ISO/TR 14121-2 is not a harmonized standard, and that is exactly why it is so useful. It is a technical report that gives practical guidance for risk assessment under ISO 12100. No ritual. No signature parade. Just one hard rule that too many companies still dodge: the risk assessment team must have competence that matches the problem being assessed.
That sounds obvious, but this is where weak assessments are born. Not because one person did the work. A single experienced designer, mechanic, or controls engineer can do an excellent job on a straightforward machine. The real failure starts when one person assesses everything, including hazards they do not truly understand, and then calls it experience. That is not competence. That is guessing with a spreadsheet.
If you want a machine risk assessment that stands up in the real world, and not just in a meeting room, you need to know where one competent person is enough, where specialist input is required, and how to leave a decision trail that shows why the machine was accepted.
What ISO/TR 14121-2 really says about competence
First, let us clear away a common myth. A risk assessment team is not there to inflate the attendance list. Five names on the document do not automatically make the assessment better than one name. If the people at the table do not know the machine, do not understand the process, and cannot challenge assumptions, you do not have a team. You have an audience.
Risk assessment does not need an audience. It needs competence.
ISO/TR 14121-2 takes a practical line. A team approach will often make the assessment more complete and more effective, but team size depends on the method, the complexity of the machine, and the process in which the machine operates. It also points out something that companies often forget: a team can be too large. Too many people can make it harder to focus, harder to challenge bad assumptions, and harder to reach a clear technical position.
That is the key difference. Do not invite everyone with a remote connection to the machine. Pick the people who actually understand the hazard under review.
On a simple machine with typical mechanical hazards, an experienced designer may be enough. During a controls retrofit, the critical person may be the controls engineer who understands operating modes, reset logic, stop functions, interlocks, fault behavior, and how the machine behaves after a failure. If there is a well-known type-C standard and the protective measures are obvious, you do not need to stage an interdisciplinary theatre production.
But ISO/TR 14121-2 also makes another practical point: the composition of the risk assessment team can change during the assessment. That matters. The noise specialist does not need to sit through every discussion about a crushing point. The occupational health physician does not need to review every guard. The ATEX specialist does not need to debate the HMI layout. Yet if their knowledge affects whether residual risk is acceptable, their input belongs in the process. Not as decoration. As a basis for the decision.
And if a full team is not practical or necessary, confidence in the outcome can still be increased through consultation with a person who has the right expertise, or by review from another competent person. That is not bureaucracy. That is engineering discipline.
When one competent person is enough
Let us be honest: not every machine risk assessment needs an interdisciplinary panel. That idea is good for consulting invoices and bad for real engineering.
If the machine has typical mechanical hazards, simple automation, well understood modes of operation, and obvious protective measures, one experienced person may be enough. A designer, mechanic, or controls engineer can carry out a solid risk assessment if they understand the machine and use conservative engineering judgement.
This is where a lot of teams overcomplicate the obvious. A moving part approaches a fixed part. There is a crushing point. You do not need a seminar to see the injury mechanism. The operator can reach a drive. Fine. Ask the basic questions. Is that access needed at all? Can it be blocked by a guard? Does it require an interlocking guard? Can the adjustment point be relocated outside the hazard zone? ISO 12100 explicitly treats reduction of exposure by locating adjustment and maintenance points outside hazard zones as a valid direction for risk reduction.
The same goes for tool change, cleaning, and minor intervention. If a person has to enter the hazard zone, ask the blunt questions first:
- Has the machine stopped?
- Has energy been isolated and dissipated?
- Is there a risk of unexpected start-up?
- Are sharp edges, hot parts, or crushing points present?
- Does the person have safe access and safe egress?
In these cases, the bigger problem is often not lack of a large team. It is lack of courage to record the risk as it really is.
Too many weak assessments are written backward. The author wants the result to be 'acceptable', so the hazard description gets softer, the severity gets diluted, and the residual risk magically lands where the project wanted it to land. That is not risk estimation. That is result shopping.
ISO/TR 14121-2 is useful here because it keeps both feet on the ground. When you estimate severity, the target is not the most extreme consequence you can imagine in theory. It is the most severe harm that can realistically occur. But the opposite mistake is just as dangerous: choosing only the most likely minor harm because it looks good in the table.
If fingers can be crushed, do not write superficial abrasion. If a hand can be cut by a blade, do not pretend you are only dealing with a scratch. If a sleeve can be caught and draw the arm into a drive, do not hide that behind a polite phrase such as contact with moving parts.
At the same time, do not become theatrical. Not every sharp edge means fatality. Not every rough contact means amputation. Good risk estimation is not a competition for the darkest story. It is an honest description of harm that can actually happen.
One competent person is often enough when they can answer three questions honestly:
- Do I understand how the harm can occur?
- Can I choose a credible conservative scenario?
- When uncertainty exists, does it push me toward stronger risk reduction rather than easier acceptance?
That last point matters most. Conservative judgement is good when it leads to a better protective measure. It is not good when uncertainty is used as cover for approval. That is not conservatism. That is gambling packaged in a table.
