Short answer
A job hazard analysis breaks a job into steps, names the hazards in each step and records the controls. That can make it a ready-made risk annotation layer for task data, potentially useful for hazard reasoning and safety-aware planning. It holds no motion or force data, and incident details, names and site specifics usually need redaction before licensing.
What a job hazard analysis contains
OSHA describes job hazard analysis as a technique that focuses on job tasks in order to identify hazards before they occur, looking at the relationship between the worker, the task, the tools and the work environment (OSHA 3071). Operating businesses write them for routine and high-risk jobs, and safety teams revise them after incidents, near misses or equipment changes.
Check whether the supplier's format is a table with one row per job step, with columns for what is done, what could go wrong and how the risk is reduced; formats vary. That is useful to a robotics buyer for a specific reason: the step structure is already there, and the hazard and control fields are the part that is hard to get from video, SOP prose or product catalogs.
A related document type is the energy-control procedure. Federal lockout/tagout rules require employers to develop, document and use procedures for controlling hazardous energy when employees service equipment; the content covers scope, purpose, authorization, rules and techniques, plus specific steps for shutting down, isolating and securing equipment (OSHA, 29 CFR 1910.147). The regulation has a narrow documentation exception, so not every site will have a written procedure for every machine. These are explicit safety-critical step sequences, and safety-critical steps in procedures discusses how to treat them in a corpus.
What this data can support
- Hazard reasoning. Pairing task steps with hazards and controls gives a model material for answering "what could go wrong at this step and what is the control?"
- Safety-aware planning. A planner that proposes steps can be checked against a hazard list, for instance flagging steps that need isolation before access.
- Annotation layers for other data. Hazard tags can be attached to demonstration clips or procedure steps that share the same task, giving a risk dimension to existing task data.
- Evaluation questions. Hazard identification from a described task is testable without any robot hardware.
- Escalation rules. Controls such as "stop and notify" show where a system should hand a decision to a person.
Equipment-specific operating limits are usually in equipment operating procedures, and abnormal situations in exception and emergency procedures; a JHA is often cross-referenced to both, so ask for the linked documents. These uses are about reasoning and constraint checking. For the related layer of physical-property constraints, see safety data sheets as handling constraints and hazard properties of objects.
Limitations and additional preparation
No physical content. A JHA says "keep hands clear of the pinch point"; it does not show how a person or robot achieves that. It contains no motion, force or perception data, and cannot train physical control.
Written for people. Controls often rely on judgment ("ensure area is clear") or on personal protective equipment, which a robot does not wear. Some controls do not translate into robot behavior; rewriting human procedures for robots covers what changes.
Uneven quality. Many are completed to satisfy a requirement. Steps may be copied between jobs, hazards listed generically, and controls out of date. A document with every row reading "slip, trip, fall" has little training value.
Coverage bias. JHAs tend to be written for high-risk or regulated tasks. Common low-risk tasks may have none, so a corpus may over-represent the dangerous work.
Local specificity. Hazards depend on the equipment model, layout and materials at one site. Transfer to another site is a claim a buyer should test, ideally with an evaluation set held out by site.
Preparation. Expect to normalize step granularity, map hazards to a controlled vocabulary, and link steps to the underlying procedure. A buyer that wants structure should plan for validating extracted steps if any parsing is automated.
Evaluating a sample
Sample at least a few JHAs per job family and read them as a reviewer would. Look for hazards that are specific to the step (a named pinch point, a stored-energy source, a chemical) versus generic entries repeated down the page. Look for controls that follow the usual hierarchy from removing the hazard to protective equipment, and check whether the document says who reviewed it and when. A JHA with a recent revision after an incident is often richer than one untouched for years, but it is also more likely to contain incident text that needs removal.
Decide the target vocabulary before purchase. If you intend to train a hazard classifier, free-text hazards need mapping to a fixed list, and that mapping is annotation work that should be scoped with the supplier or done in house. If you intend to use JHAs as retrieval context for a planner, structure matters less than clean, current text.
Redaction and sensitivity
JHAs are often written after incidents. They may carry employee names, injury descriptions, incident dates, claim numbers, investigation findings, equipment serial numbers and notes about regulatory contact. Some of this is personal information and some is legally or commercially sensitive. Redaction removes visible identifiers but does not by itself settle whether the document can be licensed, and a supplier's counsel may want to review any material tied to an incident.
Diligence questions: Who authored the document, and does the author's employment terms or union agreement affect use? Do any documents refer to open claims or investigations? Does the supplier treat safety records as confidential? Is any text copied from an equipment manufacturer's manual, which would carry its own rights? This is not legal advice. The general approach is covered under redacting SOPs.
Illustrative package
# Illustrative only - fields and values are examples, not an available dataset
package: redacted_jha_v0
supplier_type: operating business, EHS team
documents: 120 JHAs across 3 sites, revision dates preserved
fields:
job_id: pseudonymous
step_no: integer
step_text: string
hazard_text: string
hazard_class: controlled vocabulary (e.g. pinch, energy, chemical)
control_text: string
control_type: elimination | engineering | administrative | PPE
linked_procedure_id: optional
removed: names, injury narratives, incident IDs, serial numbers
excluded: incident reports, video, motion and force data
rights: training and internal evaluation; no redistribution; deletion on request
Questions to ask suppliers
- Who authored the JHAs, and who reviewed and approved them?
- How many jobs and distinct equipment types do they cover, and how many are generic templates?
- Are the revision dates and revision histories available?
- What was removed in redaction, and is any incident-derived text left?
- Are hazards and controls written in free text only, or tagged with a vocabulary?
- Can each JHA be linked to the SOP or work instruction for the same job?
- Are energy-control procedures included, and for which equipment?
- Does any text come from a third party, such as an equipment manufacturer's manual?
- Who in the company must approve licensing of safety records?
Explore sourcing and licensing options
If hazard annotations belong in your corpus, describe the task domain, document types and permitted uses to SourceX. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and coordinates the license and delivery. A request does not guarantee a match; SourceX follows up about potential sources. Start from the humanoid training data guide or the procedures hub, and see the site's page on SOPs and playbooks.
Explore sourcing and licensing options with SourceX
Sources
- Job Hazard Analysis (OSHA 3071) — JHA as a task-focused technique for identifying hazards.
- 29 CFR 1910.147 The control of hazardous energy (lockout/tagout) — documented energy-control procedures during servicing.