Guide · For production managers and plant directors

OEE: the formula, a worked example with real numbers, and how to actually improve it.

OEE (Overall Equipment Effectiveness) is manufacturing's most quoted KPI — and one of its most mistreated: computed ten different ways, inflated for slides, compared across lines that measure different things. Used well, it's a single number that holds downtime, slowdowns, and scrap together, and tells you how much of your planned time actually produces good parts at the right speed. This guide takes it apart piece by piece: formula, full worked example, the six big losses, honest benchmarks, and the path to making it climb.

Published
10 min read
1 · The KPI

What OEE is, and why it's the KPI that matters in production.

OEE measures the overall effectiveness of a machine or line: how much of the planned production time turns into good parts, produced at the speed the equipment was designed for. An OEE of 100% would mean: never a stop, always at nominal speed, zero scrap. It doesn't exist in nature — and that's exactly why the number is useful: the distance from 100% is the map of your losses.

OEE's strength is that it forces three worlds to look at the same number: maintenance (accountable for downtime), production (accountable for speed), and quality (accountable for scrap). When OEE is measured honestly, Monday-morning discussions move from opinions to causes.

2 · The calculation

The formula: Availability × Performance × Quality.

OEE = Availability × Performance × Quality. Availability is the share of planned time the line actually ran (unplanned stops erode it). Performance is the real speed versus the ideal one (micro-stops and slowdowns erode it). Quality is the share of good parts out of total parts produced (scrap and rework erode it). Three ratios, multiplied: one of them dips and the total collapses.

A full worked example. An 8-hour shift = 480 minutes; 30 minutes of planned break → planned time 450 minutes. Unplanned stops: 50 minutes → run time 400 minutes → Availability = 400/450 = 88.9%. Ideal cycle: 2 parts per minute → theoretical output in 400 minutes = 800 parts; actual 680 → Performance = 680/800 = 85%. Good parts 646 of 680 → Quality = 95%. OEE = 0.889 × 0.85 × 0.95 = 71.8%. One line, three different losses — and now you know where to look.

OEE isn't estimated: either it comes from field data, or it's an opinion with decimals. Every number in the example — downtime minutes, counted parts, scrap — must have a measured source, not a box filled in from memory.
3 · Where efficiency goes

The six big losses: where your OEE ends up.

The TPM tradition groups losses into six families, two per factor of the formula. Their job is to turn an aggregate number into actions: each family has different causes, owners, and countermeasures.

  • Breakdowns (Availability)

    Unplanned stops from failures or malfunctions. Fought with preventive and predictive maintenance — and with honest cause logging, without which every maintenance plan is blind.

  • Setup and changeovers (Availability)

    The time between batch A's last good part and batch B's first. The classic lever is SMED: separating what can be done while the machine runs from what truly requires the stop.

  • Micro-stops (Performance)

    Jams, sensors to clean, upstream or downstream waits: stops so short nobody logs them by hand — and which, added up, often outweigh breakdowns. They're why automatic collection changes the game.

  • Reduced speed (Performance)

    The line runs, but below the ideal cycle: wear, out-of-spec materials, parameters «tamed» to avoid trouble. Invisible without an honest ideal-cycle reference.

  • Startup scrap (Quality)

    The parts lost at every start or changeover until the process stabilizes. Reduced by stabilizing startup parameters and standardizing recipes.

  • Process defects (Quality)

    Scrap and rework at steady state. Per-station, per-parameter traceability lets you attribute them to a cause — without it, you're debating perceptions.

4 · Manual vs automatic

Whiteboard, Excel, or PLC: how OEE data gets collected.

Manual collection — a whiteboard at the line, paper forms, Excel at end of shift — is where almost everyone starts, and it has one real merit: it forces you to define the causes. But it has three structural flaws: it only records long stops (micro-stops vanish), it arrives hours or days late, and data quality depends on whoever transcribes it at end of shift, in a hurry.

Automatic collection reads counts, states, and cycle times directly from the PLCs, through the SCADA/MES layer: every stop is logged the instant it happens, with exact duration; the operator keeps the value-adding task — assigning the cause, not reconstructing the minutes. It's the ARIA MES/SCADA model: numbers are born in the field, people add the context. The typical gap between declared OEE and measured OEE in the first weeks is double-digit — and that's where you find out where the time really goes.

5 · Reference numbers

Honest benchmarks: what a «good» OEE is.

The number quoted everywhere is 85% as «world class». Take it for what it is: a reference born in high-volume discrete manufacturing, with precise definitions of planned time and ideal cycle. Change the industry, the product mix, or the definitions, and the comparison loses meaning: a small-batch line with frequent changeovers isn't comparable to a single-product line running wall to wall.

What we see in practice: many lines, at their first honest measurement, sit between 45 and 65%. That's not failure — it's the starting point, and discovering it is already the project's first result. The comparisons that matter: the same line over time, the same shifts against each other on the same product, and losses in absolute terms (minutes, parts) next to the percentages.

Be wary of «conference OEE» above 90% declared without definitions: almost always the planned time generously excludes whatever is inconvenient, or the ideal cycle has been adjusted. A useful OEE hurts a little the first time you read it.
6 · The path

How to improve OEE: measure, attribute, attack, repeat.

