← Insights

Task safety

JSA vs risk assessment: which one, when

July 2026 · 6 min read

A Job Safety Analysis (JSA) and a Job Hazard Analysis (JHA) are the same document under two names. Both break one specific task into its sequential steps and look hard at the hazard sitting inside each step. "JSA" and "JHA" are mostly US terminology - the same idea in UK practice is more often just called a risk assessment, which causes more confusion than it should when a team is trying to work out which document a job actually needs.

The method, in four moves

Strip a JSA back to its mechanics and there are four moves, done in order. List the steps of the task as it is actually performed, not as the manual describes it. For each step, identify the hazards present at that exact point - not the hazards of the job in general, the hazards of that step. Decide the controls that address them. Then score the risk of each step twice: once initial, before any controls are applied, and once residual, after the controls are in place and assumed to be working.

That second score is the part a lot of paperwork skips, and it is the part that matters. An initial score on its own just restates that the task is dangerous, which everyone doing it already knew. The residual score is the one that earns its place in a file: it is the evidence that the controls actually move the number, not just that a hazard was written down and acknowledged. If the initial and residual scores land in the same place, either the controls are not real or the step needs another look.

Narrow and deep, versus broad

The contrast with a risk assessment is really a contrast in shape. A JSA is narrow and deep - one task, followed step by step, from first move to last. A risk assessment is broad - a whole activity, a work area or a process, surveyed for the hazards present across it. Neither is a smaller or lesser version of the other; they answer different questions and most safety-conscious teams end up using both, often on the same job.

The step-by-step view earns its keep at the transitions - the moment between two steps, where a hazard exists that neither step shows on its own. A broad assessment, working at activity level, rarely has the resolution to see it.

This template is the US-style sibling of the Axiom Risk Assessment. Where that one covers an activity or area on a single 5x5 matrix, this one drills into one task at a time and scores every step of it before and after controls.

When to reach for a JSA

Pull out a JSA specifically when the task is high-consequence, non-routine, or when the order of operations is itself part of the risk:

• A die or tooling change on a press or machine. The danger is not the machine in general - it is the specific point in the sequence where a guard is open or the ram is under manual control.
• A confined-space entry. Atmosphere testing, isolation, entry and rescue arrangements each carry their own hazard, and each depends on the one before it being done correctly.
• A heavy or complex lift. Rigging, positioning and landing the load are three different risk pictures, not one.
• A lockout or tagout sequence. The whole point of the procedure is the order, so the analysis has to follow the order too.

In every one of these, a step is only safe because the step before it was completed properly, and a document written at activity level will not show that dependency. A JSA is built to show exactly that.

When a broad risk assessment is the right tool

The reverse case is just as common. A general activity or a whole work area, where the goal is coverage across many different hazards rather than depth on one sequence, is a risk assessment job. Think of a warehouse floor, a workshop area, or "working at a client site" as an activity - there is no single sequence of steps to walk through, so a step-by-step tool would be the wrong shape for the problem. Reach for the broad assessment when you need to see everything present in a space or an activity; reach for the JSA when you need to see everything present inside one task.

Getting the grain right

A JSA lives or dies on how finely the task is broken into steps. Two failure modes show up constantly, in opposite directions:

• Too coarse. A single step reading "carry out the die change" is not a step, it is the whole task restated. Nothing useful can be scored against it, because it hides every transition inside it.
• Too fine. Breaking the same task into forty steps, down to individual hand movements, buries the real hazards in noise and makes the document unusable in practice. Nobody reads a forty-row JSA before starting a job.

The right grain is usually somewhere between six and twelve steps for most tasks - each one a distinct, observable action with its own hazard profile. Write the steps with the people who actually do the task, not from a desk. They know where the awkward reach is, where the guard sticks, where the shortcut everyone takes actually lives - and a JSA drafted without them tends to miss precisely those points.

Finally, a JSA is a snapshot, and snapshots go stale. Set a review date and revisit the analysis whenever the task changes, the equipment changes, or the people doing it change - a new machine, a modified fixture, a different crew. A review date that has to be remembered will eventually be forgotten; a review date that flags itself on a register is the difference between a JSA that stays current and one that quietly stops matching the job it describes.