What is a good OEE score?

85% OEE (overall equipment effectiveness) is a target, not a measured benchmark. It is the product of the minimums Seiichi Nakajima set in his TPM work: 90% availability, 95% performance and 99.9% quality. Published measurements sit lower, from 20–30% on one packaging line to about 76% across 15 presses. Set each machine’s goal from its own baseline and trend instead.

This post traces the number to its source, lists the public measurements, and shows why a job shop and a high-volume line start from different places. For the definitions, start with what OEE measures.

Where does the 85% figure come from?#

The 85% comes from total productive maintenance (TPM), the program Seiichi Nakajima led at the Japan Institute of Plant Maintenance from the early 1970s.^[2] In his 1984 book Introduction to TPM, he gave four world-class numbers: OEE above 85%, availability above 90%, performance above 95% and quality above 99.9%.^[1]

The glossary entry for OEE defines the three factors the way IoTFlows measures them.

Check it the other way round: 95% × 99.9% is 94.9%, and 90% of 94.9% is 85.4%. Many pages use 99% quality instead, which gives 0.90 × 0.95 × 0.99 = 84.6%. Both round to the 85% people quote.

The factor targets aren’t equal, and that matters. Three factors at 90% each give only 72.9%, so 85% needs near-perfect quality and close-to-ideal speed on top of few stops.

Nakajima set these numbers from practical experience as minimums for plants to strive for. He also noted that every plant winning Japan’s Distinguished Plant Prize for TPM scored above 85%.^[2] The figure describes award-winning TPM plants in Japan in the 1970s, in discrete manufacturing rather than process industries.^[2] It was never an average across plants.

What do plants actually score?#

Public measurements are few, and each covers a small number of machines under its own definitions. These four name their setting and their source.

Ljungberg found OEE of only around 55% across about 20 cases.^[3] The 15 presses in Raju’s tool and die shop averaged 75.99%, with availability at 87.02%, performance at 87.34% and quality at 99.98%.^[1] Quality beat Nakajima’s target there. Stops and speed losses kept OEE 9 points short.

The two NIST MEP projects start lower. Leggett & Platt Aerospace lifted OEE on its most problematic CNC lathe from 39% to 45% with a TPM program.^[4] Pladis Global’s snack packaging line in Horseheads, New York, started at 20–30% and passed 50%, against a goal of more than 60%.^[5]

And the 60% average that many pages repeat? None of them cites a survey, and the closest published measurement is the 55% across about 20 cases above.^[3] Quote 60% as a rule of thumb, if at all.

Two plants can also report different OEE for the same machine. Whether changeovers sit inside planned time and how the ideal cycle time is set both move the result, as the mistakes that change OEE show. For the same reason, OEE compared across dissimilar processes, or against an outside benchmark, says little.^[2]

Why do job shops and high-volume lines score differently?#

A job shop and a high-volume line lose time to different things, so the same OEE asks for different work. A job shop makes many part numbers in short runs. A high-volume line runs one product, or a few, for days at a time.

Every job change in a job shop is a setup, and setup time counts against availability. A line making 10 different products loses far more availability to changeovers than a line making one.^[2] Ljungberg also expected OEE to be lower when a company is new to a production process,^[3] and a job shop meets a new process with each new part number.

The same 40-minute changeover once a week on a high-volume bottling line costs 1.8% of the same 2,250 minutes. That line loses its time elsewhere: jams, short stops and slow running, which land in performance. Ljungberg found performance losses the dominant category across his cases,^[3] and the Pladis line shows that a high-volume line can start as low as 20–30%.^[5]

Loading sets a ceiling too. The utilization ranges by machine type put a manual CNC at 40 to 60% uptime and a fully automatic one at 75 to 85%. Uptime is availability, so a manual job-shop CNC sits far below Nakajima’s 90% before speed or quality count.

How do you set a goal from your baseline?#

Set the first goal 5 to 10 points above the machine’s current average, and pick the end of that range from its trend. The monitoring rollout plan uses the same rule for a first uptime goal: a line at 55% gets a goal of 60%.

The baseline is the machine’s average over the last full quarter, measured with the same shifts and the same rule for planned time. The trend is the direction of its monthly figures over the last six months.

1. Measure a full quarter

   Collect OEE and its three factors for one complete quarter, with the same definition of planned production time throughout.

2. Read the trend

   Plot the last six monthly figures and decide whether they are flat, rising or falling.

3. Set the goal

   Add 5 points when the trend is flat and 10 when it is rising. When it is falling, find what changed before you set any goal.

4. Hold it for a quarter

   Leave the goal alone, then raise it once the machine meets it. A stretch target should be achievable within three or four months.^[2]

Set OEE goals per machine#

IoTFlows keeps one OEE goal for the organization and lets you override it on a single machine, so a manual job-shop mill and an automatic cell each get a goal from their own history. The goal colors the Uptime gauge and the parts gauge green at or above it and red below.

A SenseAi on each machine measures uptime and counts parts with no PLC connection, and IoTFlows reports OEE factor by factor. To find a baseline and its trend, open the Advanced Report over the last quarter at a weekly interval: the bars show the trend, and each machine’s row gives its own figure. The guide to per-machine goals shows when an override is worth setting.

Which factor should you chase first?#

Chase the factor that loses the most minutes on the machine that limits output. In a job shop that is often availability, through setups; on a high-volume line it is often performance, through short stops and slow running.

A factor’s percentage sits on a different base of time from the next one’s, so convert each into lost minutes before you compare them. The guide to which OEE factor loses the most works that conversion through a week on a mill.

Choose availability when setups lead, and measure changeover time per job before you start a setup project. Choose performance when the machine runs most of the shift but output still lags its ideal. You don’t need a new OEE goal to start either one: the factor goal you set from the baseline tells you when the project worked.