OEE doesn't improve by «pushing harder»: it improves by removing losses, one family at a time. The path that works is always the same, and it's a cycle, not a one-off project.

  • 1 · Measure honestly

    Written definitions of planned time, ideal cycle, and causes; automatic collection where possible. The first two weeks are for surfacing the true picture, not for improving anything.

  • 2 · Pareto the losses

    Almost always, 2-3 causes account for more than half the lost minutes: a long changeover, a recurring micro-stop, a bottleneck station. The Pareto decides where to invest energy — not the latest complaint that reached the office.

  • 3 · Attack the top cause, with an owner

    One cause, one owner, one countermeasure, one date. SMED on the changeover, targeted maintenance on the recurring failure, revised parameters on startup scrap. Small and finished beats big and endless.

  • 4 · Standardize and start again

    The countermeasure that works becomes the standard (recipe, procedure, maintenance frequency) — and the cycle restarts from the new top cause. Andon and line-side KPIs keep the score visible to the people playing the game: operators and shift leads.

7 · What not to do

The calculation mistakes that inflate OEE (and make it useless).

An inflated OEE is worse than a low one: it removes pressure where it's needed and drives decisions on fake numbers. These are the mistakes we run into most often.

  • «Elastic» planned time

    Excluding inconvenient stops from planned time — material waits, meetings, extra cleaning — raises Availability without the line producing one more part. Exclusions are defined once, in writing, and never adjusted to make the numbers work.

  • Optimistic (or pessimistic) ideal cycle

    An inflated ideal cycle crushes Performance and demoralizes; a cautious one pushes it past 100% and hides slowdowns. The right reference is the nameplate cycle or the best demonstrated repeatable cycle — documented per product.

  • Comparing lines with different definitions

    If line A excludes changeovers from planned time and line B doesn't, the comparison rewards whoever measures worse. Align definitions first, compare numbers second — never the other way around.

  • Looking only at the percentage

    A +2% OEE can be worth hundreds of hours a year, but nobody notices unless you translate it: minutes recovered, extra good parts, shifts saved. Percentages convince charts; minutes convince people.

FAQ

Frequently asked questions about OEE

Questions Plant Managers ask before starting an MES/SCADA project.

How is OEE calculated, in short?

OEE = Availability × Performance × Quality. Availability = run time / planned time; Performance = actual output / theoretical output at the ideal cycle; Quality = good parts / parts produced. Example: 400 minutes running out of 450 planned (88.9%), 680 parts out of a theoretical 800 (85%), 646 good out of 680 (95%) → OEE = 71.8%.

What is a good OEE value?

The «world class» reference in the literature is 85% for discrete manufacturing, but it depends heavily on industry, mix, and calculation definitions. In practice, many lines sit between 45 and 65% at their first honest measurement — that's normal, and it's the starting point. The most useful comparison isn't the conference benchmark: it's the same line over time, with constant definitions.

Can I calculate OEE with Excel?

Yes, and it's a good way to start: it forces you to define planned time, ideal cycle, and causes. The limits arrive quickly: micro-stops go unrecorded, data arrives at end of shift, and maintaining the sheet becomes a job. Once OEE starts driving decisions, it pays to move to automatic collection from the PLC via MES/SCADA — numbers are born in the field and the operator only adds the cause.

What's the difference between OEE, TEEP, and OOE?

The denominator changes. OEE measures effectiveness over planned production time. OOE (Overall Operations Effectiveness) also includes part of operations' unplanned time. TEEP (Total Effective Equipment Performance) uses full calendar time — 24/7, 365 days — and tells you how much of the asset's theoretical capacity you're actually using: useful for deciding whether you need a new line or just better use of the one you have.

Does OEE make sense for manual operations too?

Yes, with judgment. On manual or mixed stations the «ideal cycle» is a standard time and collection is partly declarative, so the number is less precise than on an automatic line. It remains useful for trends and loss attribution — just don't compare it head-to-head with the PLC-measured OEE of an automatic line.

How does ARIA measure OEE?

ARIA MES/SCADA reads counts, machine states, and cycle times directly from the PLCs (OPC UA, S7): stops and micro-stops are logged the instant they happen, with exact duration, and the operator assigns the cause at the line. The Analytics suite in the Pro plan computes OEE, availability, performance, quality, and MTBF per station, line, and shift — on field data, not declarations.

Go deeper

Related pages for anyone who wants to raise OEE with data taken from the field.

ARIA MES/SCADA · Supervision and analytics

The platform that measures OEE as described in this guide: counts and states from the PLC, causes assigned at the line, analytics per shift and station.

Resources · How much does an MES or SCADA cost

The budget for moving from the whiteboard to automatic collection: pricing models, market ranges, and ARIA's public price list.

Resources · MES vs SCADA for Plant Managers

Where OEE calculation lives in the system chain: what SCADA measures, what the MES organizes, and why the two shouldn't be confused.

Let's talk

Got a project in mind?

Tell us about your idea or your operational challenge. Our team is ready to listen and propose the right technology fit.

Let's talk about your project