HyperQ AI Safety deploys on existing ONVIF-compatible CCTV in under 1 hour, producing an event index across the installed camera network from the first shift. Within 2–3 weeks of initial configuration — after zone sensitivity thresholds settle — the system generates measurements that most EHS management systems currently cannot produce: a daily count of PPE compliance gaps per zone, a zone dwell-time distribution for restricted areas, and a near-miss detection frequency. These are not dashboard cosmetics. They are leading indicators in the specific sense that ISO 45001 and the WSH Council's SMS framework use the term: measurements that precede incidents rather than count them.
This post covers why Total Recordable Incident Rate fails as a primary safety metric, what leading indicators actually measure and why they survive an audit when lagging metrics do not, which leading indicators AI cameras can measure reliably, which ones they cannot, and what the monitoring data shows in the first 90 days.
Why TRIR hides risk
TRIR — Total Recordable Incident Rate, calculated as recordable incidents per 200,000 hours worked — is the default metric on most EHS management dashboards. It is also the metric most likely to tell you nothing until after something has gone wrong.
TRIR is a lagging indicator. It records events that have already occurred. A TRIR of zero in a given quarter means no recordable incidents were logged. It does not mean the site was safe. It means no incident completed during that period.
The specific problem is what a sustained low TRIR conceals. A site where workers routinely bypass machine guarding, skip hard hats in certain areas, and take the shortest path through restricted zones can maintain a low TRIR for an extended period — until the exposure sequence completes. When it does, the incident investigation reveals a compliance pattern that was visible for months. The TRIR reported zero throughout. The hazard accumulated continuously.
When a Busan-based manufacturer first deployed camera-based leading indicator monitoring, their TRIR for the preceding 18 months showed two recordable incidents and a rate of 0.8 per 200,000 hours — well within WSH Act reportable thresholds, and a number any EHS manager would present to their board as satisfactory performance. The first 8 weeks of camera monitoring showed a different picture: 34 zone breach events, 67 PPE gap corrections at entry, and a night-shift PPE compliance rate 18 percentage points lower than the day shift for the same zones. The TRIR was low. The leading indicators showed where the next incident would come from if the pattern continued. The plant's EHS manager ran a targeted intervention on the night-shift crew in weeks 9 and 10. By week 12, the compliance gap had narrowed to 7 percentage points. The TRIR stayed at zero during that period. The leading indicators showed the work that kept it there.
What auditors actually want
EHS auditors conducting WSH Act SMS audits in Singapore and DOSH facility audits in Malaysia ask the same type of question: "Show me that your safety management system is operating, not just documented."
A TRIR trend line is documentation. It records what happened. A PPE compliance trend line is evidence of system operation: it shows that the organisation monitors specific behaviours continuously, identifies gaps, and responds. That is the corrective action loop that ISO 45001 Clause 9.1 (monitoring, measurement, analysis, and performance evaluation) requires organisations to demonstrate.
The specific standard is: monitoring and measurement of the extent to which operational controls are implemented and effective. For a plant where PPE is an operational control for a specific hazard class, monitoring of PPE use is a Clause 9.1 requirement. A once-annual site survey satisfies the documentation requirement in a minimal way. A monthly trend line produced by continuous camera monitoring satisfies it in the way that demonstrates the control is operating.
Auditors distinguish between these. A safety poster saying "PPE required in this area" is an administrative control. An EHS manager who can show that PPE compliance in that specific area was measured at 81% in month 1, 88% in month 2, and 94% in month 3, and can show the corrective action that moved the number, is demonstrating that the administrative control is backed by an active monitoring system. The trend line, not the poster, is the audit evidence.
What AI cameras measure as leading indicators
The leading indicators that camera-based monitoring can produce fall into three categories.
PPE compliance rate by zone and shift. For each camera zone where PPE is required, the system produces a daily and weekly compliance rate: the fraction of worker-entries where the required PPE was detected as present. This rate can be trended over time, broken out by zone, and compared across shifts. A zone where compliance is consistently lower on the night shift than the day shift is a zone where the management intervention is shift-specific, not site-wide.
The compliance rate is a population-level measure, not a binary count of violations. Of 180 worker-entries through Zone A this week, 163 had full required PPE — a non-compliance rate of 9.4%, down from 14.2% the prior week. That is a measurement the EHS manager can act on, trend, and present to an auditor as evidence of continuous monitoring.
Near-miss detection frequency. Zone boundary breaches, man-down events, fall alerts, and PPE gap events in the event index are each a category of near-miss: situations where the exposure sequence began but the incident did not complete. The weekly frequency of these events, categorised by type and zone, is a leading indicator showing which areas carry the highest active hazard exposure.
A zone with 12 near-miss detection events in a week is a zone where the hazard is actively being encountered, regardless of whether any of those events resulted in a recordable incident. The event frequency is the signal. The TRIR from that zone this week is zero; the near-miss frequency is 12. An EHS manager using only TRIR has no signal. The same manager using near-miss detection frequency has 12 entries to review and act on.
Zone dwell-time distribution. For hazard zones that workers should pass through rather than stay in — an area adjacent to a press during operation, a chemical storage corridor, a high-noise zone where extended presence creates hearing risk — the system logs the duration of each worker presence event. The dwell-time distribution shows whether workers are exceeding safe exposure limits, and how frequently.
Extended dwell-time is a different kind of leading indicator than near-miss frequency: it measures a risk-accumulating behaviour rather than a risk-realisation event. The worker who spends 45 minutes in a zone with a 15-minute exposure limit has not had an incident yet. The dwell-time log shows they will have accumulated three times the recommended exposure if the pattern continues uncorrected.
