Ask two supervisors whether a trainee is ready to run a machine alone and you can get two answers, both defensible. That is the hard part of learning how to assess ILUO levels. The four letters take a minute to define, and the line between two of them decides whether the whole board tells the truth.
ILUO is a four-step scale for how far one person has come on one task: In training, Limited, Unsupervised, Operator. To assess an ILUO level, watch the person do the whole task and answer one question: could they do this right now, alone, to standard? The answer has to rest on something you can show, and if any of the four tests below isn’t met, mark Limited and set a date to watch again. The Training & ILUO Skills Matrix gives each task a place to record the evidence behind a sign-off and counts the coverage for you, but the test below works on paper too.
This post is about that judgment call. For the definition, see what an ILUO matrix is. For the board itself, see how to build an ILUO skills matrix.
What evidence does each ILUO level require?
Each ILUO level requires different proof, and the proof gets stricter at the Unsupervised line. A level is a claim about what someone can do, so decide what would prove the claim before you watch anyone work.
| Level | The question to answer | Evidence to record | Counts as coverage? |
|---|---|---|---|
| I: In training | Have they started learning this task? | Date training began and who is teaching it | No |
| L: Limited | Can they do it with help, or do parts of it unaided? | The parts they can do alone and what still needs supervision | No |
| U: Unsupervised | Can they do the whole task alone, to standard? | More than one observed solo run, with the date and the assessor | Yes |
| O: Operator | Can they also bring someone else up to standard? | A person they trained who then passed the Unsupervised test | Yes |
A blank cell means nobody has assessed that person on that task. Treat it as a gap.
Plants don’t all draw these lines in the same place. Merca’s guide to lean skills matrices (opens in new tab) expects a Limited operator to work under normal conditions, respect the cycle time and identify manufacturing faults, and an Unsupervised operator to prepare the workstation, manage start-up and resolve routine anomalies. Your wording can differ from that. It only has to be written down, so that two assessors reach the same answer.
The case for marking someone Unsupervised now
The strongest argument for signing someone off is that a board can be wrong in both directions. If a capable person sits at Limited for months, the board shows a coverage gap that doesn’t exist, and you spend training hours closing it.
Holding people back has other costs too:
- Your Operators become the bottleneck. Every extra observation comes out of the time of the people who are hardest to spare.
- People stop believing the board. If someone has been running a task alone on night shift for a year and still shows as Limited, the crew will treat the whole grid as paperwork.
- Trust is part of the training. At some point a person learns the task by being responsible for it.
All of this is true, and it is why “when in doubt, keep them at Limited” can’t be the whole policy.
The case for holding someone at Limited
The strongest argument for holding someone back is that an overstated board fails at the worst moment. You read the board when the qualified person is out, which is exactly when nobody is there to catch a mistake.
The two errors don’t cost the same:
- A wrong Limited is visible. The person is frustrated, says so, and you can fix it within a week.
- A wrong Unsupervised is invisible. Nothing looks wrong until the day they run the task alone and hit something they have never handled.
A false Unsupervised also removes the warning itself. A task with one qualified person shows up as a single point of failure on your team. Round a second person up and the flag disappears, while the actual risk stays exactly where it was.
The four-part test for an Unsupervised sign-off
The dividing line between the two sides is evidence. Someone is Unsupervised when all four of these are true, and “nearly” on any one of them means Limited.
- The whole task. Not just the steady running in the middle. Include start-up, changeover, shutdown and the routine faults the task throws up.
- Alone. The assessor watches and says nothing. A question answered or a prompt given means that run doesn’t count, however good the result.
- To standard. Judge the output against a written standard you chose before watching, such as a work instruction, a spec or a checklist. If the task has no written standard, write the SOP first. Without one, you are rating confidence.
- More than once. One clean run can be luck. Use separate days, and include at least one run where something isn’t routine.
Two or three observed runs is our suggested starting point for an ordinary task. It is a rule of thumb and not an industry standard, so raise it wherever a mistake could hurt someone or scrap a batch.
A worked example: one trainee, four observed runs
Picture a trainee working toward Unsupervised on final inspection. The written standard is a checklist: measure the sample, compare it to the spec, hold and log anything out of tolerance.
| Run | Conditions | Prompts given | Result against the checklist | What it proves |
|---|---|---|---|---|
| 1, Monday | Routine batch | 0 | Met standard | Can do the routine version alone |
| 2, Wednesday | Batch with an out-of-tolerance part | 1 (asked what to do next) | Met standard | Doesn’t count: they needed help on the non-routine case |
| 3, next Monday | Routine batch | 0 | Met standard | Second clean solo run |
| 4, next Thursday | Batch with an out-of-tolerance part | 0 | Held and logged the part correctly | Handled the non-routine case alone |
After run 2, the honest mark is Limited, even though the result was fine. Run 2’s output met the standard only because someone answered a question. After run 4 the trainee passes all four parts of the test and moves to Unsupervised.
