Safety

Most incidents are not random. They happen at predictable moments, after a change, after a break, when something new is introduced. The Safety pillar builds the capability to see those moments coming and act before the incident occurs.


A site I worked with had a persistent problem with lost time incidents in the days immediately after a planned shutdown. The data was clear. The pattern was consistent. Every shutdown, the same window, the same elevated risk.

The team had tried a number of things. Additional briefings before the restart. Supervisor walk-rounds on the first morning back. Reminders about safe working procedures. None of it moved the number significantly.

Then someone asked a different question. Instead of asking how to make the first day back safer, they asked whether the first day back needed to be a production day at all.

It did not. There was training that needed to happen. Regulatory checks that needed to be completed. Equipment that needed to be verified before it ran. All of it had been squeezed into the margins of a production day. It could just as easily fill a non-production day.

So that is what they did. The first day back after each shutdown became a structured non-production day. Training, verification, equipment checks, team briefings. The things that needed to happen anyway, given the time they needed.

The lost time incidents in that window dropped dramatically. Not because the hazard was removed. Because the people were not near it at the moment of highest risk.

That is what a proactive safety system looks like. Not more warnings. Not more rules. A clear picture of when and where the risk is highest, and a deliberate decision about what to do about it.


Pillar Intent

The Safety pillar exists to build the capability to identify hazards before they cause harm, reduce the risk they present through the most effective controls available, and investigate incidents and near misses in a way that changes something.

That is not a compliance function. Compliance sets the minimum. The Safety pillar works above the minimum, on the conditions, behaviours, and systems that determine whether the minimum is ever tested.

The starting point is always the hazard. Not the incident. An incident is what happens when a hazard was not identified, or was identified and not controlled, or was controlled and the control failed. By the time the incident occurs, the opportunity to prevent it has passed. The Safety pillar works upstream, on the hazards, the controls, and the moments of highest risk.

This pillar works in close partnership with Autonomous Maintenance. The safety map, the documented picture of every significant hazard in the work area, including isolation points, energy sources, and restricted access zones, is a prerequisite for AM Step 1. No team gets close to equipment without it. As AM progresses and the team's understanding of their equipment deepens, their understanding of its hazards deepens with it. The two develop together.

The Safety pillar also works alongside Problem Solving. Every incident and every near miss is investigated using the same six-stage problem solving discipline used across the site. The finding must change something, a condition, a design, a standard, a control. An investigation that closes with "the operator did not follow the procedure" has not found the cause. It has found the last person in the chain.


Pillar Owned Losses

Safety losses are not always counted in the same way as production or quality losses. Some are visible and recorded. Others are absorbed into the way work is done, accepted as normal, until they are not.

⚠️
Lost time incidents. The lagging indicator. An incident has occurred, someone has been hurt, and time has been lost. The incident is recorded, investigated to some degree, and a corrective action is raised. The corrective action closes the record. The condition that produced the incident may or may not have changed. Lost time incidents are the measure of a safety system that is behind the risk.
⚠️
Near misses that were not reported. A near miss is an incident that did not cause harm, this time. It is the clearest signal the safety system has that a hazard is active and a control is not working. A site with very few reported near misses and occasional lost time incidents almost certainly has an under-reporting problem, not a safe workplace. Near misses that are not reported are near misses that will not be investigated. The next event may not be a near miss.
⚠️
Hazards that are known but uncontrolled. Every site has a list of known hazards that have not yet been addressed. Some are waiting for a capital decision. Some are waiting for a maintenance window. Some have been on the list long enough that they have become part of the furniture. A known, uncontrolled hazard is not a risk that has been accepted. It is a risk that has been deferred. The difference matters when the incident occurs.
⚠️
Controls that exist on paper but not in practice. The procedure says isolate before entry. The permit system requires sign-off. The guard must be in place before the machine starts. These controls are real when they are followed. When they are not followed, because the process is slow, because the pressure is high, because it has always been done this way, they are not controls. They are documentation.
⚠️
Incidents that cluster around predictable moments. The first day back after a shutdown. The first week for a new team member. The first run of a new product. The first shift after a process change. These are not random. They are predictable. A safety system that does not recognise the pattern will keep responding to the same events in the same windows, without changing the conditions that produce them.
⚠️
The cost of managing risk rather than reducing it. PPE added to compensate for a hazard that could have been engineered out. A warning sign placed in front of a condition that could have been redesigned. A procedure written to manage a risk that a physical control would have eliminated. These are not wrong. Sometimes they are the only option. But when they are the first response rather than the last, the hazard remains, and the control depends on human behaviour being perfect every time.

