Task-Based Risk Assessment: The Method Standards Now Mandate

When ISO rewrote its industrial robot standard in February 2025, it made one method mandatory by name: ISO 10218-2:2025 requires a task-based risk assessment covering every mode a robot cell operates in, fault recovery and cleaning included. Machinery standards had been drifting that way for a decade.

A task-based risk assessment scores hazards where they actually live: in the task, the mode, and the moment, instead of averaging them across a machine or a department. The distinction sounds academic until the injury data lands, and NIOSH and BLS supply that data below.

Task-Based Risk Assessment in Six Lines
ISO 10218-2:2025 made task-based risk assessment mandatory by name for robot cells in February 2025, and ANSI B11.0 has long required it for US machinery.
NIOSH’s fatality analysis found 78% of robot-related deaths involved struck-by events, mostly in maintenance and fault recovery, the modes averaged assessments skip.
The method is five moves: task inventory across every mode, task-hazard pairs, a score per pair, controls in hierarchy order, and change-triggered review.
The worked press changeover below shows two of four steps scoring 20 before controls, and neither is normal production.
A JSA stops at steps and controls; scoring each pair is what lets one register rank tasks across a whole plant.
Daily task talks pay their way: ABC’s 2026 report credits them with a 59% TRIR reduction across 1.3 billion work hours.

The case in one number: NIOSH found 78% of robot-related deaths involved the machine striking a worker, most during maintenance or troubleshooting, exactly the modes an area-level assessment averages away. Tasks are where people die; tasks are where the scoring has to happen.

What a Task-Based Risk Assessment Changes

The unit of analysis moves from place to activity. An area assessment asks what is dangerous in this room; a task-based risk assessment asks what can hurt the person doing this specific step, in this mode, with this tool, and the answers rarely match. The room did not climb onto the machine; a person did.

The Modes a Task-Based Risk Assessment Must Cover

Figure 1. Five modes per cell, and the last three carry the injuries. Per ISO 10218-2:2025 and ISO 12100 practice.

Fault recovery and cleaning carry the injury load precisely because averaged assessments skip them. The 2024 Applied Ergonomics study of OSHA severe injury reports found unexpected activation during interventions among the leading incident archetypes, matching what our robot risk assessment guide documents in depth.

Vocabulary first, then method. Keep the difference between a hazard and a risk crisp, anchor terms with the definition of hazard and risk assessment, and the rest of this page turns those definitions into a register your supervisors will actually use.

Standards That Now Demand a Task-Based Risk Assessment

ISO wrote the requirement into machinery law’s reference texts. ISO 10218-2:2025 demands task-based assessment of robot applications across every operating mode, ANSI B11.0 has required it for US machinery for years, and ANSI/ASSP Z244.1 builds alternative lockout methods on that exact task logic.

Standard or guide	Scope	What it says about tasks
ISO 10218-2:2025	Robot applications and cells	Task-based risk assessment across all modes, mandatory
ISO 12100:2010	All machinery	Hazard identification per task and lifecycle phase
ANSI B11.0	US machinery safety	Task-based risk assessment as the core method
ANSI/ASSP Z244.1	Hazardous energy control	Task-based analysis justifies alternative methods
OSHA JHA guidance, publication 3071	General industry	Break jobs into steps, control step by step
ISO 31000	Enterprise risk	The umbrella process the task level feeds

OSHA never says task-based out loud, and its Job Hazard Analysis booklet teaches exactly that: break the job into steps, find the hazard in each, fix the worst first. The agency’s recommended practices then wire the worker input into the loop.

The enforcement backdrop rewards the granular. Fall protection led OSHA’s FY2025 citations at 5,914, lockout sat fourth at 2,177, and both standards fail at task level, on a specific rooftop step or one skipped isolation, never as abstractions in a binder.

The Numbers Driving Task-Based Risk Assessment Adoption

Figure 2. Four numbers, one direction. Sources: BLS, NIOSH, ABC, ISO.

Running a Task-Based Risk Assessment from Task Inventory to Controls

Method beats enthusiasm here, and the method is five moves. Inventory the tasks including every mode, pair each task with its hazards, score each pair, control in hierarchy order, and set the review triggers; our guide to conducting a risk assessment holds the generic loop.

Move	What you do	Output
1. Inventory	List every task in every operating mode, operator in the room	Task inventory with modes
2. Pair	Attach each hazard to the specific task it lives in	Task-hazard pair register lines
3. Score	Likelihood times severity per pair, worded scales	Inherent scores on a 5×5
4. Control	Apply the hierarchy: eliminate, engineer, then administer	Residual scores and owners
5. Review	Reassess on change, incident, or the calendar	Dated review log and triggers

From Job to Controls: The Task-Based Risk Assessment Pipeline

Figure 3. Hazards attach to task-hazard pairs, never to the job as a whole.

The task inventory is where most programs under-count. Watch a full shift, ask operators what they do when the machine jams, and mine the risk identification approaches in our library; the undocumented workaround is usually the highest-scoring line you will find. Ninety minutes of watching beats nine pages of assuming.

Scoring stays plain: likelihood times severity per pair, on scales defined in words. The definition of likelihood in risk assessment supplies the anchors, the 5×5 versus 4×4 comparison settles grid size, and qualitative and quantitative risk assessment explains when hard numbers are justified.

A Worked Task-Based Risk Assessment: Press Changeover in Four Steps

Theory proves nothing until it scores a shift. The table below runs a task-based risk assessment on one job, an injection molding press changeover, using the four steps a setter actually performs; our injection molding risk assessment covers the machine’s full file.

