OEE for small factories: finding your hidden machine hours

I like OEE because it puts one honest number on a question every factory owner asks. How much of this machine am I really using?

OEE stands for overall equipment effectiveness. The Lean Enterprise Institute defines it as a Total Productive Maintenance measure of how effectively equipment is being used, calculated as availability × performance × quality. It is also one of the standard factory KPIs in ISO 22400-2, the international standard for manufacturing performance indicators, first published in 2014.

You do not need software to start. You need one machine, a printed sheet, a counter and an operator who trusts you.

The three parts, in plain words

1. Availability asks whether the machine was running when it was supposed to. Availability = run time ÷ planned production time. Breakdowns and changeovers pull it down.
2. Performance asks whether it ran at full speed while it was running. Performance = (ideal cycle time × total units made) ÷ run time. This is where small stops and slow running show up.
3. Quality asks how much of the output was good the first time. Quality = good units ÷ total units made. Every reject, every reworked unit and the waste after each start count against it.

Planned production time is the shift minus the planned breaks, such as lunch and tea. If the machine stands idle because there are no orders, take that time out as well, and record it separately. That is a sales or planning problem, not a machine problem.

The ideal cycle time is the fastest the machine should make one unit of that product. Use the maker's rated speed, or the best speed you have held for a full hour, and write it down. Do not change it later to make the numbers look better.

A worked example on one shift (assumed figures)

All the figures here are made up, to show the calculation. Take a packing machine with an ideal speed of 30 packs a minute, which is 2 seconds a pack.

1. Availability = 390 ÷ 450 = 86.7%.
2. Performance = (2 seconds × 9,360) ÷ (390 × 60 seconds) = 18,720 ÷ 23,400 = 80.0%.
3. Quality = 9,080 ÷ 9,360 = 97.0%.
4. OEE = 86.7% × 80.0% × 97.0% = 67.3%.

You can check it with the shortcut. 9,080 good packs × 2 seconds = 18,160 seconds, which is 302.7 minutes. Divide by the 450 planned minutes and you get 67.3% again.

Each of the three figures looks respectable on its own.

Multiplied together, they say that 147 minutes of a 450-minute shift produced nothing good. That is nearly two and a half hours.

Over 2 shifts a day and 25 days a month, this example machine hides about 123 hours a month. That is roughly 15 extra eight-hour shifts, from a machine you have already paid for.

Where the hours go: the six big losses

TPM groups machine losses into six big losses. The Lean Enterprise Institute lists them as failures, adjustments, minor stoppages, reduced operating speeds, scrap and rework. Some TPM books split the quality losses differently, into start-up rejects and production rejects. The idea is the same.

Here is the example shift, split by loss.

Look at the biggest row.

It is not the breakdown. It is the performance loss, the time when the machine was "running".

There is a simple reason this loss hides. A breakdown gets noticed, because the whole floor hears the machine stop. A 30-second jam that the operator clears by hand, or a machine run a little slower because it behaves better that way, rarely gets written down anywhere.

How to capture the data without software

Use one sheet per machine per shift. At the top, write the date, shift, operator, product, ideal speed, planned production time, the counter readings at the start and finish, and the reject count.

Below that, keep a stop log with five columns: time stopped, time restarted, minutes, reason code and a short note.

Keep the reason codes few and plain.

1. B for breakdown.
2. C for changeover or adjustment.
3. M for waiting for material.
4. P for waiting for a person.
5. Q for stopping because of a quality problem.

Decide a cut-off for small stops, for example any stop under 5 minutes. Do not log each one in full. Make a tally mark for each, so they are counted without slowing the operator down. The cut-off is your choice, but keep it the same every week.

If the machine has no counter, an electrician can usually fit a basic counter with a sensor. Until then, count output from finished cartons (cartons × units per carton). Keep rejects in a separate, marked bin and count them when the shift ends.

Run this on one machine for four weeks before you spread it. Sit with the operator once a week and look at the sheets together.

Mistakes that make OEE useless

1. Comparing OEE between different machines or products. Compare a machine with itself, week after week.
2. Lowering the ideal speed to make performance look better.
3. Chasing a benchmark figure you read online before you know your own baseline.
4. Leaving out small stops because writing them down is boring.
5. Using OEE to blame operators. The day that happens, the sheets stop being honest, and then the number is worth nothing.

The first improvements to try

Start with the biggest loss on your own sheets, not the one in this example.

1. For changeovers, record one on your phone. Then separate the work that can be done while the machine is still running, such as bringing the next reel, the tools and the settings sheet. Only the rest needs the machine stopped.
2. For breakdowns, start a routine of daily operator checks and a logbook. Our preventive maintenance starter plan shows how.
3. For small stops, stand at the machine for one hour with a notebook. Write down every stop and what caused it. Often the pattern shows up within that hour.
4. For start-up waste, write a settings sheet for each product, so the first good unit comes sooner after every start.

Quality losses respond well to small, steady work. On our tape plant, our plant head took up scrap with the "1% better every day" idea from Atomic Habits. He did not launch a big project. He changed small things on the line, one at a time. Every kilo of scrap saved is material we had already paid for.

Use OEE before you buy more capacity

In the example, raising OEE from 67.3% to 75% would give about 11.5% more good packs from the same machine, the same operator and the same shift.

Before you sign for a second machine, put a value on that extra output. Then compare it with the payback of a new machine. Sometimes the new machine still wins. But you should know the hidden hours in the old one before you decide.

Print the sheet tonight and give it to one operator tomorrow. Tell him or her that you want honest numbers, and that nobody will be blamed for what the sheet shows.