Pillar Owned Processes

The Safety pillar owns three processes. They are introduced in sequence. Hazard identification and ranking tells you where the risk is. Risk reduction tells you what to do about it. Incident and near miss investigation tells you what the system missed when something goes wrong.

Hazard Identification and Ranking. The starting point. A structured assessment of every significant hazard in the work area, energy sources, isolation points, chemical risks, mechanical risks, working at height, manual handling, and any other condition that could cause harm. Each hazard is assessed by likelihood and consequence and classified into tiers that drive the priority and type of control. The output is the safety map: a documented, visual picture of the hazards in the area, the controls in place, and the residual risk. The safety map is a prerequisite for AM Step 1. It is reviewed whenever the work area, the equipment, or the process changes.

Hierarchy of Controls. For each ranked hazard, the question is: what is the most effective control available? The hierarchy runs from most effective to least — eliminate the hazard entirely, substitute it with something less dangerous, engineer a physical barrier, reduce exposure, add warnings and procedures, and as a last resort, add PPE. The most effective controls work regardless of what anyone decides to do. The least effective depend on human behaviour being correct every time.

Incident and Near Miss Investigation. Every lost time incident is investigated. Every near miss is investigated with the same rigour. The investigation uses the six-stage problem solving process, Define, Return to Standard, Understand Why, Select the Solution, Validate the Fix, Make it Stick, applied to the conditions that produced the event. The finding must change something. A procedure that was not followed is not a finding. It is a symptom. The finding is the reason the procedure was not followed, the pressure, the design, the standard that made non-compliance the easier path.


When incidents are most likely

Incidents are not evenly distributed across time. They cluster. Understanding when the risk is highest is one of the most practical things a safety system can do, because it turns a reactive response into a proactive one.

These are the moments that consistently produce elevated risk. They are not excuses. They are signals. A system that recognises them can act before the incident, not after.

After a shutdown or extended break. Routines are disrupted. Equipment may be in a different state than when it was last running. People are mentally returning to the work, not yet fully in it. The physical environment may have changed during the shutdown. This is one of the most consistently high-risk windows in any operation that runs planned shutdowns. The response is not more warnings. It is a deliberate decision about what work happens in that window and what does not.

After any change. New equipment. New process. New layout. New product. New procedure. Change introduces unfamiliarity, and unfamiliarity is where incidents live. The hazard assessment for the area needs to be reviewed before the change goes live, not after the first incident reveals what was missed. A change that has not been assessed for new or altered hazards is a change that has been introduced without a safety review.

When new people are present. New to the site. New to the role. New to the team. New people do not yet know what the experienced team knows, which equipment behaves unexpectedly, which area has a hazard that is not obvious, which procedure is critical and which is routine. The induction covers the formal requirements. It does not cover the knowledge that the experienced team carries without thinking about it. That knowledge needs to be made explicit and transferred deliberately.

When production pressure is high. The schedule is tight. The batch is late. The customer is waiting. In these moments, the shortcuts that are always available become more tempting. The isolation that takes five minutes feels like five minutes the line cannot afford. The guard that slows the changeover gets left off for one run. The pressure does not create new hazards. It reduces the likelihood that existing controls will be followed. A safety system that only works when the pressure is low is not a safety system.

After a near miss that was not investigated. A near miss is a signal that a hazard is active and a control is not working. When that signal is not investigated, when it is recorded and filed, or not recorded at all, the hazard remains. The next event in the same location, with the same hazard, may not be a near miss. The investigation is not optional. It is the mechanism by which the system learns.


Pillar Behaviours

A near miss is reported on the line.

Without the system: The near miss is noted. The supervisor asks what happened. The team member explains. The supervisor says they will look into it. The near miss is not formally recorded. Nothing changes. Three months later, a lost time incident occurs in the same location, with the same hazard.
With the system: The near miss is recorded immediately. It is reviewed at the Tier 1 meeting the same day. The hazard is identified and the control is assessed, was the control in place? Did it fail? Was it absent? An investigation is opened with a named owner and a close-out date. The finding changes something before the shift ends, or a temporary control is put in place until the permanent fix is confirmed. The near miss rate is tracked as a leading indicator. A rising near miss rate is treated as a signal, not a problem.

A new team member joins the line.