And remember this: the estimated risk level is only one input into the decision to stop risk reduction. Legal requirements, actual use, work organization, technical limits, and operating practice matter too. In high-risk situations, more detailed risk estimation may be necessary. That is exactly where one person can stop being enough.
When ISO/TR 14121-2 says one person is not enough
The turning point is simple. The question changes from 'What is the obvious protective measure?' to 'Do we really have grounds to accept the residual risk and release the machine?' That is a different level of responsibility.
For a straightforward crushing point, you can design a guard. For access to a drive, you can add an interlocking guard. For a dangerous maintenance point, you can relocate it outside the hazard zone. But some hazards do not yield to intuition. Some demand specialist knowledge, measurements, validation, or at least a competent review.
Noise: you are not measuring hearing loss by ear
Noise fools people because it feels familiar. A machine runs. It sounds loud. Someone stands nearby for five minutes and says it is tolerable. The table gets filled out: hazard is noise, harm is hearing damage, protective measure is hearing protection, residual risk is acceptable. Done. Except it is not done at all.
Do you know the emission level? The operator exposure time? Whether the operator is at that machine for twenty minutes a day or eight hours a shift? Whether the noise comes from one machine or the whole line? Whether hearing protection is actually worn? Whether the person must still hear alarms, speech, or other machines?
Long-term health harm is not assessed the same way as a sudden cut or crush injury. For noise, dose, level, and duration matter. If that information affects whether the machine is accepted, then measurement data, manufacturer emission data, and input from an occupational hygiene specialist, acoustician, or occupational health physician may be necessary. You do not need to bring them to every workshop. But you do need their competence when the decision depends on it.
'It did not seem that loud' is not a risk assessment. It is an impression.
Radiation: invisible hazards do not care about your confidence
With mechanical hazards, intuition often helps. You can see the blade. You can see the motion. You can see the crushing point. Radiation is different. Optical radiation, laser radiation, ultraviolet, infrared, electromagnetic sources — these are not hazards you should approve by instinct.
A mechanic may spot the source. A designer may build a guard. A controls engineer may interlock the access. Fine. But someone has to understand what is actually being assessed. What wavelength? What power? What exposure time? Are reflections possible? Is the radiation visible or invisible? Are the eyes at risk, the skin, or both? Does the guard attenuate the relevant range? Are the protective glasses selected for the actual source, or do they just look professional?
ISO 12100 lists radiation as its own hazard group for a reason. If you do not know enough, you can adopt a conservative position and design stronger protection. What you cannot do is approve the machine from a position of ignorance.
Dust, fumes, oil mist and ATEX: are you assessing the machine or the worker's lungs?
Emissions often vanish inside tidy spreadsheets. Dust. Fumes. Aerosols. Oil mist. Vapors. The document looks neat: extraction, ventilation, mask, cleaning instruction. Problem solved. Except dust is not just dust.
Wood dust is not metal dust. Plastic dust is not oil mist. Welding fume is not a mild nuisance by default. Some substances irritate. Some sensitize. Some are toxic. Some can create an explosive atmosphere.
So ask the hard question: do you actually know what the worker is inhaling, in what quantity, for how long, and under what operating conditions?
If not, the word ventilation does not rescue the assessment. Sometimes a safety data sheet, material data, extraction documentation, and conservative judgement are enough. Sometimes you need measurement or specialist advice. Sometimes the issue moves directly into ATEX territory. And when combustible dust can create an explosive atmosphere, housekeeping alone is not your main protective measure. It may support the strategy, but it does not replace assessment of ignition sources, dust accumulation, ventilation performance, material parameters, and process conditions.
Ergonomics: the operator will get used to it is not a protective measure
Ergonomics is where many assessments become visibly dishonest. Awkward posture. Repetitive motion. Excessive reach. Twisting the trunk. Poor HMI angle. Heavy parts handled at the wrong height. Then the record says training, medium risk, acceptable.
No. That is not serious work.
The body is not an accessory attached to the machine. If the task is repeated hundreds or thousands of times per shift, the harm may develop gradually rather than in one dramatic event. And that is exactly why simple risk estimation tools are often better at sudden mechanical or electrical harm than at long-term health effects such as ergonomics and noise.
Does every risk assessment need an ergonomist? No. But if the decision depends on repetitive work, force, posture, recovery time, or long-term strain, the team needs to know what basis it is using to declare the residual risk acceptable. 'They have always done it that way' is not evidence of safety. Sometimes it is just the history of a problem nobody bothered to name.
ISO/TR 14121-2 and safety functions: a green light proves nothing
Controls engineers are often critical members of the risk assessment team, but let us separate two different skills. Being able to write PLC code is not automatically the same as being able to assess or validate a safety function.
Documents love reassuring labels: interlocking guard, light curtain, emergency stop, safety scanner, safety PLC. It all sounds good. The real question is harder: does the safety function actually do what the hazard requires?
Do you know which hazardous situation and hazardous event it is supposed to prevent? Has the required PL or SIL been determined where relevant? Has stop time been considered? Has reset behavior been checked? Does a fault drive the system to a safe state? Has bypassing been considered? Does setup mode punch a hole through the whole safety concept? Has validation been done in line with ISO 13849 and the system architecture?
