A Risk Assessment Is a Prediction

When must you review a risk assessment? The law sets no fixed interval. Regulation 3 requires a review when there is reason to suspect the assessment is no longer valid, or when there has been a significant change. HSE also says to review your controls to make sure they are working, and to consider a review after accidents or near misses. The rest of this essay is about the first of those triggers, the hard one.

What does a risk assessment actually claim?

On paper, a risk assessment is a list: hazards, who might be harmed, the controls in place, what else is needed. Read another way, every line is a prediction. "Pedestrians use the marked walkway" predicts that people will walk there. "Reversing is done with a banksman" predicts that one will be there at four in the morning. "One-way system in the yard" predicts that drivers will follow it when the yard is full.

The law sets the purpose clearly. Regulation 3 asks every employer for "a suitable and sufficient assessment" of the risks, made "for the purpose of identifying the measures he needs to take". The assessment exists to choose the controls. Whether those controls keep doing their job is a separate question, and it is the one that matters on the ground.

HSE's own guidance on managing risk makes the same distinction in plain words: "Do not rely purely on paperwork as your main priority should be to control the risks in practice."

The assessment is right on the day it is written. Everything after that is a forecast.

When does the law say you must review it?

Regulation 3(3) is short. An assessment "shall be reviewed" if:

HSE's guidance adds a duty that is easy to read past: "You must review the controls you have put in place to make sure they are working." It suggests a review if controls "may no longer be effective", if the workplace changes, and "if your workers have spotted any problems or there have been any accidents or near misses."

Most organisations handle the second trigger well. A new building, a new process or a new piece of plant prompts a review almost automatically, because the change is visible and someone owns it. The first trigger is harder. It asks you to notice that something has stopped being true when nothing obvious has changed.

Why does the prediction stop being true quietly?

Because the conditions it rested on move gradually, and usually for good reasons.

Volumes grow, and the yard that had space for a one-way loop now has trailers parked across it. A walkway is moved a metre to fit a new racking bay. Agency staff join for the peak, trained on the written rules but not on the local habits that made them work. The night shift finds a quicker route between two doors. Each change is small and sensible. None of them, on its own, feels like a "significant change in the matters to which it relates".

James Reason's work on organisational accidents describes this well. He drew a line between active failures, the slips and decisions close to the event, and latent conditions: weaknesses that sit in a system, sometimes for a long time, until they combine with something else. A control that has quietly stopped holding is a latent condition. It is invisible in the document, because the document still says it is there.

This is not a failing of the people who wrote the assessment. They made a good forecast. The site simply moved on, as sites do. Our essay on work-as-imagined and work-as-done looks at why practice drifts from procedure; the point here is narrower. The assessment carries a date, and the world it describes does not.

What would give you a reason to suspect?

Incidents and near misses are one answer, and HSE names them. But they arrive after the control has already failed at least once. They tell you the forecast was wrong; they do not tell you it was becoming wrong.

The earlier signals are about the controls themselves. Is the walkway being used? Is the exclusion zone empty when plant is working? Does the one-way system hold at the busiest hour, not just at the quiet one? These questions can be asked of a control directly, without waiting for it to fail. Our piece on active and reactive monitoring sets out how HSG65 frames this; and the Check stage of Plan, Do, Check, Act explains why it is the hardest part to keep up.

What makes this practical is treating each important control as a claim that can be tested. Take the three or four controls the assessment leans on most. For each, write down what you would expect to see if it were working, on an ordinary shift. Then go and look, at a time you would not normally be there.

A control is a claim about tomorrow. The only way to test it is to look.

A question worth asking on Monday

Pick one risk assessment that was revised recently, perhaps after a change in layout or a new contractor. Find the control it relies on most. Then ask, without blame: when did anyone last see that control working, on a normal shift, and how do we know?

If the honest answer is "when we wrote it", that is not a failure. It is the most useful finding a review can produce, because it tells you exactly where to look first. Our worked transport risk assessment shows one way to write controls so that each can be checked.

Where this is heading

The idea that an assessment should be reviewed is as old as the regulations. What has changed is how much of the working day can now be seen, and how cheaply. The field has always known that paperwork is not the same as control. It is now much better placed to check the difference, shift by shift, and to treat each assessment as what it always was: a good forecast, worth testing.

Common questions

When must a risk assessment be reviewed?

Under regulation 3(3) of the Management of Health and Safety at Work Regulations 1999, an assessment must be reviewed if there is reason to suspect it is no longer valid, or if there has been a significant change in the matters it relates to. Where the review shows changes are needed, the employer must make them. HSE's guidance also says you must review the controls you have put in place to make sure they are working.

How often should a risk assessment be reviewed?

The regulations do not set a fixed interval. They tie review to reasons: a suspicion that the assessment is no longer valid, or a significant change. Many organisations also set a regular review date as a backstop, but a calendar date alone will not catch a control that has quietly stopped working between reviews.

What counts as a significant change?

Changes to the work, the people or the place that the assessment relies on. HSE's guidance gives examples such as changes to staff, a process, or the substances or equipment used. New plant, a new layout, new contractors or a large change in volume are common triggers in transport and logistics.

What does "suitable and sufficient" mean?

It is the standard regulation 3 sets for a risk assessment. In practice it means the assessment identifies the significant risks, considers who might be harmed, and identifies the measures needed to comply with the law, in proportion to the risks of the work.

Is a near miss a reason to review a risk assessment?

HSE's guidance suggests considering a review if workers have spotted problems or there have been accidents or near misses. A near miss at a point the assessment says is controlled is a direct sign that a control may not be holding.

References

Published 18 November 2026 · Last updated 18 November 2026

This piece starts from the ideas set out in How We Think About Safety.

Wondering what a camera would see at your gate? Request a Safety Mapping — one camera, two weeks, at no cost.