Without the system: The new team member completes the site induction. They are assigned to work alongside an experienced operator for the first week. The experienced operator shows them the job. The hazards that are obvious to the experienced operator are not explained, because they are obvious. The new team member does not know what they do not know. The first incident involving a new team member is treated as a training failure.
With the system: The new team member's induction includes a walk of the safety map for their area. Every significant hazard is pointed out, explained, and the control is demonstrated. The experienced operator does not assume knowledge. The new team member is not left unsupervised near high-ranked hazards until their capability has been confirmed on the capability matrix. The first week is treated as the highest-risk window and managed accordingly.

A process change is being planned.

Without the system: The change is designed, approved, and implemented. The safety team is notified after the change is live. A near miss occurs two weeks later involving a hazard that the change introduced. The hazard assessment is updated. A new control is added. The incident is recorded as a consequence of the change.
With the system: The hazard identification and ranking process is reviewed before the change goes live. The safety map is updated to reflect the new or altered hazards. New controls are designed into the change, not added afterwards. The hierarchy of controls is applied — can the new hazard be designed out? If not, what is the highest-order control available? The change does not go live until the safety review is complete and the controls are confirmed.

The planned shutdown is approaching.

Without the system: The shutdown plan covers scope, contractors, and schedule. The restart plan covers production targets and batch priorities. The first day back is a production day. The team returns, the line starts, and the elevated-risk window begins. The incidents that occur are treated as individual events.
With the system: The data on incident patterns around shutdowns has been reviewed. The first day back is structured deliberately, the work that needs to happen anyway, training, equipment verification, regulatory checks, is scheduled for that day. Production begins on day two, when routines are re-established and the equipment has been confirmed in the correct state. The elevated-risk window is managed by changing what happens in it, not by adding warnings to a day that was already high-risk.

A hazard is identified during an AM Step 1 .

Without the system: The team finds a damaged guard. They tag it and continue cleaning. The tag is recorded. The guard is repaired two weeks later. In the meantime, the equipment runs without the guard in place, because the line cannot afford to stop.
With the system: The damaged guard is tagged and the equipment is isolated immediately. The hazard is assessed, what is the risk of running without the guard? The answer determines the response. If the risk is significant, the line does not run until the guard is repaired or a temporary control is in place. The repair is escalated through the TAM with a named owner and a same-day close-out date. The incident that would have occurred in the two-week window does not occur.

What Good Looks Like

Safety pillar team member

  • Maintains the safety map for each area. Reviews it whenever the equipment, process, or work area changes. Does not wait for an incident to trigger a review.
  • Leads hazard identification and ranking assessments with the operating team and the Maintenance function. Does not complete the assessment in isolation and hand it over. Builds it with the people who work in the area.
  • Applies the hierarchy of controls to every ranked hazard. Starts at the top, can this hazard be designed out? and works down. Does not accept PPE as the primary control for a hazard that could be engineered out.
  • Investigates every near miss with the same rigour as a lost time incident. The investigation asks: what condition produced this event, and what needs to change so it does not produce the next one?
  • Tracks near miss reporting rate as the primary leading indicator. A low near miss rate on a site with active hazards is not a sign of a safe workplace. It is a sign of an under-reporting problem.
  • Is present at AM step preparation for each line. Ensures the safety map is complete and the team understands the hazards in their area before AM Step 1 begins.
  • Knows the high-risk windows for the site, shutdowns, changes, new people, and has a plan for each one that goes beyond additional briefings.

Frontline leader

  • Reviews the safety map for their area and can explain every significant hazard and the control in place for it. Does not treat the safety map as a document that lives in a folder.
  • When a near miss occurs, records it immediately and raises it at the Tier 1 meeting the same day. Does not decide whether it is worth reporting. Reports it and lets the investigation decide.
  • Confirms safety-critical controls through the Process Confirmation routine, isolation points, guards, permit requirements. Does not assume controls are in place because they are on the procedure.
  • Manages the high-risk windows deliberately. Knows when new people are on the line, when a change has been introduced, when the team is returning from a break. Adjusts supervision and confirmation accordingly.
  • Does not allow production pressure to displace safety controls. When the schedule is tight and a shortcut is available, names it explicitly and refuses it. The team learns what the standard is by watching what the leader does when it is inconvenient.

Operations leader

  • Ensures near miss investigations are completed, not just opened. A named owner and a due date are not enough. The investigation needs a finding that has changed something.
  • Reviews the near miss rate and the lost time incident rate together. Asks: what is the ratio telling us? A site where near misses are rising and lost time incidents are falling is a site where the reporting system is working and the investigation system is working. A site where near misses are low and lost time incidents are occasional is a site where the signal is not being captured.
  • Supports the Safety pillar team's access to the line for hazard assessments and investigations. A hazard assessment that cannot get to the equipment is an assessment that will not find the hazard.
  • Ensures the high-risk windows are managed in the production plan, shutdowns, changes, new team members. Does not treat these as safety team responsibilities alone. They are operational decisions.