Task step	Hazard	Inherent	Control and residual
1. Lock out the press	Stored hydraulic and electrical energy	20 (4×5)	LOTO with verified isolation; residual 5
2. Remove tooling	Pinch points, 40 kg mold halves	16 (4×4)	Lifting aid plus two-person rule; residual 6
3. Clean hot surfaces	Barrel burns above 200 C	12 (3×4)	Cool-down interlock and gloves; residual 4
4. Restart, first cycle	Unexpected motion with guards open	20 (4×5)	Interlocked guard check before mode change; residual 5

Residual Scores in the Worked Task-Based Risk Assessment

Figure 4. Inherent versus residual per task step, from the table above. Controls applied in hierarchy order.

Two of the four steps score 20 before controls, and neither is production. That is the argument for the method in one table: the changeover an area assessment files under routine maintenance holds the two highest-risk moments of the machine’s week, and nobody had budgeted a second glance for either of them before the scoring made it visible.

Controls follow NIOSH’s hierarchy, engineering before procedure before gloves. Note the residuals never reach zero; a signed decision that 5 is acceptable, against written criteria, is the honest end state, and scenario-based risk assessment stress-tests the pairs you scored optimistically.

JSA, JHA, or Task-Based Risk Assessment: Which Does What

The three names overlap more than the acronyms admit. A JSA or JHA breaks one job into steps and controls; a task-based risk assessment does that and adds scored likelihood and severity per pair, which is what lets a register rank one task against another.

Tool	What it produces	Where it fits
JSA / JHA	Steps, hazards, and controls for one job	Designing or changing a task
Task-based risk assessment	Scored task-hazard pairs feeding a register	Prioritizing across many tasks
Daily toolbox talk	A spoken, task-level refresh each shift	Keeping controls alive in heads
Permit to work	An authorization gate for high-risk tasks	Hot work, confined space, live energy

The daily layer already exists on most sites. Our risk assessment toolbox talk guide gives the five-minute format its structure, CPWR’s free topic sheets feed it, and ABC’s 2026 data credits daily task talks with a 59% TRIR reduction across 1.3 billion work hours.

Concrete crews saw this pattern already. The concrete pouring risk assessment we published runs task-by-task through pour day, and manufacturing sites can wire the outputs into the KRI set our manufacturing examples post carries, so leading indicators track the worst pairs.

Task-Based Risk Assessment FAQs Practitioners Send Us

What is a task-based risk assessment?

A hazard evaluation that decomposes a job into tasks and operating modes, scores likelihood and severity for each task-hazard pair, and applies controls in hierarchy order. The scored pairs feed a register, which is the practical difference from a checklist-style walkthrough.

How is a task-based risk assessment different from a JSA?

A JSA stops at steps, hazards, and controls for one job; the task-based version adds a score to every pair so tasks can be ranked across the whole operation. Run the JSA logic inside it and you get both documents for one effort.

Which tasks should a task-based risk assessment cover first?

Start where the fatality data points: maintenance, fault recovery, changeover, and anything with stored energy or open guards. Our hazard identification and analysis guide ranks the hunting grounds, and the undocumented jam-clearing routine belongs at the top of the queue.

How long does a task-based risk assessment take per task?

Thirty to ninety minutes per task once the inventory exists, faster with the operator in the room. A press changeover like the worked example above runs about an hour including scoring, and our risk assessment templates cut the formatting time to zero.

Does OSHA require a task-based risk assessment?

Not by that name, and the agency’s own JHA guidance teaches the method step for step. Specific standards then demand task-level thinking anyway: lockout procedures are written per task, fall protection plans per exposure, and citations land on tasks, which settles the practical question.

How often should a task-based risk assessment be refreshed?

On any change to the task, tool, material, or crew, after any incident or near miss, and on a fixed cycle besides; our guide on how often risk assessments should be conducted defends the calendar. New modes are the trigger teams forget, so audit for them.

Shortcuts That Put the Averages Back Into a Task-Based Risk Assessment

Every shortcut below re-averages what the method just separated. They all look efficient in the meeting where someone proposes them, and each one hands back the blind spot that area-level assessment had in the first place; everyday safety risk management catches most of them early.

Shortcut	What it re-averages	Keep instead
Copying scores across similar tasks	Different modes hiding in look-alike jobs	Score each task-hazard pair on its own
Assessing only normal production	The maintenance modes where fatalities cluster	One pass per operating mode
Skipping the operator interview	The workarounds no procedure documents	Operator in the room for every inventory
Scoring the job, not the pair	Twenty hazards mashed into one number	One register line per task-hazard pair
Controls chosen by cost first	The hierarchy inverted at the worst moments	Engineering options priced before PPE
Reviews on anniversaries only	Every mid-year change to task or tool	Change-triggered reassessment plus the calendar

The 2026-2028 Horizon for Task-Based Risk Assessment

Standards will keep converging on the task. The ISO 10218 adoption into US practice arrived as ANSI/A3 R15.06-2025 in September, machinery revisions cite ISO 12100’s task logic each cycle, and the drafting direction across committees reads in exactly one direction.

Wearables and vision systems will feed the task inventory next. Cameras that flag guard-open interventions and sensors that log how often crews clear faults turn the under-counted modes into data, which fixes the inventory problem faster than any annual audit ever has.

ISO 31000’s next review will keep pulling operational detail upward. Boards increasingly ask which tasks carry the top residuals, insurers price on the answer, and a register built from scored task-hazard pairs answers in one query instead of one quarter.

The 5,070 deaths BLS counted in 2024 each happened during a task someone could have named. That sentence is the whole method: name the task, score the pair, control the worst first, and reassess when anything about the task changes.

 

Pilot a Task-Based Risk Assessment With Risk Publishing

One task is enough to prove the method. Send us your nastiest changeover or jam-clear, and Risk Publishing returns it as a finished register extract inside a week; the services page shows the full program version, and the contact form takes the task description.