Repeat-Incident Case: How 250+ Projects Fixed CAPA

Repeat incidents rarely repeat because the organization lacks a form. They repeat because the first investigation closes the visible event while leaving the work system almost untouched. A broken guard is replaced, a worker is retrained, a toolbox talk is delivered, and the dashboard shows closure. The same exposure then returns through a different shift, contractor crew, product mix, or maintenance window.

Across more than 250 safety projects supported by Andreza Araujo and ACS Global Ventures, the pattern is consistent enough to treat as a case in itself. Organizations that cut recurrence did not become better at writing corrective actions. They became stricter about proving that the action changed the condition that produced the event.

The thesis of this case is narrow and practical. Corrective and preventive action, often called CAPA, stops repeat incidents only when leaders treat closure as a field-verification decision, not as an administrative deadline. That distinction changes who owns the action, what evidence counts, and when the investigation is truly finished.

Key takeaways

• Treat repeat incidents as evidence that the first corrective action did not reach the operating condition.
• Separate document closure from field verification before declaring CAPA effective.
• Assign action ownership to the leader who controls the risk, not only to the person who manages the investigation file.
• Use recurrence reviews to test decision quality, supervision rhythm, and control reliability.
• Connect corrective actions to SIF precursor review when the repeated exposure can lead to serious harm.

Initial scenario

The typical starting point was not a catastrophic failure. It was a familiar recurrence that the organization had normalized. A hand injury returned after a glove campaign. A line-break exposure returned after another permit refresher. A forklift near miss returned after the warehouse supervisor signed a traffic checklist. None of these responses was absurd, but each one stayed too close to the last visible act.

In many sites, the investigation file looked disciplined. The root-cause field was completed, the action had an owner, the due date was visible, and the status changed from open to closed. The weak point sat elsewhere. The evidence used for closure did not show that the operating condition had changed. It showed that the assigned activity had been performed.

This is where Andreza's critique in The Illusion of Compliance becomes operational. A system can satisfy its own rules while failing to control the risk those rules were supposed to address. In incident investigation, that illusion appears when a closed action becomes a substitute for a changed task.

Decision

The decision in these projects was to stop treating CAPA as a task list. The action plan had to become a control-change plan whose evidence could be tested in the field. That meant some familiar action types lost status. Training, communication, awareness campaigns, and revised procedures could still be valid, but only when paired with evidence that the work changed after the intervention.

For an EHS manager, this decision creates tension because it slows closure. A fast action that reaches the database is easier to defend than a slower action whose effectiveness depends on operations, maintenance, procurement, engineering, or contractor management. Yet the sites that reduced recurrence accepted that tension because the old speed had been false economy.

The first rule was simple in practice, although difficult politically. If the action did not modify a barrier, a decision right, a physical condition, a supervision routine, a maintenance trigger, or a work-planning rule, it could not be the main corrective action for a recurring exposure. It could support the change, but it could not carry the case alone.

Execution

The execution model began with a recurrence map. Teams grouped recent incidents, near misses, and audit findings by exposure rather than by department. A maintenance hand injury, a contractor hand injury, and a packing-line hand injury might sit in different files, but the map asked whether all three depended on the same weak control, such as poor line-of-fire separation, weak tool selection, or supervision that accepted improvised work.

The second move was to rewrite the action in observable language. Instead of "retrain operators on safe line breaking," the action became "change the line-break permit so the supervisor verifies isolation points in the field before flange opening, then sample five permits across two shifts within 14 days." The revised action named the condition, the control, the verifier, and the evidence.

The third move was delayed effectiveness testing. A corrective action verified on the same day it is completed often measures enthusiasm, not stability. Better sites returned after the work had been exposed to ordinary pressure, including night shift, contractors, overtime, production recovery, rain, heat, equipment downtime, or short staffing. The article on rebuilding corrective actions after recurrence expands that short-cycle repair model for teams that need a procedural route.

The final move was leadership review. Senior leaders did not ask whether the action was closed. They asked what changed in the field, which evidence proved it, and whether the same exposure could still appear through another path. That question moved recurrence from EHS administration into operational control.

Measured result

The measured result across these projects was not a single universal percentage, because the 250+ projects covered different sectors, countries, maturity levels, and risk profiles. The defensible result was a change in the quality of recurrence decisions. Sites began rejecting weak closures, reopening actions whose evidence did not match the exposure, and moving repeat events into leadership review faster.

That matters because repeat incidents are rarely a pure investigation problem. They are a management-system signal. When the same exposure returns, leaders should assume that one of three conditions is present. The first is that the cause was described too narrowly. The second is that the action was too weak for the cause. The third is that the organization closed the action before the field had tested it.