Site or function leader

  • Sets the expectation that near misses are reported and investigated, not managed. A near miss that is recorded and filed is not a near miss that has been addressed.
  • Asks in reviews: what are our top three near miss types this month, and what have the investigations found? Not: what is our lost time incident rate?
  • Ensures the hierarchy of controls is being applied. When a new PPE requirement is introduced, asks: what higher-order control was considered and why was it not used? PPE as a first response is a signal that the hazard assessment did not go far enough.
  • Connects safety performance to the site's strategic objectives. A lost time incident rate that is not improving is a strategic problem, not just a safety metric.
  • Treats the first day back after a shutdown as a managed event, not a normal production day. Reviews the incident data. Makes a deliberate decision about what work happens in that window.

Audits and Confirmations

✅
Is the safety map current for every work area? A safety map that has not been reviewed since the last equipment change is not current. The map should be reviewed whenever the equipment, process, or work area changes, and at least annually. If the team cannot point to the map and explain every hazard on it, the map is not being used.
✅
Does every significant hazard have a control at the highest level the hierarchy allows? A hazard with PPE as its primary control, where an engineering control was available, has not been adequately assessed. Review the top-ranked hazards and ask: was the hierarchy applied? Is there a higher-order control that was not considered?
✅
Is the near miss reporting rate meaningful? A site with active hazards and very few reported near misses has an under-reporting problem. The near miss rate should be tracked and reviewed. A rising near miss rate, combined with a falling lost time incident rate, is the sign of a system that is working, more signals being captured, more investigations being completed, fewer incidents occurring.
✅
Does every near miss have a completed investigation? A near miss that is recorded but not investigated is a near miss that has been accepted. Every near miss should have a named investigation owner, a finding, and a countermeasure that has changed something. Check the backlog. If near misses are accumulating without findings, the investigation system is not keeping up.
✅
Are safety-critical controls confirmed through Process Confirmation? A guard that is assumed to be in place is not a control. An isolation point that is assumed to be followed is not a control. Safety-critical controls should be on the frontline leader's Process Confirmation schedule and confirmed on a defined frequency. If they are not, they are being assumed, not confirmed.
✅
Are the high-risk windows being managed deliberately? Shutdowns, changes, new people, high production pressure, these are the moments when incidents are most likely. Ask: what is the plan for each of these windows? If the answer is "additional briefings," the plan is not sufficient. The plan should address the conditions that make the window high-risk, not add warnings to a situation that is already elevated.
✅
Is the Safety pillar team present at AM step preparation? The safety map must be complete and the team must understand the hazards in their area before AM Step 1 begins. If the Safety pillar team is not part of AM step preparation, the safety prerequisites for the step have not been confirmed.
✅
Are incident investigations finding conditions, not people? An investigation that closes with "operator error" or "failure to follow procedure" has not found the cause. It has found the last person in the chain. Review the last five investigation findings. If most of them name a person rather than a condition, the investigation process needs to go deeper.

Where to Start

Before selecting a process to build, take stock of where you are. These questions are a quick read of the safety system as it stands today.

🔍
Do you have a current safety map for every work area?

If not, start with Hazard Identification and Ranking. You cannot apply controls to hazards you have not identified. The safety map is the prerequisite for everything else in this pillar, and for AM Step 1 on any line.

🔍
Are your controls at the highest level the hierarchy allows?

If your primary controls are procedural or PPE-based for hazards that could be engineered out, the hierarchy has not been applied. Review the top-ranked hazards and work through the hierarchy from the top. Design it out first. Engineer a barrier second. Reduce exposure third. Add warnings fourth. Write a procedure fifth. Add PPE last.

🔍
Are near misses being reported and investigated?

If near miss reporting is low, or if near misses are being recorded but not investigated, the leading indicator system is not working. A site that only counts lost time incidents is always one step behind the risk. Start by building the near miss reporting and investigation discipline before the next lost time incident makes it urgent.

🔍
Do you know your high-risk windows?

If you have not analysed when and where your incidents cluster, you are managing safety reactively. Pull the incident and near miss data. Look for patterns, time of day, day of week, after shutdowns, after changes, in the first weeks of new team members. The pattern tells you where to focus the proactive effort.

🔍
Are safety-critical controls confirmed through Process Confirmation?

