OEE is the product of three numbers: availability, performance and quality. If a machine was available for 90% of the planned time, ran at 95% of its nominal rate and produced 98% good parts, its OEE is 83.8%. Below: the formula, a worked example from a single shift, a calculator for your own numbers, and an honest list of what the indicator does not show.
What OEE is
OEE (Overall Equipment Effectiveness) answers one question: how much of the planned time did the machine spend making good parts at the rate it was designed for. Everything that took that time away — breakdowns, changeovers, slow running, micro-stops, scrap and rework — ends up in a single number.
The indicator makes sense for one machine or for the bottleneck of a line. Averaged across a whole plant it produces a figure that looks good on a slide and leads to no decision: a machine running 40% of the time and one running 100% average out to 70%, and nobody knows where to go with a spanner. So calculate OEE per asset, and aggregate at most within one line — stating which machine sets the pace.
The formula and a packaging-line example
OEE = availability × performance × quality, where:
- Availability = run time ÷ planned time;
- Performance = parts produced ÷ parts possible in the run time at the nominal cycle;
- Quality = good parts ÷ parts produced.
An eight-hour shift on a packaging line, with no planned breaks and no idle time for lack of orders:
| Item | Value | Source |
|---|---|---|
| Planned time | 480 min | 8-hour shift |
| Unplanned downtime | 45 min | breakdown 25 min, changeover 20 min |
| Run time | 435 min | 480 − 45 |
| Nominal cycle | 18 s/part | machine documentation |
| Parts possible in the run time | 1,450 | 435 min ÷ 18 s |
| Parts produced | 1,200 | machine counter |
| Scrap | 36 parts (3%) | quality control |
- Availability = 435 ÷ 480 = 90.6%
- Performance = 1,200 ÷ 1,450 = 82.8%
- Quality = 1,164 ÷ 1,200 = 97.0%
- OEE = 0.906 × 0.828 × 0.970 = 72.8%
Check it the other way round. In 480 minutes at an 18-second cycle the machine could have made 1,600 good parts. It made 1,164. 1,164 ÷ 1,600 = 72.8% — the same number. If the two calculations disagree, the definitions are wrong and there is no point going further until you find out where.
The breakdown matters more than the headline. Those 72.8% say that the biggest loss here is neither breakdowns nor scrap but pace: over 435 minutes of running, the line gave up 250 parts it could have made. That is usually micro-stops and reduced-speed running — the loss that is hardest to see by eye and most often recoverable without investment.
OEE calculator
Enter the three components as percentages — the result updates as you type:
The calculator multiplies three numbers, and that is all it does. The difficulty starts earlier: those three numbers have to mean the same thing in your plant six months from now and on the second shift too.
Definitions to agree before the first measurement
- Planned time. Do you include breaks, washdowns, planned maintenance and shifts without orders? If you do, you are measuring asset utilisation (TEEP) rather than OEE. Both are legitimate — they just need different names and must never be compared with each other.
- Nominal cycle. From the nameplate, from the process documentation, or from the best documented shift? An inflated cycle depresses performance and within a month nobody believes the report; an understated one produces performance above 100%, which signals an error rather than a success.
- Stop threshold. From how many seconds is a stoppage a stop? Everything below the threshold lands in performance as a micro-stop, everything above it in availability. Changing the threshold changes the loss picture without changing anything on the shop floor.
- Good part. Is a reworked part good? Does start-up after a changeover count as scrap? Settle it with quality control and write it down.
This is one workshop with production, maintenance and process engineering. Without it, any OEE figure can be dismissed in a single sentence, and the argument about the indicator replaces the argument about the losses.
Typical values: what counts as good OEE
TPM literature has long used 85% as the world-class level for a single machine — availability 90%, performance 95%, quality 99.9%. Just as often quoted is the estimate that an average plant sits around 60%. Treat both as reference points rather than targets: for a continuous single-product process 85% can be low, and for short runs with a changeover every two hours it is physically unreachable.
We know of no publicly available, reliable survey of OEE in Polish plants, so we quote no "industry average" here. The only comparison that means anything is your machine today against the same machine a quarter ago, with the definitions unchanged.
What drags OEE down, by industry
- Foundry. Furnace and mould availability and overhauls. Losses are large and rare, so a monthly average says little — look at the distribution of stops, not at one number.
- Food and beverage. Washdowns and format changes. The key decision is how washdown is classified: as planned time (OEE rises and measures the line) or as downtime (OEE falls and measures the shift). Both answers can be right, but only one at a time.
- Packaging and print. Format changes and start-up — start-up scrap can eat the whole quality component, which is why it is counted separately from production scrap.
- Assembly and appliances. Micro-stops at stations and response time to a call. Here performance is usually the weakest component, and the cause is invisible without automatic measurement.
What OEE does not show
OEE does not know whether the batch was worth making at all. A machine building stock has an excellent indicator and a worsening effect on the company. Nor does it see money: an hour recovered on a machine that is not the bottleneck adds nothing to sales. So before you optimise anything, establish which machine sets throughput — and measure that one first.
Maintenance needs two further numbers that OEE does not contain: MTBF, the mean time between failures, and MTTR, the mean time to repair. The first says how often the machine fails, the second how long it takes to get it running. They lead to completely different decisions (preventive maintenance versus spare-part availability and repair procedures), so report them separately.
Collecting the data automatically
Calculating OEE from the shift sheet has one advantage: you can start tomorrow. It has two drawbacks. The data arrives a day late, so it suits reporting rather than reacting. And it is recorded by the same people whose work it evaluates — not out of bad faith, but because a supervisor facing a breakdown has more urgent things to do than note down minutes.
Automatic measurement removes both problems: states, counters and alarms are read straight from the controllers, and the operator adds only what the machine cannot know — the downtime reason, chosen on the panel from a list agreed in advance. How that works in practice is on our production monitoring and OEE page; if part of the fleet has no controllers, or the machines are CNC, start with machine data collection.
Before you ask for a quotation it helps to know what such a rollout is made of — we break the price down in what a production monitoring system costs. And if you are wondering whether OEE needs a separate system at all, the comparison is in MES vs SCADA vs ERP.