Most safety programs do not have a detection problem; they have a decomposition problem. Violations get found at a healthy rate and then decompose quietly in tracking spreadsheets — assigned to someone who changed roles, due dates that slid twice, "in progress" statuses that have meant nothing for a quarter. The failure is architectural: detection, assignment, corrective work, and verification live in different tools owned by different people, so no single system ever holds the whole lifecycle. A closed loop puts them in one place: the violation is reported with type, priority, photos, and forfeit consequences; a corrective program — the sequence of steps that makes it right — is attached to it; the program is assigned to a named person with a tracked closure status; and closure is driven through explicit statuses until the original reporter's evidence confirms it is done. The audit question "show me every open finding and its age" becomes a filter, not a fortnight of spreadsheet archaeology.
The four statuses that keep programs honest #
- Reported — evidence attached, priority set, nothing owed to anyone yet except acknowledgment.
- Assigned — a named owner and a corrective program exist; the clock starts here, not at detection.
- In progress — steps completing against the program, each with its own timestamp and evidence.
- Closed — verification evidence attached by someone other than the fixer, ideally the reporting source.
The separation of fixer and verifier is the single rule that separates theater from closure. When the person who did the corrective work also marks it closed, programs drift toward checkbox resolution — the barrier is re-erected, the photo is taken, and the underlying cause (the missing tool, the skipped permit step, the incentive to rush) survives untouched. Verification by the original reporter or an independent reviewer forces the closure question to be "would this violation have happened today?" rather than "does it look fixed?" Over a year, the ratio of verified closures to total closures is the most honest KPI a safety program can publish, because it is the one number you cannot inflate without doing the work.
Programs, not tasks: the unit that survives personnel changes #
Define corrective programs as reusable correction sequences — a type with steps, certificate implications, and closure criteria — rather than one-off task lists. "Missing PPE" as a program means: immediate stop-work assessment, equipment replacement with photo evidence, toolbox talk within seven days, and certificate status review for the worker involved. When the same violation recurs at another site next month, you assign the program, not your memory of what you did last time. Programs accumulate institutional knowledge in a way task lists cannot, and the violation history they build becomes the dataset your year-end risk review actually needs: which programs run longest, which sites reopen the same finding, which correction types actually reduce recurrence.