An investigation report might identify what went wrong, but who makes sure the organisation actually does something about it?
On 19 October 2026, the US Chemical Safety and Hazard Investigation Board (CSB) released a rather disturbing update concerning a fatal explosion at the Yenkin-Majestic/OPC Polymers facility in Columbus, Ohio.
In April 2021, an explosion and subsequent fire at the facility killed one worker and injured eight others. Some of those injuries were horrific, including severe burns, fractures and an amputation.
The CSB subsequently conducted an investigation and, in November 2023, issued its final report containing five safety recommendations intended to address the serious deficiencies identified.
Fast-forward almost three years, and the CSB has publicly reported that the company has neither implemented nor substantively responded to any of those five recommendations. According to the CSB, repeated attempts to engage with the company over more than two years were met with silence.
Think about that for a moment. A worker died, eight others were injured, a federal investigation identified serious failures, and the agency responsible for investigating the incident says it has received no meaningful response to its recommendations.
The investigation identified deficiencies involving pressure-vessel integrity, emergency preparedness, evacuation alarms, emergency training and protective clothing. These weren't trivial administrative matters. They concerned fundamental controls intended to prevent a catastrophic event or protect people when one occurs.
The CSB's public statement is particularly concerning, although it is important to recognise that the agency's account does not, by itself, establish the current condition of every control at the facility.
Nevertheless, it raises a question that every organisation conducting workplace investigations should be asking.
What is the point of conducting an investigation if nobody is accountable for ensuring its findings lead to meaningful change?
We've Become Very Good at Completing Investigations
Across the world, organisations invest considerable time and money investigating workplace incidents. They assemble investigation teams, collect evidence, interview witnesses, construct timelines, identify absent or failed defences, examine organisational factors and develop recommendations.
Some investigations are excellent. Others are considerably less so, but that's a discussion for another day.
The point is that we put enormous emphasis on the investigation process itself. We measure how quickly investigations commence, how long they take to complete, whether reports are submitted on time and whether recommendations have been entered into the organisation's corrective action system.
And then, in many organisations, something rather strange happens.
The investigation team disbands, the report is filed, and responsibility for implementing the recommendations is transferred to an action management system.
The investigation is considered complete.
The investigators move on to their next assignment. Management receives a summary of the findings. A series of actions appears on somebody's dashboard, complete with due dates and traffic-light indicators.
Eventually, those actions are marked as closed.
Everyone moves on.
But here's the uncomfortable question: has the organisation actually reduced its exposure to the circumstances that contributed to the incident?
Because completing an action and reducing risk are not necessarily the same thing.
The Difference Between Closing an Action and Making Something Safer
Consider a fairly common investigation finding. A worker is injured during an interaction between mobile equipment and pedestrians. The investigation identifies deficiencies in the site's traffic management arrangements, including inadequate segregation, poor visibility and inconsistent application of existing procedures.
The organisation accepts the findings and develops several recommendations.
The traffic management procedure is reviewed. Workers receive refresher training. Additional signage is installed. Supervisors are reminded of their responsibilities.
Three months later, every action is marked complete.
Excellent. The dashboard is green.
But has anybody returned to the workplace to observe how traffic movements are actually being managed? Have the changes addressed the physical interaction between pedestrians and mobile equipment? Are workers still having to cross operating areas because of the way work is organised? Have supervisors developed informal workarounds because the revised arrangements are impractical?
Or have we simply updated some documents, delivered some training and installed a few signs?
There may have been genuine improvements, but the administrative closure of those actions tells us very little about whether the original exposure has been adequately controlled.
This is one of the fundamental weaknesses in how organisations manage investigation recommendations.
We confuse evidence that an action has been completed with evidence that the action has been effective.
They're two very different things.
Recommendations Are Management Decisions, Not Investigation Team Decisions
One of the principles I regularly reinforce when discussing ICAM investigations is that investigators investigate. They examine the evidence, identify the contributing factors and absent or failed defences, and develop recommendations to address the systemic weaknesses identified.
But investigators don't necessarily have the authority to implement those recommendations.
They may recommend redesigning equipment, changing operating arrangements, modifying maintenance systems, increasing resources or addressing weaknesses in organisational governance. These decisions often sit well above the authority of the investigation team.
And that's entirely appropriate.
The problem arises when the organisation fails to establish clear accountability for what happens next.
A recommendation that requires a significant capital investment cannot simply be assigned to a maintenance supervisor and forgotten. A recommendation concerning organisational resourcing cannot reasonably be left with somebody who has no authority over budgets or staffing.
Accountability needs to sit with the people who have the authority to make the necessary decisions.
That means senior management, and in some cases executive leadership or the board, must take ownership of the response to significant investigation findings
Not simply receive the report.
Not just acknowledge the findings.
Actually take responsibility for ensuring the organisation makes an informed decision about what needs to change, and then follows through.
Accepting a Recommendation Is Not the Same as Implementing It
There's another issue that deserves attention.
Organisations should not be expected to blindly accept every recommendation produced by an investigation team.
Investigators can make mistakes. Recommendations can be impractical, disproportionate, insufficiently developed or occasionally fail to address the underlying contributing factors.
Management has every right, and indeed a responsibility, to critically evaluate recommendations before committing resources.
But there is an important distinction between a recommendation being formally considered and rejected for defensible reasons, and a recommendation simply disappearing into an administrative black hole.
Where management decides not to implement a recommendation, that decision should be documented, supported by a clear rationale and made by somebody with the appropriate authority.
If an alternative control is proposed, there should be evidence that it adequately addresses the risk identified.
And where a significant exposure remains, the organisation should be explicit about who has accepted that residual risk and on what basis.
