OEE

Running OEE as a Daily Management System

OEE only pays when it drives daily decisions. Here is how to compute it honestly, where the benchmarks sit, and the tiered cadence that turns a percentage into recovered capacity.

OEE decides whether you buy your next machine or find it inside the one you have. A line at 60 percent OEE that reaches 75 percent just created 25 percent more good output from the same capital, labor, and floor space. On a line worth 4 million dollars a year in throughput, that gap is a million dollars, roughly the cost of the new machine you were about to order. That is why OEE belongs in the daily management system, not in a monthly slide deck where it arrives three weeks too late to change anything.

The math has to be honest to be useful. Take a 480 minute shift with 30 minutes of planned breaks: 450 minutes of loading time. Lose 45 minutes to breakdowns and changeovers and availability is 405 divided by 450, or 90 percent. If ideal cycle is 30 seconds and you made 690 units, that is 345 minutes of ideal run against 405 available, so performance is 85.2 percent. With 669 good units out of 690, quality is 97 percent. OEE equals 0.90 times 0.852 times 0.97, or 74.4 percent. The OEE calculator runs this chain from raw shift inputs so every line computes it the same way.

Know the benchmark ranges before setting targets. World-class discrete manufacturing sits around 85 percent, built from roughly 90 percent availability, 95 percent performance, and 99 percent quality. Typical plants measure in at 55 to 65 percent, and a first honest measurement often lands at 40 to 50 percent, which is normal and useful, not shameful. Continuous process plants run higher, often above 90 percent, so never benchmark a packaging line against a refinery. The target that matters is a 3 to 5 point gain per year on your own constraint, not someone else's number.

Improvement comes from attacking the six big losses in order of pareto, not preference. Breakdowns and changeovers eat availability; typical plants find 30 to 50 percent of changeover time is external work done while stopped. Minor stops and reduced speed eat performance and are the most underreported, often 10 to 15 points hiding in 30 to 90 second jams nobody logs. Startup rejects and running scrap eat quality. Pull the loss tree monthly: if changeovers are 40 percent of losses, run SMED events; if minor stops lead, put counters on the top three fault codes.

The failure modes are predictable. Padding ideal cycle time so performance reads 98 percent, which hides speed loss forever; ideal must be the best demonstrated cycle, not the standard. Reclassifying breakdowns as planned downtime to protect the availability number. Averaging OEE across dissimilar lines into one plant number that nobody can act on. And rewarding the score itself, which guarantees gaming within two quarters. If your OEE ever reads above 100 percent on any element, your ideal cycle is wrong, and every improvement decision downstream is wrong with it.

Run it on a tiered cadence. Each shift, operators log downtime with reason codes at the station within 5 minutes of the event, because end-of-shift recall loses 20 to 30 percent of minor stops. Daily, the tier 2 meeting reviews OEE by line in 10 minutes and assigns countermeasures on the top loss. Weekly, engineering refreshes the loss pareto and reports progress on the top three projects. Monthly, plant leadership reviews the constraint line's trend and moves capital or kaizen resources to whichever loss category owns the biggest slice.

World-class OEE practice is visible on the floor: 85 percent or better on constraint equipment, loss data entered at the point of occurrence, a pareto no more than 7 days old, and operators who can name their line's top loss and the countermeasure in flight. Those plants recover 15 to 25 points of OEE over 2 to 3 years, and each point on a constraint line is roughly 1 percent more sellable output. The score is the byproduct. The system, honest data, daily review, and relentless loss attack, is the product.

Published 2026-07-02.