Contractor Safety Governance: Which Model Holds?

Contractor work often begins with a familiar sentence: “The contractor owns safety.” That sentence sounds efficient, but it can hide the control failures created by the host employer’s schedule, design, access rules, procurement decisions, and supervision.

The better question is not who carries the contract. It is which governance model keeps critical controls verifiable when work crosses organizational boundaries.

Why contractor governance is a control problem

The International Labour Organization estimates that construction alone accounts for at least 108,000 work-related deaths each year, about 30% of occupational fatal injuries. The same ILO source reports that construction workers in industrialized countries are three to four times more likely to die at work than workers in other sectors.

Those figures do not prove that every contractor arrangement is unsafe. They do show why contractor governance belongs inside the host employer’s risk system rather than in a procurement appendix. The host controls the site, the interfaces, the work sequence, and often the commercial pressure that shapes the job.

OSHA’s recommended practices for host employers, contractors, and staffing agencies organize the problem around communication and coordination. NIOSH’s 2022 guidance for protecting temporary workers uses three practical areas, evaluation and contracting, training and supervision, and injury reporting and recordkeeping. Neither source treats a signed contract as evidence that the work is controlled.

Evaluation criteria for the three models

Compare contractor governance on five dimensions before choosing a model. First, ask who defines the critical controls. Second, identify who can stop work without commercial permission. Third, test whether the host and contractor share the same work plan. Fourth, check who verifies field conditions after the job starts. Fifth, examine whether incidents and near misses enter one learning and corrective-action process.

ISO 45001:2018 applies to organizations of any size, industry, or geography, including high-risk sectors such as construction, manufacturing, mining, oil and gas, and agriculture. Its broad applicability matters here because contractor control is not a specialist add-on for one sector. It is a management-system question about processes that the organization controls or influences.

A model that scores well on paper but fails one of these five tests is not mature. It is merely well documented.

Model one: client-led governance

In a client-led model, the host employer defines the safety requirements, approves the work method, controls access, sets the permit conditions, and monitors execution. The contractor supplies competence and labor, while the client retains the operating decision.

This model is strongest when the host owns the process hazard. A chemical plant, refinery, warehouse, or hospital cannot outsource knowledge of its own energy sources, traffic routes, emergency systems, or simultaneous operations simply because another company performs the task.

The weakness appears when client-led governance becomes client-controlled paperwork. A host may approve a method statement without observing whether the crew understood the isolation boundary, the rescue route, or the changed site condition. OSHA and NIOSH both emphasize coordination because the host’s knowledge and the contractor’s task knowledge are different forms of control evidence.

Use this model when the host owns the hazard, the work affects production systems, or the contractor is entering a tightly controlled operating environment. Put a named host decision-maker at the workface, not only a contract administrator in an office.

Model two: joint-control governance

Joint-control governance treats the host and contractor as one temporary operating system for the duration of the work. The host owns site and interface risks. The contractor owns task competence and execution. Both parties define the work plan, verify controls, review changes, and share stop-work authority.

This is the strongest default for high-risk work because it matches the actual distribution of knowledge. The contractor knows the task, equipment, and crew constraints. The host knows the plant, surrounding work, emergency response, and production interfaces. A control that requires both sets of knowledge cannot be verified by one party alone.

The Headline Podcast has repeatedly surfaced the danger of an “us and them” contractor culture. One guest described contractors as receiving the most dangerous jobs while being treated as a separate population. The practical implication is not that the two companies become legally identical. It is that their daily risk decisions must become visible to each other.

Joint control fails when shared ownership becomes vague ownership. The work plan should name who confirms isolation, who releases the permit, who checks the crew, who responds to deviation, and who closes the action. Shared goals need sharper roles, not softer accountability.

Choose this model for shutdowns, construction inside operating plants, maintenance with simultaneous operations, major lifts, confined-space work, and any project where a contractor’s task can change the host’s exposure.

Model three: contractor-owned governance

Contractor-owned governance gives the contractor primary responsibility for its safety system, supervision, task planning, and workforce controls. The host verifies that the contractor meets entry requirements and monitors the interface, but it does not manage each task.

This model can work for low-interface work in a defined area, such as a specialist service performed away from production hazards, provided the host has assessed the boundaries and the contractor can demonstrate competent control. It is not a license for the host to stop asking questions.

The central trap is confusing independence with separation. A contractor may own its method while the host still controls access, energy, traffic, weather exposure, emergency response, and the timing of simultaneous work. When those conditions change, the host’s responsibilities return to the center of the decision.

Use contractor-owned governance only when the task, area, hazards, and emergency arrangements are genuinely bounded. Require evidence of competence, supervision, equipment condition, reporting, and response before work begins, then define the triggers that move the job into joint control.

Decision matrix: which model fits the work?

The matrix is not a ranking of companies. It is a test of where exposure sits. A contractor-owned model may be appropriate for a bounded service and completely inadequate for a confined-space entry inside a live plant. A client-led model may protect a process hazard and still fail if the contractor’s task competence is not verified.

Recommendation by operating context

For work inside a live process, use joint control as the starting point. The host should define the site and interface barriers, while the contractor explains how the task will be performed and how the crew will respond to deviation. The permit should record decisions, not replace them.

For major projects, use joint control during design, mobilization, commissioning, and every change that alters access, sequencing, or energy. ISO 45001’s focus on operational planning and control supports this approach because the risk is created by the work system, not only by the person holding the tool.

For bounded specialist services, contractor-owned governance can be efficient, but the host should set a clear escalation trigger. A change in location, crew, equipment, weather, energy state, or adjacent work should force a new review. If the trigger is not defined before the job, the site will debate it under pressure.

For emergency work, use client-led command with contractor task input. The host knows the emergency system and must coordinate the broader response. The contractor knows the equipment and task constraints. A command structure that excludes either side creates blind spots at the worst moment.

A Headline Podcast conversation with Andreza Araujo and Dr. Megan Tranter would frame this as a leadership test. The question is whether leaders are willing to design one operating system for the work, even when the payroll, contract, and reporting lines remain separate.

What leaders should change before the next contract

Start with the work, not the vendor. Classify each contractor activity by host hazard ownership, interface complexity, simultaneous operations, and emergency dependence. Then assign the governance model before procurement finalizes the scope.

Write five decisions into the contract and the work plan. Name who sets critical controls, who verifies them, who can stop work, who must be notified when conditions change, and who closes corrective actions. These statements should match the actual authority at the workface.

Measure evidence rather than attendance. Count the percentage of high-risk jobs with a jointly verified work plan, the time between a reported deviation and control restoration, the number of interface changes reviewed before execution, and the recurrence of contractor-related corrective actions. These measures are useful because they test whether governance changes exposure.

Finally, review contractor performance with the same seriousness as internal performance. OSHA’s recommended practices aim to protect the entire worksite, while NIOSH emphasizes that host employers must address temporary-worker protection. A separate contractor dashboard can be useful for analysis, but it should not create a separate standard of care.

Headline Podcast note. Contractor safety becomes credible when the host and contractor can disagree about a control without disagreeing about who belongs in the safety system. Explore more leadership and safety conversations at Headline Podcast.

Conclusion: choose the model that matches exposure

Client-led governance fits host-owned hazards, contractor-owned governance fits bounded specialist work, and joint-control governance is the safest default when interfaces can change the exposure. The decisive factor is not who signs the contract. It is who can see, verify, and stop the work when the plan no longer matches reality.

Before the next contractor mobilizes, make that decision explicit. If the work crosses a live operational boundary, treat governance as a shared control and give both parties the authority and evidence needed to use it.