When the Behaviour-Based Safety Programme Drives Reporting Underground
A cement bagging plant launches a behaviour-based safety (BBS) programme. Forty observers are trained; each is required to complete four observations a month and each observation is logged as 'safe' or 'unsafe', with most entries reading 'PPE not worn' or 'unsafe act by worker'. The observer tells the worker what he did wrong, and observations are copied to the supervisor's monthly appraisal. After six months the number of observations has fallen from 160 to 42 per month, near-miss reporting has dropped by 60%, and workers now warn each other when observers are coming. Dust masks are issued at the gate in one size only, are uncomfortable in 38 °C heat, and no fit testing has ever been done. Management concludes that the workforce has a poor safety culture and asks for a disciplinary element to be added.
1. Diagnose why the programme is failing. Is management's conclusion correct?
Read the worked answer
Management's conclusion is wrong, and adding discipline will make the data worse, not the safety better. The programme has four design faults:
- It blames the person, not the system. Nearly every observation records 'PPE not worn' or 'unsafe act'. But the mask is one size, uncomfortable in 38 °C heat, and never fit-tested - so not wearing it is a rational response to a badly designed control. The programme is detecting a management failure and recording it as a worker failure.
- It records outcomes, not causes. 'Safe/unsafe' ticks capture no antecedent (what made the act attractive) and no consequence (what the worker gets out of it). That data cannot be acted on, so observation becomes a paperwork ritual.
- The quota destroys quality. Four observations per month per observer forces drive-by ticking at the end of the month, so the count measures compliance with the quota, not exposure.
- It is punitive. Copying observations into appraisals turns the observer into an enforcer. Workers warn each other, near-miss reporting drops 60% - the classic signature of a reporting system that has become unsafe to use. Falling observation numbers are a symptom of lost trust, not of a better culture.
The plant has not got a bad culture; it has got a programme that punishes people for telling the truth.
2. How would you redesign the programme, and what would you fix first?
Read the worked answer
Fix the barrier first. Order a range of mask sizes and models, run a fit test for every wearer, and trial alternatives for heat and humidity. If the control cannot be worn comfortably for a full shift, no amount of behavioural coaching will make it work. Until that is done, PPE observations tell you about the mask, not the worker.
Redesign the observation:
- Record antecedent, behaviour and consequence, plus the conditions seen (housekeeping, access, guarding), not just a safe/unsafe tick.
- Separate the person from the finding: log the barrier ('no mask fits this worker'), and route it to the action tracker with an owner and a date.
- Remove the monthly quota; measure observation quality by a sample check and by the proportion of findings converted into closed actions.
- Replace 'you were wrong' with a two-way conversation: ask why, thank the worker, act on what they say.
- Firewall the data from discipline and appraisals, and be seen to do it - announce it and honour it.
Use the hierarchy, not just behaviour: the underlying issue is that the plant is trying to control dust with PPE. Enclosure, extraction at the bagging spouts and dust suppression reduce exposure for everyone regardless of behaviour. Behaviour is the last layer, not the first.
3. What should the officer propose for measuring whether the redesigned programme is working, and how should new and existing workers be trained and inducted?
Read the worked answer
Leading indicators, not just injury counts:
- Near-miss and hazard report rate (expect a rise first - that is recovery of trust), with time-to-close for each report.
- Proportion of observation findings converted into closed corrective actions; overdue actions on the tracker.
- Observation quality audit score from co-observation of a sample; participation across shifts and contractors.
- Barrier fixes completed (masks fit-tested, extraction installed) and exposure measurements trending down.
- A short anonymous safety climate pulse survey, repeated, with the same questions, and acted on publicly.
Use lagging data only alongside these; a low injury rate with collapsing reporting is a warning, not a success.
Training and induction:
- Induction must cover the site's real hazards, stop-work authority, how to report, and what happens to a report - with a walk-round of the actual area, not a video.
- Deliver in the languages spoken on site, with pictograms for the bagging crew, and check understanding by having workers demonstrate or explain, not by signature on an attendance sheet.
- Train observers separately on observation, feedback and bias, and re-certify them.
- Frontline supervisors need coaching skills before they are asked to run feedback conversations.
- Include contractors and new hires in the same programme, and record induction so that nobody reaches the bagging line without it.
Finally, report back to the workforce what changed as a result of their reports - visible follow-through is what rebuilds the reporting culture.