Availability vs Performance vs Quality

OEE is one number, but you cannot improve a number, you improve its causes. That is what the three factors are for: they split the loss into Availability, Performance and Quality so you know exactly where to push. (New to OEE? Start with What is OEE?)

The three factors at a glance

Availability

The share of planned time the machine was actually running. It is dragged down by the two availability losses: breakdowns (unplanned stops) and setup & changeovers. To raise it: cut breakdowns with predictive maintenance and faster repair (MTTR), and shrink changeovers with SMED. Full playbook: how to reduce unplanned downtime.

Performance

How fast it ran versus how fast it could. Dragged down by minor stops (micro-stops) and reduced speed. This is the sneaky factor: those losses are too short to log, so they show up as an unexplained gap rather than identified downtime. On many lines it is a large recoverable loss, see the hidden factory. Performance can read over 100% only if your ideal cycle time is set wrong, fix the input, not the machine. To recover it, see how to reduce micro-stops.

Quality

The share of parts that were good first time. Rework counts as a defect, not a good unit. Dragged down by process defects (steady-state scrap/rework) and reduced yield (rejects at startup/changeover). To raise it: SPC, process control and root-cause work on the defect modes. See how to reduce scrap and rework.

Which to attack first

Improve your lowest factor. The maths makes this obvious: because OEE multiplies them, lifting a factor that is at 60% returns far more than polishing one already at 95%. A worked example: at 90% × 70% × 99% = 62% OEE, Performance (70%) is the lever. Chasing the already-high Quality would barely move OEE; fixing the micro-stops behind Performance moves it a lot.

Map each loss to its factor with the six big losses, then go after the pair under your weakest factor.

A worked example from one shift

Take an 8 hour shift with a 30 minute break, so 450 minutes of planned production time. The line stops for 45 minutes in total, so it runs for 405 minutes and availability is 405 / 450 = 90%. The ideal cycle time is 1 minute and the line makes 360 parts, so performance is 360 / 405 = 88.9%. Of the 360 parts, 350 are good first time, so quality is 350 / 360 = 97.2%. OEE is 0.90 x 0.889 x 0.972 = 77.8%. The same answer comes from good parts times ideal cycle time over planned time: 350 x 1 / 450 = 77.8%. Try your own shift in the OEE calculator.

Common measurement mistakes

• Using a slow "standard" cycle time instead of the ideal one, which makes performance look better than it is.
• Leaving short stops out of the downtime log (see OEE data collection methods), so they vanish from availability and hide inside performance.
• Counting reworked parts as good, which lifts quality above the true first pass yield (see how to calculate first pass yield).
• Taking breaks out of planned time on one line and not on another, so the two lines cannot be compared.

The common thread

All three factors depend on measuring the line accurately, and Performance especially falls apart when short stops go uncounted. The partner we recommend, Fabrico, reads OEE straight from the PLC, shows camera video of stops, where cameras are fitted, with a suggested cause, and lets a technician open the work order from the stop, so each factor becomes visible and fixable. It is EU-built, with data hosted in the AWS EU region and holds ISO 27001 / 20000-1 / 9001 (supports audit-readiness). The tools here are free regardless.

The three factors on real machines

Jams and short stops on a filling machine or capping machine show up in performance, while rejects at a checkweigher or metal detector show up in quality. Each guide lists the common causes with a fix for each.

More guides · OEE Calculator · Six big losses