If safety controls are assumed rather than confirmed, they will drift. Add safety-critical controls to the Process Confirmation schedule before the next audit reveals they are not being followed.

If the answer to most of these is no, start with Phase 1, Establish. Complete the safety map for the pilot area. Apply the hierarchy of controls to the top-ranked hazards. Build the near miss reporting and investigation discipline. These are the foundations. Without them, the safety system is reactive by design.

If the safety map exists and near misses are being investigated, but incidents are still clustering in predictable windows, start with Phase 2, Strengthen. The most common gap at this stage is not in the tools. It is in the proactive management of high-risk moments. Identify the windows. Build a plan for each one that goes beyond additional briefings.

If near miss investigations are completing with findings that change conditions, and high-risk windows are being managed deliberately, move to Phase 3, Sustain. Extend the safety map and the hierarchy of controls to all areas. Build the capability across the full site so that hazard identification and near miss investigation are not dependent on the Safety pillar team to lead them.


Phases

The Safety pillar is introduced in sequence. Each phase builds the capability and the infrastructure that the next one depends on.

Phase 1 — Establish

The safety map is completed for the pilot area. Every significant hazard is identified, assessed, and classified. The hierarchy of controls is applied to the top-ranked hazards. Controls that are procedural or PPE-based, where a higher-order control is available, are reviewed and upgraded where possible.

The near miss reporting and investigation system is established. The team understands what a near miss is, why it is reported, and what happens when it is. The first near miss investigations are completed with findings that change something. The near miss rate is tracked as a leading indicator alongside the lost time incident rate.

The safety map is integrated into AM Step 1 preparation for the pilot line. The team understands the hazards in their area, the isolation points, the energy sources, the restricted access zones, before the first cleaning session begins.

The measure of Phase 1 is not the lost time incident rate. It is whether the foundations are in place: a current safety map, controls at the highest level the hierarchy allows for the top-ranked hazards, and a near miss investigation system that is producing findings that change conditions.

Phase 2 — Strengthen

The safety map expands to cover all areas in the pilot zone. Hazard identification and ranking is run by the Safety pillar team with the operating team and the Maintenance function as active participants, not completed in isolation and handed over.

Near miss investigations are running consistently. The findings are reviewed at the monthly safety review. Every top near miss type has either a completed investigation with a confirmed fix, or an open investigation with a named owner and a due date. The near miss rate is rising, because reporting is improving, and the lost time incident rate is falling.

The high-risk windows are being managed deliberately. The data on incident patterns has been reviewed. The plan for shutdowns, changes, and new team members goes beyond additional briefings. For each high-risk window, the question has been asked: what is the highest-order control available?

The measure of Phase 2 is whether the near miss investigation system is producing findings that change conditions, and whether the high-risk windows are being managed with a plan that addresses the conditions, not just the symptoms.

Phase 3 — Sustain

The Safety pillar's capability reaches all areas across the site. Hazard identification and ranking is a standard capability, area leaders and process owners can run an assessment without the Safety pillar team leading it. Near miss investigation is embedded, the team closest to the work can define the problem, examine the conditions, and find the cause without waiting for the Safety pillar team.

The hierarchy of controls is applied consistently. PPE is the last control considered, not the first. Every new change, new piece of equipment, and new process is assessed for hazards before it goes live, not after the first incident.

The measure of Phase 3 is not the absence of incidents. It is the speed and quality of the response when a near miss occurs, and the consistency with which the hierarchy of controls is applied to new hazards. A site that has reached Phase 3 does not manage safety reactively. It identifies hazards, applies the most effective controls available, and investigates every signal the system produces, consistently, quickly, and without needing to be directed.

The number of people who hold the capability grows across the phases. Phase 1 builds it in the Safety pillar team and the AM pilot team. Phase 2 extends it to the leaders and process owners in the pilot area. Phase 3 reaches all process owners and people leaders across the site.

The depth of capability also grows. In Phase 1, the goal is that the Safety pillar team can complete a hazard assessment and run a near miss investigation. In Phase 2, the goal is that operations leaders and process owners can participate in and lead assessments and investigations. In Phase 3, the goal is that safety capability is embedded, the team closest to the work can identify a hazard, apply the hierarchy of controls, and investigate a near miss without waiting for the Safety pillar team to lead it.

Sample Phase Implementation Plan

Ready to build your pillar?

Before you start, make sure your team is in place, your charter is drafted, and your plan connects to the site's strategic objectives. The Setting Up Your Pillar guide takes you through each step in sequence.