Your Last Incident Investigation Probably Didn’t Change Anything

Every story in this issue involves an organization that already had the information it needed. Brent Industries had a fatality and an OSHA finding in 2015. NaturPak had a death in February and a federal inspection already underway. Watsontown Brick had exposure data and a prior citation. None of them lacked warning. All of them lacked a mechanism that converts a warning into a changed condition on the floor. Here is how to test whether yours works.

Pull your last five investigations and audit the corrective actions, not the findings. For each action item, answer three questions: Is it closed? Who verified it? What did they physically look at? An action closed because a supervisor said it was done is not verified. An action closed with a photograph of the guard, a copy of the revised procedure, or a signed retraining roster is.

Check whether any corrective action changed equipment. Investigations drift toward the cheap end of the hierarchy of controls — retrain the operator, add a reminder to the pre-shift, update the JSA. Those are the actions most likely to close on schedule and least likely to prevent recurrence. If your last five investigations produced no engineering change, no procedure rewrite, and no equipment modification, your process is producing paperwork.

Find the practices nobody wrote down. NaturPak’s February incident started with a clogged vent line. Ask maintenance and operations a direct question — what do you do when the equipment doesn’t cooperate? When the valve won’t seat, the line clogs, the lid sticks, the sensor won’t read. The workaround they describe is your unwritten procedure. It has no hazard analysis behind it because it does not officially exist.

Look specifically at repeat-prone standards. Lockout/tagout, fall protection, respiratory protection, and confined space account for a disproportionate share of repeat citations. For each, ask whether you have a written program, whether it matches current equipment, whether training is documented with names and dates, and whether anyone has watched the task performed in the last year.

Close the multi-site loop. Brent Industries’ 2015 finding was in Alabama; the 2026 fatalities were in Ohio. If your organization operates more than one facility, a finding at one site should generate a documented check at every site running the same process. Ask who owns that transfer at your company. If the answer is nobody, that is your finding.

Set a verification date separate from the closure date. Close the action when the fix is installed. Then verify it 90 days later, on the floor, during production. Controls decay. The only way to know whether yours held is to go look after the attention has moved on.

The organizations in this issue did not fail to investigate. They failed to make investigation consequential. That is a different problem, and it is fixed by verification, not by better forms.