Manufacturing Execution

Your OEE looks high and the line falls short

In this article

Nori.mx6 min read

Monthly report87%Line 3and overtimeevery SaturdayHonest mathAvailability86.0%Performance85.3%Quality95.0%OEE69.7%The shift313 minuseful of 450

The monthly report says line 3 is running at 87% OEE. And yet line 3 does not make the week’s volume, there is overtime every Saturday, and the production manager has already asked for a quote on a second machine.

Both things can be true at once. The OEE math is almost never wrong. What is wrong is the numbers that go into it, and each of them can be inflated without anybody lying.

The math, in three lines

OEE multiplies three percentages:

  • Availability: of the time the line was supposed to run, how much it actually ran.
  • Performance: while it ran, how close it got to the ideal cycle.
  • Quality: of what came out, how much was good the first time.

An example with round numbers. An eight-hour shift with a half-hour meal break leaves 450 planned minutes. There were 63 minutes of downtime, so the line ran for 387. The ideal cycle is 30 seconds, which in 387 minutes allows 774 parts, and 660 came out. Of those, 627 passed the first time.

Math Result
Availability 387 ÷ 450 86.0%
Performance 660 parts × 30 s ÷ 387 min 85.3%
Quality 627 ÷ 660 95.0%
OEE 86.0% × 85.3% × 95.0% 69.7%

There is another way to read the same number, and it is the one worth remembering: 627 good parts at 30 seconds are about 313 minutes of useful work, out of 450 possible. That is what OEE is: the share of the shift that actually turned into product.

Five ways to inflate it without lying

Same shift, same parts, same operator. A few of these decisions, each one defensible in a meeting, are enough to turn 69.7% into 87%.

1. Calling “planned” what is not planned. The changeover, the start-up meeting, waiting for quality to release the first part, or material that did not arrive. If those minutes are taken out of planned time instead of being counted as downtime, availability goes up on its own. In the example, removing a 25-minute changeover takes it from 86.0% to 91.1%.

2. Using today’s cycle as the ideal cycle. The ideal cycle should be the machine’s design cycle or the best the line has demonstrated. Many plants use the engineering standard, which was set with slack, or the time the line makes on a normal day. With an “ideal” cycle of 34 seconds instead of 30, performance goes from 85.3% to 96.6%, and the line looks maxed out when it has four seconds a part to spare.

3. Counting a reworked part as good. If a part repaired at the rework station counts as good, quality measures what shipped, not what came out right. In the example, 23 reworked parts take quality from 95.0% to 98.5%. Rework costs time and people, and in that number it disappears.

4. Not logging short stops. Nobody writes a two-minute jam on the sheet. Those micro-stops add up and should show up as a performance loss. But with the inflated ideal cycle from point 2, they hide under the four seconds of slack and nobody sees them.

5. Averaging percentages. Plant OEE is not the average of each line’s OEE, and the month is not the average of its shifts. A short shift at 95% weighs the same as a long one at 60% if you average the percentages. The honest math adds up minutes and parts first and divides at the end.

The first three are enough: the same shift gives 91.1% × 96.6% × 98.5% = 86.7%. That is 17 points. In time, more than an hour per shift that the report says the line is using and that does not exist on the floor.

What believing it costs

An inflated OEE is not a reporting problem. It is a decision problem, because every decision made with it goes wrong in the same direction.

You buy capacity you already have. If the line “is at 87%”, the only way to get more out is another machine. If it is at 70%, there is an hour per shift to recover with a shorter changeover and material that arrives on time, and that costs far less than new equipment.

You quote at a pace the line cannot hold. The cost per part and the date you promise the customer come from the same capacity assumption. If that assumption is 87%, the part costs more than you think and the date you promised was calculated with hours that do not exist.

You attack the wrong loss. With all three factors inflated, the big loss does not show. The improvement team ends up working on quality, which is already good, while the hour is lost in changeovers nobody counts as downtime.

One more: there is a figure going around as “world class” OEE. Forget it. Comparing your OEE with another plant’s says nothing unless they calculate it the same way, and they almost never do.

How to measure an honest one in two weeks

You do not need to buy anything to start. You need one line, one sheet and written agreements.

  1. Pick the line that limits output. Not the one with the worst number, the one that decides how much leaves the plant. Improving the OEE of a machine that is not the bottleneck improves the report, not the volume.
  2. Write down what counts as planned time and what does not. Meals and breaks do. Changeovers, meetings, missing material and waiting for quality do not: they are downtime, and they get logged as such. That rule alone is half the work.
  3. Agree on the ideal cycle and sign it. The design cycle or the best demonstrated, per part number. It does not change when the line runs slow; it changes when the line proves a better one.
  4. Count first-pass quality. Reworked parts go in their own column. They are still shippable, but they are not good the first time.
  5. Log stops with a code, not free text. A short list, the same for every shift, as we covered in what must be ready before you connect a machine. With free text nobody can add anything up at the end of the week.
  6. Report all three factors, never just the product. OEE only tells you there is a loss. The three factors tell you where.

At the end of two weeks you will have a lower number and a list of losses sorted from largest to smallest. The second is worth more than the first.

Where a system fits

The sheet works for one line and two weeks. It stops working at six lines and three shifts, because hourly counting by hand becomes one more task for the operator and short stops stop being logged right when they matter most.

That is where an MES earns its place. Parts are counted straight from the machine, the stop is recorded the moment the line halts, and the operator only has to pick the cause on a screen. The three factors are calculated with the rules you agreed on, not the ones that suit the monthly report.

If you already have one line’s downtime sheet, it works as a sample. With it we build a free prototype: that line’s dashboard with its three factors calculated on your real data, so you can see how much the number moves before you decide anything. And if you do not know yet whether your problem is measurement or process, Chira asks you the questions in five minutes and leaves you a document with what to fix first.

Related service

Manufacturing Execution

Your production line connected, monitored and optimized in real time.

See the service

Articles on this service

One shift's week, 2027MTWTFS48 hours → 46 hours−2 hThrough 2030462027442028422029402030−2 h every Jan

Manufacturing Execution

Your 48-hour shift won't fit in January

On January 1, Mexico's maximum workweek drops to 46 hours and overtime is capped at 9. What happens to your plant's capacity and what your plan must know.

6 min read

One lineone monthNetworked PLCSignals onlyManual entryStop reasonsA closed list,same on every shift10–12Before connecting1Named downtime2What a good part is3The machine talks4Network on the floor5The operator gains6An owner from ops

Manufacturing Execution

What has to be ready before you connect a machine

An MES project rarely fails because of the software. It fails because the plant was not ready, and you can know that before signing anything.

5 min read