Target Zero and the Disappearing Near-Miss Cards
A 320-person food packaging plant launched a 'Target Zero' program in January: any work group that finishes a quarter without a recordable injury receives a $250 bonus per employee. In the same quarter, the new operations manager began personally interviewing every employee who filed a near-miss card and asking each to explain in writing what they had done wrong. Near-miss cards fell from an average of 14 per month the prior year to 3 per month. Recordables are flat at about 2 per month, but severity is rising - two of the last four involved amputations that coworkers say had been reported informally to shift leads as 'accidents waiting to happen.' First aid visits are up 40%. At the quarterly review the plant manager cites the falling near-miss number as proof the program is working.
1. Interpret the data. What is the most likely explanation for a 79% drop in near-miss reports accompanied by rising severity, and how would you confirm it before acting?
Read the worked answer
What the numbers say. Near-miss reports are a leading indicator only while reporting is voluntary and low-blame. A drop that coincides in time with (a) a financial reward for not having injuries and (b) a punitive interview process is far more likely to be suppressed reporting than improved conditions. The supporting signals point the same way:
- Recordables flat while severity rises - the hazards are not being removed; only the small ones are disappearing from the system.
- Amputations that 'everyone knew about' - warnings existed but travelled through informal channels (shift leads) that never reached the corrective-action system.
- First aid up 40% - minor events are being absorbed at the first aid room rather than generating reports, which is what you expect when reports have consequences.
How to confirm before you act.
- Tabulate informal reports captured by shift leads and compare to the formal card count.
- Read the last four recordables and search for prior mentions of the same equipment or task.
- Ask a sample of hourly employees, without the supervisor present, whether they would file a card; a single question is enough.
- Check whether anyone was disciplined or had a bonus withheld following a report.
- Reconcile the first aid log against the 300 log, and confirm the amputations were evaluated against the 24-hour reporting requirement for amputation and in-patient hospitalization - if they were not reported, that is a compliance exposure on top of a culture problem.
The temptation is to run a poster campaign to 'get reports back up.' That treats the symptom and tells employees their read of the situation was wrong.
2. What do you recommend doing about the bonus and the manager's interviews? Give the specific changes and how you would announce them.
Read the worked answer
Separate learning from accountability. Write a one-page just-culture standard with a decision rule employees can apply to themselves: honest error and at-risk behavior get corrected through the system; reckless behavior, intentional violation, and anything involving harassment or discrimination go to HR through a defined route. The current practice collapses all of it into one interview, so the only rational employee response is silence.
Restructure or retire the bonus. A payout conditioned on the absence of recordables pays people for the absence of reports. If a recognition program is kept, tie it to process measures employees control - hazards closed on time, corrective actions verified, safety observations completed, participation in JSAs - and never let a work group's payout depend on whether a coworker got hurt.
Replace the interview. The near-miss conversation becomes 'walk me through what made this possible,' led by the area supervisor with the employee, focused on conditions. The written statement is dropped. Findings get logged and tracked.
Announce it in a way that is believable. A written statement from the plant manager, repeated face-to-face on every shift by supervisors, saying plainly: no one will be disciplined for reporting a hazard or an injury, and here is what happens when you do. Then prove it within 30 days by publicly closing two or three issues that came in through the new channel. Employees believe behavior, not memos - the first test case after the announcement will determine whether the change is real.
3. Define the leading indicators and targets you would use to judge whether the change worked in six months, and explain why raw near-miss count is a poor target.
Read the worked answer
Why raw count fails as a target. It is easily gamed in both directions - capped by fear or inflated by junk submissions - and it depends on headcount and shift pattern. Setting a number also invites the wrong question ('are we hitting 14?') instead of the right one ('are hazards closing?'). Expect the count to rise in the first quarter after the culture change; if leadership reads that as regression, the whole effort collapses.
Better measures, with six-month targets.
| Indicator | Why it matters | Target direction |
|---|---|---|
| Reports per 100 employees per month, split by source (hourly vs. supervisor) | Volume plus who is willing to speak | Rising, with hourly share above 60% |
| % of reports receiving feedback to the reporter within 7 days | The single strongest predictor of continued reporting | Above 90% |
| Median days from report to verified hazard closure | Tests whether reporting produces change | Down by half |
| Repeat-hazard rate (same equipment/task reported twice) | Tests whether fixes hold | Falling |
| First aid case rate | Early signal of absorbed events | Falling toward the pre-program baseline |
| Anonymous survey item: 'I would report a near miss without fear of consequences' | Direct measure of the mechanism | Up meaningfully from baseline |
Pair volume with quality. Require that a report be counted as 'closed' only when a corrective action was implemented and verified - otherwise the program learns to generate paperwork instead of removing hazards. Review the whole set monthly with the plant manager, and treat the survey result, not the card count, as the headline number.