The leading indicator matrix
The following maps each indicator to the camera measurement, the audit evidence it produces, and the ISO 45001 clause it directly addresses.
| Leading indicator | Camera measurement | Audit evidence produced | ISO 45001 clause |
|---|---|---|---|
| PPE compliance rate by zone | Fraction of worker-entries with full required PPE per zone per shift | Weekly compliance trend with corrective action history | 9.1.1 — monitoring of operational controls |
| Near-miss frequency by type | Count of zone breach, PPE gap, fall, man-down events per week | Event log with clips, alert routing, and response records | 10.2 — incident, nonconformity and corrective action |
| Zone dwell-time | Duration of each worker-presence event in restricted zones | Dwell-time distribution with events exceeding exposure limits | 9.1.1 — monitoring of operational controls |
| Night-shift vs day-shift compliance gap | Compliance rate by shift segment | Shift-level comparison trend for targeted interventions | 6.1.2 — hazard identification and risk assessment |
| Post-corrective-action compliance trend | Rate before and after a specific intervention | Evidence that corrective actions moved the measured number | 10.2 — corrective action effectiveness |
| Contractor PPE compliance | Per-crew entry compliance rate and zone breach count | Per-contractor compliance record for SMS contractor management requirements | 8.1.4 — contractor and outsourcing management |
How thresholds and trends work together
The leading indicator value is only useful when measured against a threshold. A PPE compliance rate of 88% in zone A is a measurement. It is not a signal until compared against last week's rate, the target rate, or the rate in a comparable zone.
The threshold configuration in HyperQ AI Safety sets two parameters per zone: the alert threshold (the event type that generates a real-time notification) and the trend threshold (the compliance rate floor that triggers a management review when breached). These are separate decisions with different purposes.
A zone might have an alert threshold that triggers an immediate notification for any hard-hat absence, and a trend threshold that triggers a management review when weekly compliance falls below 90%. The alert threshold is for immediate response. The trend threshold is for system-level management — and it is the audit-relevant setting.
When an auditor asks what the target is for PPE compliance in the press area, the documented answer is "90%, measured weekly, with a management review triggered when the rate falls below that." When the auditor asks what happened when compliance dropped to 87% in February, the EHS manager shows the review record and the corrective action taken. That exchange — target, measurement, gap, response — is what ISO 45001 Clause 9.1 requires and what an EHS management system that relies only on TRIR cannot produce.
Where camera monitoring does not produce leading indicators
Several safety dimensions commonly tracked as leading indicators are not measurable by camera-based monitoring.
Noise exposure and vibration. A camera can detect that a worker is in a high-noise zone. It cannot measure the decibel exposure the worker received. For hearing damage risk indicators, dosimetry is the appropriate instrument.
Near-misses that leave no visual trace. A worker who has a near-miss event without a visible component — a slip that does not result in a fall, a chemical splash that does not produce a visible reaction — will not be detected by camera monitoring. The event index represents the subset of near-miss events that are visually distinguishable. It is not a complete near-miss register.
Skills and procedure compliance. A camera detects whether a worker is wearing required PPE. It cannot detect whether the worker is following the correct operating procedure. Checklist-based procedure compliance requires direct observation by a trained supervisor.
Reported near-misses vs detected near-misses. The near-miss frequency from camera detection is distinct from the near-miss reporting rate. Camera detection adds events that would not have been voluntarily reported. When comparing leading indicator data across sites or presenting to an auditor, the EHS team needs to hold the camera-detected event log and the voluntary near-miss register separately — they measure different things and neither replaces the other.
What the first 90 days shows
The consistent pattern in the first 90 days of camera-based leading indicator monitoring is a declining event frequency alongside an improving compliance rate — not because the hazards change, but because visibility changes behaviour. Workers whose PPE compliance is measured are more consistent. Supervisors who receive weekly compliance reports by zone have a specific, evidenced conversation to have with the crew responsible for the low-rate zone rather than a general toolbox talk.
The 90-day trend line is the first document that survives an audit on its own terms. It shows the system was operating, the measurement was continuous, and the number moved. The period before monitoring was established shows nothing of use: the TRIR may show zero, but the compliance rate and near-miss frequency for that period are unknown. That unknown is what most EHS management systems are built on. They know what incidents occurred. They do not know the leading indicator values that preceded those incidents — or the values accumulating now ahead of the next one.
The transition from a TRIR-only reporting framework to one that includes camera-derived leading indicators does not require replacing the TRIR. The TRIR stays. It is the required reportable number for MOM incident reporting in Singapore and DOSH records in Malaysia. What changes is the additional column in the weekly EHS report: PPE compliance rate by zone, near-miss detection count by type, zone dwell-time events above threshold. The lagging metric records what happened. The leading metrics describe what is happening now and what the system is doing about it. Auditors can read both columns. The leading indicators are the ones that show the management system is running.
For ISO 45001 Clause 9.1 requirements under a Malaysian DOSH framework and for WSH Act SMS requirements in Singapore, the 90-day trend is the answer to the auditor's question about continuous monitoring. For the complete clause-by-clause mapping of what camera monitoring evidence satisfies and what stays outside the system's scope, the Malaysian manufacturer's guide to ISO 45001 evidence covers this in detail.
For EHS teams building a measurement programme from the existing camera network, the HyperQ AI Safety solution overview covers the zone configuration and alert routing options that produce the specific leading indicator outputs described in this post.
Share your current safety KPI dashboard — we'll map which leading indicators HyperQ AI Safety generates automatically from your existing CCTV, set the zone thresholds before the contract is signed, and show you what the first 90-day trend looks like on your own site.