Overall equipment effectiveness: what the measure is for, and how it gets gamed

A loss ledger compressed into a single figure

The purpose is not the figure. The purpose is the structure underneath it: stops with reasons, running slower than the equipment can go, and product that had to be scrapped or reworked. Each category decomposes into causes different people can act on, such as changeover, minor stoppages, tooling, late material, absent operators and drifting settings. A plant that reports the headline and never opens the categories has bought a summary of its problems without buying any of the information needed to fix them. Ask to see the loss reasons before you ask to see the score.

Why the factors multiply rather than average

Each factor measures what survived the previous one. Of the time the machine was scheduled to work, some was lost to stoppages; of the time it did run, some capability was lost to slow running; of the units made, some were not good. Multiplying gives the fraction of scheduled time that produced acceptable output at the intended rate, which is a physically meaningful quantity. Averaging would let a strong factor compensate for a collapsed one, so a machine running constantly while producing mostly scrap could still look respectable. The multiplication is deliberately unforgiving, and that property is worth keeping.

The definitions decide the answer, so it cannot travel

Everything rests on choices that are not standardised: whether the denominator is calendar time, staffed time or scheduled time; whether planned maintenance, changeovers, breaks, trials and no-demand periods are excluded; and which rate counts as the equipment's capability, whether nameplate, best ever achieved, or the rate for this particular product. Change any of those and the figure moves without anything happening on the floor. That is why the measure is meaningful only against itself, on the same asset, with definitions frozen and written down, and why comparing your figure with another company's compares conventions.

The predictable ways the figure gets inflated

Where the score drives judgement of people, it is managed rather than improved. Slow-running losses vanish if the reference rate is quietly reset to the rate actually achieved. Downtime is reclassified as planned. Short stops go unrecorded because logging one takes longer than clearing it. No-demand time is excluded so the machine is measured only while busy. Rejects are reworked and counted as good. None of this requires dishonesty and each step can be argued individually. The defence is a written definition, an occasional independent check of how losses were classified, and never putting the score into an incentive scheme.

Improving it on a non-constraint machine makes things worse

On equipment that is not limiting plant output, a higher score means the machine made more parts than the plant can use, or made them in larger batches to avoid changeovers. Both consume material early, occupy space, extend lead time and hide problems inside stock. The measure is legitimate on a constraint, on equipment whose loss pattern is being studied for a stated reason, and on assets where availability genuinely governs delivery. Elsewhere, tracking it invites exactly the local optimisation the improvement effort was supposed to remove, and supervisors judged on it will behave accordingly.

Frequently asked questions

What score should we be aiming for on our equipment?

There is no transferable answer, and the benchmarks circulating in industry rest on undisclosed definitions and different equipment types. A continuous process, a machining centre with frequent changeovers and a manual assembly cell are not comparable in principle. The useful target is directional: fewer unplanned stops, a smaller gap between actual and demonstrated rate, and less rework, judged against your own history with the definitions unchanged. Anyone quoting a universal figure is selling something.

Should we monitor this on every machine in the plant?

No. Instrumenting everything creates a large reporting burden and a strong pull towards making every machine look busy. Measure it where availability governs delivery, meaning the constraint, shared bottleneck equipment and assets with expensive downtime, and use simpler indicators elsewhere, such as whether the area met its plan and what stopped it. Adding machines to a monitoring system is easy and rarely questioned, which is exactly why the list grows past the point of usefulness.

Our score improved but output fell. What happened?

Usually the mix changed, or the machine was scheduled less and measured only while running favourable work. Both effects raise the ratio while the plant ships less. Other candidates are a reference rate revised downwards, downtime reclassified as planned, and longer runs of one product achieved by building stock nobody ordered. Reconcile the figure against units despatched and against the loss reason log before treating the movement as improvement; that reconciliation normally identifies the cause quickly.