Check data quality before calculating OEE
Availability, performance and quality must use consistent production boundaries.
Overall equipment effectiveness (OEE) is the share of planned production time that turns into good parts made at the ideal cycle speed. It is the product of three factors: OEE = A × P × Q. A is availability, P is performance and Q is quality, and each is a ratio between 0 and 1.
The three factors measure three kinds of loss. Availability loss is planned run time lost to stops, such as failures, changeovers or waiting for material. Performance loss is running slower than the ideal cycle time. Quality loss is scrap and parts that need rework. Each factor is a ratio:
Planned stops, such as lunch breaks and scheduled maintenance, are removed from planned production time first. Under this definition they are not losses.
Hypothetical line: planned time 480 minutes, run time 420 minutes, ideal cycle time 30 seconds (0.5 minutes), 760 parts in total, 722 of them good.
As a check, run time and total count cancel in the product, leaving ideal cycle time × good count / planned time = 0.5 × 722 / 480 = 361 / 480 ≈ 0.752. Both routes give the same result.
The first mistake is counting planned stops as downtime. If lunch and scheduled maintenance stay in planned time, availability looks lower than the machine really is, and the team works on the wrong problem. The second is entering a wrong ideal cycle time. If it is set longer than the true ideal, performance can go above 100 %, which usually signals an input error. The third is counting reworked parts as good. A part that fails the first pass is a quality loss, even if it is repaired and shipped later.
It can, if it is scheduled and its length is consistent. An unplanned overrun at a shift change is a loss. Write the rule down and apply it to every line, so that numbers from different lines and months can be compared.
Yes, but first fix the definitions: stop reasons, counter logic and the ideal cycle time. If the counters cannot separate good parts from bad, quality cannot be computed from that data. Check one shift by hand against the automated figures before trusting the report.
Example numbers are hypothetical, not measured field results.