OEE tells you what you lost. It doesn't tell you what to do.

OEE is a useful score and a poor operating system. This guide covers what the metric is good for, where dashboards built on it fall short, and what to look for beyond the score.

OEE (Overall Equipment Effectiveness = availability × performance × quality) is useful as a trend metric but poor as an operating system: it reports loss after the fact, assigns no owners, and averages away causes. Sites that outgrow OEE dashboards typically adopt an operating layer where every loss event carries an owner, an action or an investigation — and OEE is computed from those events.

What OEE is genuinely good for

Overall Equipment Effectiveness multiplies availability, performance and quality into one number. As a trend line it has real value: it makes loss visible, it lets you compare a line against its own history, and it gives a site a shared definition of "good". If you have no measurement at all, OEE is the right place to start.

Where OEE dashboards fall short

It's a lagging score, not a control

OEE reports what happened after the shift is over. By the time the number moves, the loss is already taken. Nothing in the score itself catches the miss at 10am while the shift can still respond — that's the job of short-interval control.

Nobody owns a percentage

A dashboard that says "OEE was 61%" assigns work to no one. Improvement happens when a specific loss carries a specific owner and a due date — and percentages don't have owners; actions do.

Averages hide the causes

A weekly OEE figure blends a bearing seizure, a slow changeover and a prover backlog into one number. The interventions for those three are completely different, and the average points at none of them.

It becomes the argument

When OEE is computed in a side system from re-typed data, meetings argue about the number instead of the plant — whose export is right, which stops were "planned", why the dashboard disagrees with the line. The metric consumes the meeting it was meant to inform.

What to look for beyond the score

  • Actionability: every number should resolve to an owner, an action, or an investigation. If it can't, it's reporting, not operating.
  • One source of truth: the shift board, the daily review and the weekly should be the same picture at three zoom levels — not three documents that disagree.
  • Asset-level cause tracking: issues and actions attached to assets, connected to the CMMS register — so the third bearing failure this year is visible as the third, not treated as the first.
  • Built on existing systems: the CMMS, historian and planning tools stay. If the fix requires replacing them, you've bought a migration programme, not an improvement.

The operating-layer alternative

Lumen takes this approach: an operating layer on top of the systems a site already runs. Instead of scoring yesterday, it shows which assets to act on next and puts the interpretation in the row — every line item carrying an owner, an action, or an open investigation. OEE can still be reported; it just stops being the operating system.

Asset reliability — next seven daysLive
Asset reliability view in Lumen — which assets to act on in the next seven days, with the AI interpretation in the row
FAQBeyond OEELumen IOI
Should we stop measuring OEE?
No. OEE remains a useful trend metric and a common language across sites. The argument is against using an OEE dashboard as the way a site runs its day — a score can inform the day; it can't run it.
What's a good OEE benchmark?
Widely quoted benchmarks put world-class around 85% for discrete manufacturing, with many real sites operating between 40% and 60%. Cross-site comparisons are mostly noise — definitions of planned downtime vary too much. Compare a line against its own history.
Isn't this just OEE with extra steps?
It's the reverse ordering. OEE-first starts from a score and hopes actions follow. An operating layer starts from the loss events themselves — each with an owner and an action — and lets the score be computed from them when needed.
Do we have to replace our existing systems?
No — that's the point. An operating layer reads from the CMMS, historian and planning systems a site already runs. The systems stay; what changes is that their numbers land on one live board.

See it on a running site.

Let us show you what Lumen would look like in your facility.

Thirty minutes, on a live board — no slide deck.