Silence is not a decision. Inaction is not a control measure. And allowing a recommendation to sit unanswered for months or years is not an acceptable substitute for governance.
Six Things Every Organisation Should Be Able to Demonstrate
If organisations genuinely want to learn from incidents, recommendation governance needs to extend well beyond maintaining a corrective action register.
At a minimum, I would expect six things.
1. A named accountable executive. Significant recommendations need an accountable person with sufficient authority to make decisions, allocate resources and remove barriers to implementation. Tasks can be delegated, but accountability for the outcome should remain clear.
2. Formal acceptance or justified rejection. Every recommendation should receive a documented management decision. If accepted, the organisation should identify how it will be addressed. If rejected or modified, the rationale, decision-maker and treatment of any remaining exposure should be recorded.
3. Due dates and meaningful escalation. Agreed actions need realistic timeframes, with overdue items escalated according to their significance. Where implementation will take time, appropriate interim controls should be considered. Repeatedly extending a due date without reviewing the associated exposure is not effective governance.
4. Evidence of implementation. Closing an action should require objective evidence that the agreed change has occurred. Depending on the recommendation, that might include physical modifications, commissioning records, revised work arrangements or demonstrated changes in operational practice. Simply stating that an action has been completed should not be sufficient.
5. Independent verification of effectiveness. Someone sufficiently independent of the implementation should assess whether the change is actually working. That may require workplace observations, testing, worker consultation, performance data or a follow-up review after the control has been operating for a reasonable period. For significant risks, verification should be proportionate to the potential consequences of failure.
6. Consideration of exposure elsewhere. If an investigation identifies a systemic weakness at one facility, why would we assume that weakness exists nowhere else? Organisations operating across multiple sites should formally consider whether the same conditions, equipment, procedures, management systems or organisational decisions are present elsewhere.
That final point is particularly important.
An investigation into an incident at one location may identify a weakness that exists across an entire organisation. Correcting the issue at the incident site while leaving identical exposures elsewhere is a missed opportunity for organisational learning.
And importantly, none of these six expectations requires a particularly sophisticated software system.
They require leadership, accountability, disciplined decision-making and a willingness to verify that the organisation has actually changed.
Where Is the Board in All of This?
This is where the conversation needs to move beyond the safety department.
For significant incidents, particularly those involving fatalities, serious injuries or high-potential events, recommendation implementation should be a matter of executive and board-level interest.
Boards regularly receive reports showing investigation completion rates, outstanding corrective actions and percentages of actions closed within agreed timeframes.
Those measures have administrative value, but they can also create a misleading impression of assurance.
A report showing 98% of corrective actions closed on time might look impressive. But what if the remaining 2% represents the organisation's most significant unresolved exposures?
And what if half the completed actions have never been checked to determine whether they actually work?
The board might be receiving an excellent account of administrative performance while knowing remarkably little about whether the organisation has become safer.
A more useful governance conversation would explore which significant risks have been reduced, which critical recommendations remain unresolved, whether implemented controls have been independently verified, and whether similar exposures exist elsewhere in the business.
It would also examine the reasons recommendations are delayed. Is it a lack of resources? Competing production priorities? Technical complexity? Unclear accountability? Or an unwillingness to confront uncomfortable organisational weaknesses?
Those questions tell us considerably more about safety governance than a green dashboard.
Learning Doesn't Happen When the Report Is Published
We often talk about investigations as opportunities for organisational learning.
I agree with that sentiment, but I think we need to be much more precise about what learning actually means.
Identifying a contributing factor is not, by itself, learning.
Writing a recommendation is not learning.
Distributing a report or presenting findings at a toolbox meeting may improve awareness, but it does not necessarily demonstrate that the organisation has learned anything in a meaningful operational sense.
Learning becomes evident when the organisation uses what it has discovered to change how risks are understood, decisions are made, work is designed and controls are managed.
And importantly, when it checks whether those changes have produced the intended result.
This doesn't mean every recommendation will eliminate the possibility of a similar incident. Risk management doesn't work that way. But it should mean the organisation can demonstrate a reasoned and evidence-based response to the weaknesses identified.
Without that response, we risk turning incident investigations into little more than elaborate exercises in documenting what has already gone wrong.
The Real Measure of a Successful Investigation
The Yenkin-Majestic case is a particularly confronting example because of the seriousness of the original incident and the CSB's subsequent public statement.
But the broader lesson is relevant to every organisation, regardless of industry or size.
We spend a great deal of time discussing investigation methodologies, investigator competencies, evidence quality and the identification of systemic contributing factors. All of those things matter, and I certainly wouldn't suggest otherwise.
However, even the most technically proficient investigation has limited preventive value if its recommendations are ignored, poorly implemented or never tested for effectiveness.
Perhaps we need to reconsider how we define investigation success.
Rather than simply asking whether the investigation was completed on time, whether the report was approved and whether the corrective actions have been closed, we should be asking a more important question
.
What can the organisation now demonstrate has changed as a consequence of this investigation, and what evidence do we have that those changes have reduced the relevant exposure?
That's a considerably more demanding question.
It requires more than an investigation team, a report and an action register. It requires a management system that connects investigation findings to accountable decisions, implementation, verification and organisational learning.
And it requires leaders who are prepared to remain accountable long after the investigators have packed up and gone home.
Because ultimately, “report completed” and “action closed” are not measures of risk reduction.
They're measures of administrative progress.
And we really need to stop confusing the two.
Source: US Chemical Safety and Hazard Investigation Board, 19 October 2026 — Yenkin-Majestic/OPC Polymers Fails to Implement CSB Safety Recommendations.
