative to baseline. Five points in six months is aggressive and realistic for most lines; "get to 85%" is a poster, not a plan.
Review weekly with the crew. Trend, top losses, one countermeasure per loss with an owner. The score is the scoreboard; the losses are the game.

A hypothetical same-line trend: noisy baseline, a specific countermeasure, and a step change you can attribute. This is the benchmark that means something.

When is a high OEE score actually bad news?

When it is bought with the wrong trade-offs. Three patterns to watch:

OEE up, output flat. Usually a shrinking denominator, planned production time trimmed to flatter Availability. Check units shipped before celebrating.
OEE up on a non-constraint. Running a non-bottleneck machine flat-out builds inventory, not throughput. OEE matters most at the constraint; elsewhere, simple downtime tracking is often enough.
OEE up, maintenance deferred. Skipping planned maintenance boosts Availability this month and mortgages next quarter. Watch OEE next to maintenance completion, not instead of it.

A trustworthy 58% that the whole crew believes, measured from machine signals rather than end-of-shift memory, the way plants running Harmony compute true OEE from the source (see how), is worth more than a flattering 80% nobody can explain. Run your own numbers through the OEE calculator freeze the method, and start the only comparison that pays: this line, this quarter, versus last.