What's a Good OEE Score? The 85% Myth, Examined
The 85% figure gets treated like a law of physics. It's actually a 1980s illustration that escaped its textbook. Here's what real plants score, and a better way to pick your target.
Somewhere in the first month of every OEE rollout, someone asks the question: "So what should the number be?" And someone else answers, with complete confidence: "85% is world-class."
It's worth knowing where that figure actually comes from, because it carries a lot less authority than its tone suggests.
Where 85% came from
The number traces back to Seiichi Nakajima, the JIPM engineer who formalized OEE as part of Total Productive Maintenance in the 1980s. He suggested ideal-condition figures of 90% Availability, 95% Performance, and 99% Quality — which multiply to roughly 85% — as what excellent equipment operation could look like.
Two things got lost as that number traveled from Japanese maintenance literature into every consultant deck on earth. First, it was an illustration, not a survey — nobody measured a thousand plants and found the best ones at 85. Second, it described dedicated equipment running stable, mature products. Nakajima wasn't talking about a packaging line that changes over four times a day, and he certainly wasn't setting a pass mark for your plant.
What plants actually score
Ask people who've measured a lot of first-time OEE and the answers cluster tightly: most discrete manufacturing lines, measured honestly for the first time, land somewhere between 40 and 60%. Plenty land lower. The commonly repeated figure for an average plant is around 60%, and even that assumes the losses are being recorded rather than politely forgotten.
If your first honest number is 45%, nothing is wrong with your plant. Something is finally right with your measurement.
The spread between industries is enormous, and legitimate. A continuous process line filling the same product all week can genuinely run in the 80s. A job shop doing short runs with daily changeovers may be doing world-class work at 55. High-speed bottling sits somewhere in between, dragged down by changeovers and micro-stops no matter how good the crew is. This is why comparing OEE across plants — or across lines making different products — is mostly noise. The number is shaped as much by what the line is asked to do as by how well anyone does it.
The trouble with chasing a benchmark
Here's the practical danger of pinning 85% on the wall: when the gap between reality and target is unbridgeable, people close it with definitions instead of improvements.
Changeovers get reclassified as "planned time." Warm-up scrap stops being counted. The rated speed drifts down until Performance looks healthy. Every one of these edits is individually defensible, and collectively they produce a beautiful 82% on a line that hasn't changed at all. We've watched plants "improve" ten points in a quarter without a single extra case shipped. The report got better. The plant didn't.
An unreachable target doesn't motivate. It corrupts the data — and corrupt data is worse than no data, because you make decisions with it.
A better way to set the target
Skip the industry benchmark entirely and use the one line you can always beat: your own baseline.
- Measure honestly for four to six weeks. No target at all during this period — the goal is trust in the number, not the number.
- Take your average, and set the first target a realistic step above it. From 52, aim for 58. Not 85.
- Tie the target to named projects. "58 by December" should mean "because we're cutting the 1L changeover from 45 to 25 minutes and fixing the cap feeder starvation" — specific losses, specific owners. A target without a mechanism is a wish.
- When you get there, hold it for a while before raising it. Sustaining a level is its own skill, and sawtooth improvement — up for the push, back down after — usually means the gains were adrenaline, not process.
So, what's a good OEE score?
One you didn't flinch while measuring. One that's higher than your own number from six months ago, for reasons you can name. On most real lines, the honest journey looks like 45 to 55 to 62 over a couple of years, with each step paid for by specific, boring fixes.
That's less quotable than 85%. It's also the version that ships more product.