This is where confidence can become dangerous. 'It worked during acceptance' is not validation. A green indicator is not proof. It is just a signal. If acceptance of residual risk depends on the safety function, someone with real competence in that area must be involved.
Stored energy: switching off power is not the same as making it safe
Machines regularly teach the same harsh lesson: stopped does not mean safe. The motor is off. The panel is dark. Someone says go in. Then the hidden part of the story appears — pressure in the line, compressed air, a hydraulic accumulator, a loaded cylinder, gravity, a clamped component releasing, a tool dropping under its own weight.
So here is the blunt version: you isolated the electrical supply. Good. Who dissipated the stored energy?
ISO 12100 does not only talk about disconnection from the supply. It also requires stored energy to be dissipated or restrained where it can create a hazard. That means the assessment needs somebody who asks the right questions:
- Does energy remain after stop?
- Who dissipates or restrains it?
- How is zero energy verified?
- Does the LOTO procedure work in real maintenance, not just in the manual?
- Can a person end up under a suspended or pressure-supported load?
- Does a hose failure create a new hazardous event?
These are not side questions. They are design questions, maintenance questions, and release questions. And very often the best answers come from maintenance, not from the person filling out the table.
ISO/TR 14121-2 and the people who know the real machine
The most commonly ignored experts are the people who actually live with the machine: operators and maintenance.
The designer knows how the machine is supposed to work. The operator knows how it works when production pressure hits. Maintenance knows what happens when it stops working, when a sensor drifts, when an interlock gets bypassed, when a jam has to be cleared at 3 a.m., and when the neat instruction collides with reality.
ISO/TR 14121-2 is very strong on this point. When you analyse tasks and work situations, it is worth involving operating and maintenance personnel because the easiest or fastest way to do a task may not match the written instruction. That sentence should be printed above every risk assessment workshop.
On paper, a jam is cleared after stop and isolation. On the shop floor, someone may clear it 'just for a second'. On paper, the guard remains closed. In reality, it may obstruct cleaning. On paper, LOTO is a procedure. In reality, the question is whether it can be applied during a task repeated several times per shift without people finding a shortcut.
If nobody in the room has seen the machine used in anger, the team may end up assessing the instruction rather than the machine.
That does not mean the operator decides everything. It means operator and maintenance input are data for the assessment, just like drawings, schematics, measurements, standards, and failure history. Leave them out, and you may produce a very logical assessment for a machine that exists only in documentation.
How to document competence and decisions under ISO/TR 14121-2
If specialist input affects the decision, do not leave it as a corridor conversation.
This is a simple rule that saves a lot of pain later. If a consultation influenced risk reduction, machine release, or acceptance of residual risk, it should leave a trace in the documentation. That does not mean a twenty-page essay every time. It may be a note, a measurement report, a calculation, a safety data sheet, a validation record, a specialist email, or a reference to a type-C standard used as the basis for the decision.
What matters is that six months later somebody can answer these questions:
- Who assessed this hazard?
- On what basis?
- What data were available?
- What assumptions were made?
- What uncertainty existed?
- How did that uncertainty affect the risk assessment?
- Why was the protective measure considered sufficient?
- Why was the residual risk accepted?
ISO 12100 requires the risk assessment documentation to show the procedure followed and the results achieved, including significant assumptions, identified hazards, hazardous situations, hazardous events, the information used, the data and data sources, and the uncertainty linked to those data. ISO/TR 14121-2 adds the practical reason: documenting the process allows the decisions to be examined later.
That matters even more if the file supports CE under 2006/42/EC. A vague statement such as risk acceptable will not save the day when someone asks why the machine was released.
Weak documentation says: risk acceptable.
Good documentation says: risk considered acceptable because a specific protective measure was applied, its effectiveness was verified, defined assumptions were used, uncertainty was considered, and no further risk reduction was found necessary on that basis.
Those are two different worlds. One produces a table. The other leaves a defensible engineering trail.
Bottom line
Competence is not an administrative extra bolted onto risk assessment. It is part of the assessment.
Not every machine needs a room full of titles. But the opposite extreme is just as dangerous: one person filling in every line, judging everything from a crushing point to radiation, from noise to explosive dust, from ergonomics to a safety function, and then signing off as if that closes the matter.
It does not.
ISO 12100 asks you to think about the machine, the person, the tasks, the hazards, the hazardous situations, the hazardous events, the possible harm, and the required risk reduction. ISO/TR 14121-2 gives the practical filter: the people involved must have competence that matches the issue being judged.
The simplest rule is still the best one. If you understand the hazard and can choose a conservative, technically justified protective measure, act. If you do not understand the hazard, but your judgement will decide whether the machine is accepted, ask someone who does.
That does not weaken the engineer. It separates the engineer from the person who merely completes the form.
The worst machine risk assessment is not automatically the one done by a single person. The worst one is the assessment where a single person judged everything, including what they did not understand, and nobody can later show why the machine was released.