If only one other person was qualified on final inspection, this sign-off takes the task’s coverage from 1 to 2 and closes a single point of failure. That is what a defensible sign-off buys you: a higher number you can believe.
How to run an ILUO assessment
Run an ILUO assessment as a short, scheduled observation. It should take about as long as the task does.
- Write down the standard first. Agree what “to standard” means for this task before anyone is watched.
- Have an Operator on that task do the observing. On this scale, Operator is the level that includes signing others off. If nobody on the task is an Operator yet, the supervisor assesses against the written standard, and developing an Operator becomes the next training job.
- Ask the person to rate themselves before the run. Don’t share your mark until after you have watched. Merca describes evaluating in pairs, with the manager observing and the operator self-assessing (opens in new tab), and treats any difference between the two ratings as a useful signal. In our view, a trainee who rates themselves below your mark may know about a gap you didn’t see.
- Watch without helping. If you have to step in for safety, step in, and count that run as not passed.
- Record the level, date, assessor and evidence. “Observed solo, routine and fault batches, met checklist” is enough.
- If the answer is no, write down what is left to prove and set a date. “Needs one unprompted fault-handling run by the 20th” keeps Limited from turning into a parking spot.
How to tell when someone has reached Operator
Someone has reached Operator when they can do the task alone and can also bring another person up to standard. Being the fastest or most senior person on the task doesn’t prove that.
Watch for three things: they can explain why each step exists, they can spot someone else’s mistake while it is happening, and they can coach a person through it without taking over. The cleanest evidence is a result. Have them train someone, then have a different assessor put that trainee through the four-part test.
Protect this level. The number of Operators on a task is how fast you can grow its bench, and losing your only one is the core of the bus factor.
When an ILUO level should go down
An ILUO level should go down when it stops being true. Skills fade, and processes change underneath them. Re-assess when:
- The process, machine or standard changes. Yesterday’s sign-off covered yesterday’s task.
- The person has been away from the task for a long stretch. Six months is a sensible default if you have no better number for a given task.
- A quality problem or near miss traces back to the task. Watch the run again before you decide what went wrong.
Merca’s guide recommends updating the matrix for each event that modifies skills (opens in new tab), such as validated training, a job rotation, or an arrival or departure, along with a quarterly review of the matrix as a whole. A dated board is also what makes a lowered level feel routine and not like a demotion.
Where an ILUO sign-off is not enough
An ILUO sign-off records competence according to your own standard. It doesn’t replace a legal requirement for a specific task.
Forklifts are a clear US example. OSHA’s powered industrial truck standard, 29 CFR 1910.178(l) (opens in new tab), sets requirements an ILUO board doesn’t cover, including:
- Training content. Training must include formal instruction, practical training and an evaluation of the operator’s performance in the workplace.
- Evaluation schedule. Each operator’s performance evaluation “shall be conducted at least once every three years.”
- Evaluator qualifications. The people doing the training and evaluation must have the knowledge, training and experience to do it.
- Certification. The employer must certify that each operator has been trained and evaluated.
An Operator rating on your board does not meet those requirements by itself.
A level also tells you what someone can do, which is separate from whether they are cleared to do it today. If certifications and safety-talk attendance decide who can be assigned, the Crew Safety & Training-Compliance Kit joins them into one verdict per person per station.
The verdict: hold at Limited, but put a date on it
When the evidence is incomplete, mark Limited. A wrong Limited costs a week of someone’s patience, and a wrong Unsupervised can cost you the task on the day you most need it.
The people arguing for a faster sign-off are right about the danger of a board that never moves, and the fix for that is the date. Every Limited near the line should carry what is left to prove and when you will watch again. That way the conservative mark stays temporary and the board stays honest.
Common questions about assessing ILUO levels
How many observations does an Unsupervised sign-off need?
More than once, and on different days. Two or three observed runs is a reasonable starting rule for an ordinary task; raise it for anything where a mistake could hurt someone or scrap a batch.
Who should sign off an ILUO level?
An Operator on that task, because on this scale Operator is the level that includes training and signing off others. If nobody on the task is an Operator yet, have the supervisor assess against the written standard and make developing an Operator the next training job.
Does Limited count as coverage?
No. Someone at Limited still needs help or supervision, so the task stops when the qualified person is away.
Can an ILUO level go down?
Yes. Lower it when the process changes, or when the person has been away from the task long enough that you would want to watch them again before trusting it.
Keep the evidence next to the level
A level with no evidence behind it is one person’s opinion in a colored cell. To try the scale on a single task, start with the free ILUO starter board. For a whole team, the Training & ILUO Skills Matrix adds an evidence column for each task, coverage counts, key-person-risk flags and a dated quarterly review log, in one file you own.
