Teach operators three things: what the number means in minutes and parts, how to log a stop the same way every time, and what happens with the data. Do it at their own line, in about 30 minutes, with a live count instead of slides. Check it worked by testing whether two operators code the same stop the same way.
Skip the slide deck
Operators do not need the history of TPM. They need to know what to do differently on their next shift, and why it is worth the trouble.
So set one goal for the training. After it, an operator can read their own shift's result, log a stop correctly, and say what will be done with that log.
What to teach, and what to leave out
| Teach | Leave for later or skip |
|---|---|
| Three losses, shown on their own machine: stopped, slow, bad | The history of TPM and the six big losses as a list to memorise |
| What counts as a stop, and how to code it | Software menus nobody on this shift will use |
| Why a jam of 30 seconds matters when it happens often | Benchmarks from other plants |
| What happens to the log, and who acts on it | League tables of shifts or people |
Supervisors need more: the three factors, a Pareto and the usual calculation mistakes. Train them first, because operators will ask them the hard questions the next day.
A 30-minute session at the line
Run it at the machine, during a planned pause or at shift change, with six or fewer people. Schedule all shifts within the same week, so nobody hears it second hand.
What it is for, and what it will not be used for. Say it first.
Two people watch the machine. One times stops, one counts parts.
Use their own count, on a whiteboard.
Practice with the real reason list, on real stops.
Who reads it, when, and how they will hear back.
Say the opening part yourself, and have the plant manager say it too if you can. Operator buy-in covers the fears you need to answer.
The count exercise
This is the heart of the session. People believe a number they counted themselves.
Example: the crew watches for 10 minutes on a machine with a 3-second ideal cycle. The machine stops once, for 1 minute, and they count 150 parts, of which 4 are bad.
Then ask the question that makes it land: "Which of these did we see and which did we not?" The crew will likely name the 1-minute stop, and may not notice that the slow part cost more.
In this example the 1-minute stop cost 60 seconds, and the speed loss cost 90 seconds (540 - 450). That single comparison explains why you also ask for short stops and speed losses to be taken seriously.
The stop-coding test
A Pareto is only as good as its codes. If two operators code one stop differently, the data splits, and the top loss hides.
Test this. Write 10 real stop stories on cards, hand them around, and have each person pick a code alone.
A low score is rarely the operators' fault, and usually means the code list is at fault. Rewrite the codes people disagree on, using downtime reason codes, and run it again.
Teach the awkward cases
These come up on the first night shift, so settle them in the session and put the answers on the wall.
- Waiting for material. It is a stop, even if the operator did nothing wrong. See waiting for material.
- Very short stops. Explain the threshold, and what to do with shorter stops.
- Two faults together. Code the one that stopped the line first.
- Not sure of the cause. Use a clear "unknown" code and say so, rather than guessing. Unknown downtime shows how to reduce it.
- Good or bad parts. Rework counts as not good first time, as good count explains.
Make it stick after the session
- Start the next day with the data. Show the crew their own top stop, within one shift of the training.
- Use a one-page guide with pictures. Keep it at the machine, in the crew's language.
- Train new starters in the first week, with a mentor from the same line.
- Repeat the 10-card test when the reason list changes, or when logs start looking thin.
- Tell them what changed. A small board with "you logged it, we fixed it" does more than any refresher course.
How to tell whether it worked
Do not use a quiz score. Watch what the data does in the following weeks, compared with what it did before the session.
- The share of stop minutes with a useful reason goes up.
- Stops get logged when they happen, not at the end of the shift.
- Operators ask about the number, or disagree with it, at the daily meeting.
- Night shift's log looks like day shift's. If not, why OEE differs between shifts can help.
Hold this training early in any program: it belongs in the first week of starting an OEE program. If someone asks why it matters, how to explain OEE has plain words.
Once a machine logs its own stops, training gets shorter, and the operator's part comes down to the cause. In Fabrico, a light MES for OEE, stops, speed and counts come live from the machine, and operators tag the cause at the line. See how Fabrico handles stop tagging.
What people ask next
How long should OEE training for operators take?
About 30 minutes at the line is enough to begin, plus a few minutes at each daily meeting over the following weeks. Practice at the machine counts for more than time in a classroom.
Do operators need to know the OEE formula?
They need the idea: stopped, slow and bad, each measured in minutes or parts. Working it through once on their own count helps. Supervisors should know the full formula well.
How do I teach OEE across language barriers?
Show it at the machine, use pictures on the one-page guide, and translate the reason codes. Have a crew member who speaks both languages run the coding test, so nobody is left guessing.
Who should deliver OEE training?
A supervisor or the program owner who knows the line, with a technician nearby to answer machine questions. An outside trainer can help the first time, but the crew must see their own leaders owning it.
What is the most common OEE training mistake?
Teaching the formula and skipping what happens to the data. Operators who see no action stop logging, as operator downtime logging explains.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.