Why leading indicators beat TRIR — and how to get your team to report them

Every safety program we have ever reviewed has a TRIR chart on the executive dashboard. Almost none have a leading-indicator chart next to it. That is a measurement problem disguised as a reporting problem.

Near-misses are the canonical leading indicator, but the moment you ask field crews to stop and report one, the friction tax kills the signal. The fix is not a campaign — it is a 30-second mobile form with one required field: what happened. Everything else is optional or captured by the system (location via GPS, reporter via session, time via timestamp).

• Mobile-first capture — no laptop, no kiosk. If a foreman cannot file from the dashboard of a truck, it will not happen.
• Reward the behaviour, not the outcome. Teams that report more near-misses should get more air-time with leadership, not more scrutiny.
• Close the loop visibly. Every near-miss should produce either a CAPA, a documented no-action decision, or an alert to the responsible supervisor within 48 hours.

The three leading indicators we track across every QEHS tenant: near-miss rate per 200,000 hours, inspection closure time, and corrective-action overdue rate. Once those three start moving, TRIR follows within two quarters.

The harder question is whether a leading indicator is predictive or performative. A high count of safety observations means nothing if every observation reports no issue, and a 100% training-completion rate can hide a 10-minute video nobody watched. The indicators that actually predict harm are specific, behavioural, and tied to a known risk: the percentage of hot-work permits with a gas-test record attached, the percentage of Lockout-Tagout procedures verified by a second competent person, and the percentage of incident investigations closed within 48 hours with a root cause named. Research on serious injuries and fatalities is consistent — organisations that surface hazards early through high near-miss reporting have lower SIF rates.

The reporting only matters if supervisors can see it in time to act. When the near-miss rate at a site drops 40% month-over-month, the right response is to ask whether operations got safer or whether a chilling effect stopped people reporting — and the data is what lets you tell the two apart. For the metric definitions, see the TRIR, LTIR, and leading-indicator glossary entries; for program design, the behaviour-based safety guide.

The distinction that matters operationally is between a leading indicator and a vanity number. A leading indicator has three properties: it is behavioural (it measures something people do, not something that happens to them), it is movable (the program can change it through a specific intervention), and it has a defensible causal link to a lagging outcome. A metric that fails any of the three is a dashboard ornament. Near-miss rate passes all three when the program has a low-friction reporting channel; it fails the third when the channel exists but no one closes the loop, because reporters learn that filing changes nothing and the rate collapses to whatever is forced.

The leading-indicator set a program should hold has to be small. A dashboard of forty leading indicators is not a measurement system; it is a backlog. The set that consistently predicts serious outcomes across heavy industry is narrow: near-miss rate per 200,000 hours, inspection closure time, corrective-action overdue rate, training currency, and permit compliance rate (the share of high-risk permits issued with every required control verified). Five numbers, each with an owner and a target, reported on the same cadence as TRIR, is enough to steer a program. Adding a sixth is defensible only when a specific risk demands it.

• Near-miss rate per 200,000 hours — reports divided by hours worked, normalised to the TRIR scale so the two sit on the same axis. Target: a rate that rises during a reporting campaign and holds steady at a program-defined level thereafter; a falling rate with no program change is the under-reporting signal.
• Inspection closure time — median days from inspection submitted to all findings closed or accepted as no-action. Target: under 14 days for low-severity findings, under 48 hours for high-severity. A finding older than 30 days is an unremoved hazard.
• Corrective-action overdue rate — open CAPAs past their due date divided by total open. Target: under 5 percent. Overdue is a stronger signal than open, because open is a workload number and overdue is a control-failure number.
• Training currency — the share of workers current on every role-required course. Target: above 95 percent, with the 5 percent gap named and scheduled, not hidden.
• Permit compliance rate — the share of hot-work, confined-space, and height permits issued with every required control verified before work started. Target: 100 percent; anything below means the permit-to-work system is a paperwork step rather than a control.

The serious-injury-and-fatality question is the one that has reshaped how leading indicators are used. The research consensus, built on work tracing serious outcomes back through precursor events, is that fatalities and serious injuries do not cluster where the recordable cases cluster — they cluster around high-energy hazards and a small set of life-saving controls. A program that steers only on TRIR optimises the frequency tail and leaves the severity head untouched. The leading indicators that predict SIFs are control-specific: the percentage of energy-isolation events verified by a second competent person, the percentage of working-at-height jobs with a rescue plan in place before the work starts, and the percentage of confined-space entries with a fresh gas test on the day of entry. These are narrower than the general set, and they are the ones a serious-injury program is built around.

The OSHA Voluntary Protection Programs and the ISO 45001:2018 clause 9.1 monitoring expectation both push toward leading indicators, though neither prescribes a specific set. The ISO 45001 language asks the organisation to determine what needs monitoring and the methods for it, which is the opening a safety leader uses to argue for the leading set on the management-review agenda rather than in a side report. The regulator-readable position is that a program which measures leading indicators and can show them moving is a program exercising due diligence; a program which measures only lagging indicators is a program counting what it failed to prevent.

1. Stand up one mobile form for near-misses with a single required field (what happened) and automatic capture of reporter, location, and time. The 30-second test: a foreman in a truck cab can file one without putting down a phone.
2. Route every medium-and-above near-miss to an auto-created CAPA with a suggested owner pulled from the location hierarchy. The supervisor can close it as no-action, but the decision is logged.
3. Build the leading-indicator dashboard against the five-number set, reported on the same weekly cadence as the incident log so the two read together.
4. Add a guard on the permit-to-work system that refuses to issue a high-risk permit when any required control is unverified; the verification rate feeds the permit-compliance leading indicator automatically.
5. Review the set at every management review, and retire any indicator that has not moved and has no program action behind it. A leading indicator that never changes is a weather report.

The last step is the one most programs skip. A leading-indicator program that is never reviewed and never pruned turns into the same kind of inert dashboard it was meant to replace. The discipline is to treat the set the way the hierarchy of controls treats a hazard: walk it up as far as you can, retire what does not work, and keep the few that do in front of the people who can act on them. For the related metrics, see the TRIR, LTIR, DART, and leading-indicator glossary entries; for the program design, the behaviour-based safety guide and the near-miss to CAPA handoff post.

The cultural failure mode is the one the dashboard cannot see. A leading-indicator program in a low-trust culture produces a high, clean near-miss rate that a high-trust site with a real hazard load would envy — except the reports are the same three near-misses recycled, filed to hit a target. The defence is the stop-work-authority leading indicator: the percentage of workers who report having used stop-work authority in the last quarter, and the percentage of stop-work events upheld by supervision. A program where stop-work authority exists on paper but is never exercised is a program where the leading-indicator dashboard is measuring compliance, not safety. The indicators that require courage to move are the ones that distinguish a reporting culture from a reporting theatre.

The reporting loop is the unit of measurement, not the report. A near-miss that is filed, triaged, owned, and closed is one cycle; a near-miss that is filed and never opened is half a cycle and it teaches the reporter that filing is theatre. The leading-indicator that captures the loop is the closure rate, and the closure rate is the one a safety leader watches before the near-miss rate, because a reporting program without a closure program is a complaint box.