James Reason's work on latent failures helps explain why this shift matters. The last visible act may trigger the event, but repeat events usually reveal upstream conditions whose ownership sits away from the injured person. When CAPA focuses only on the worker, recurrence becomes predictable because the system has left its stronger causes intact.

Generalizable lessons

The first lesson is that recurrence should raise the level of review. If a hazard produces a second similar event, the case should not return to the same closure routine that failed the first time. A repeat event deserves escalation because it shows that the organization misunderstood, underpowered, or under-verified the previous action.

The second lesson is that evidence must match the claim. Attendance proves attendance. A revised procedure proves document change. A purchase order proves procurement. None of those facts proves that the task is now controlled. For that, the organization needs field observation, permit sampling, maintenance records, exposure measurements, engineering checks, supervisor interviews, or other evidence that connects directly to the risk mechanism.

The third lesson is that investigation tools are useful only when the question is strong. A team can use five whys, fishbone, or fault tree and still write a weak action if the scope stops at the worker's decision. The comparison in Fault Tree vs Fishbone vs Five Whys is useful here because tool choice should follow consequence, complexity, and evidence quality.

The fourth lesson is that timelines matter. A late timeline lets memory decay, evidence disappear, and defensive stories harden. A rushed timeline can miss latent conditions. The stronger path is disciplined sequencing, which is why the related guide on an RCA timeline helps teams preserve the first account without pretending that the first account is the full explanation.

Before and after

What to apply in your operation

Start with the last 12 months of repeat events and near misses. Do not begin by debating root-cause labels. Begin by grouping events whose exposure looks similar, even when the department, job title, or immediate trigger differs. This gives leaders a recurrence map that is harder to hide inside isolated files.

For each cluster, ask whether the previous action changed a control or only changed a record. If the answer is only a record, redesign the action around the actual work condition. The best action sentence names the exposure, the control to be changed, the owner who can change it, the field evidence required, and the date when the delayed test will occur.

Then connect high-consequence recurrence to SIF precursor governance. If a repeated exposure could plausibly lead to fatality or permanent disability, it should not wait for a recordable injury to receive senior attention. The related piece on SIF precursor review shows how to separate severe-potential signals from ordinary activity counts.

Finally, protect evidence quality. Photos, witness statements, and equipment logs do not carry the same weight in every case, and the wrong evidence mix can send the team toward a comfortable but false closure. The comparison of photos, witness statements, and equipment logs in RCA can help investigators choose evidence before the conclusion hardens.

Common traps

The first trap is retraining by reflex. Retraining feels responsible because it is visible, fast, and easy to document. It becomes weak when the job design, tooling, permit flow, production pressure, or supervision routine still invites the same exposure.

The second trap is assigning the action to the person closest to the incident. If the corrective action requires engineering, purchasing, scheduling, or staffing authority, a frontline owner cannot make it real. Ownership must sit with the person who controls the condition, not with the person who can update the database.

The third trap is celebrating closure rate. A high closure rate can hide poor action quality when the metric rewards speed more than risk reduction. In Safety Culture: From Theory to Practice, Araujo treats culture as the pattern of decisions that people can observe. A site that rewards fast weak closure teaches everyone what the real priority is.

FAQ

What is a repeat incident in safety?

A repeat incident is an event or near miss that returns with a similar exposure, even when the department, worker, or immediate trigger is different. The repeated pattern suggests that the previous control or corrective action did not reach the real operating condition.

Why do corrective actions fail after an investigation?

Corrective actions fail when they close an activity rather than change a control. Training, reminders, and procedure edits can help, but they do not prove risk reduction unless field evidence shows that the task, barrier, or decision routine changed.

Who should own CAPA effectiveness?

The leader who controls the risk condition should own CAPA effectiveness. EHS can protect the method and challenge evidence quality, but operations, maintenance, engineering, procurement, or contractor management may need to own the actual control change.

How long should effectiveness verification wait?

The verification window depends on the risk and the task cycle. For recurring exposures, verification should occur after the changed control has faced normal operating pressure, such as night shift, contractor work, production recovery, or maintenance conditions.

When should recurrence be escalated to senior leaders?

Recurrence should be escalated when the repeated exposure can lead to serious injury, fatality, permanent disability, regulatory exposure, or loss of trust in the investigation process. At that point, the issue is no longer only a local action plan.

Conclusion

Repeat incidents expose the distance between closure and control. The strongest organizations do not ask only whether the corrective action was completed. They ask whether the field now behaves differently when the same pressure returns.

If your organization needs to rebuild CAPA around field verification, recurrence mapping, and leadership review, Andreza Araujo and ACS Global Ventures can support the diagnostic and operating rhythm. Start a conversation with Andreza Araujo.