How to measure safety culture (without relying on injury rates)

Ask most safety managers how they measure safety culture and you'll get one of two answers: "We look at our injury rates" or "We do an annual perception survey." Neither answer is wrong, but both are incomplete. Injury rates tell you about outcomes, not culture. Perception surveys tell you about attitudes at a single point in time, not about what people actually do day to day.

Safety culture is what happens when nobody's watching. Measuring it properly means looking at multiple layers: what the data shows, what people believe, and what people do. Here's how to build a measurement approach that captures all three.

The problem with injury rates as a culture measure

LTIFR, TRIFR, and DART rates are important. They're legally required in most jurisdictions, they're understood by boards and insurers, and they provide a necessary record of what's happened. But as measures of safety culture, they have three fundamental problems.

They're lagging. By the time an injury rate changes, the cultural conditions that caused it have been in place for months or years. A rising LTIFR doesn't tell you what went wrong in your culture. A falling one doesn't tell you what you fixed. The delay between cause and effect makes it nearly impossible to connect specific cultural interventions to specific outcomes.

They're statistically unreliable at the facility level. Most individual sites don't have enough incidents to produce statistically meaningful trends. A single serious injury can swing a facility's LTIFR dramatically, making it look like culture deteriorated overnight when the underlying conditions may have been unchanged for months. Conversely, a run of zero injuries can create a false sense of security when risk is actually building beneath the surface.

They incentivise the wrong behaviour. When injury rates are tied to bonuses, contracts, or performance reviews, the pressure to keep the number low can suppress reporting rather than reduce risk. Research consistently shows that organisations with strong incentive programmes tied to lagging indicators often have higher underreporting rates. The NSC's 2025 Membership Benchmarking Survey identified safety culture and employee engagement as two of the most pressing safety and health issues, partly because many organisations are measuring outcomes without measuring the cultural drivers behind them.

None of this means you should stop tracking injury rates. It means you shouldn't rely on them alone.

What leading indicators actually tell you about culture

Leading indicators are proactive measures that reflect the health of your safety system before an incident occurs. They don't predict individual events, but they reveal whether the conditions and behaviours that prevent incidents are in place and functioning.

Here are the leading indicators that most directly reflect safety culture.

Near-miss and safety event reporting rates. The volume of near misses reported is one of the strongest signals of cultural health. A high reporting rate indicates that workers feel psychologically safe enough to raise concerns, that they trust the system to respond constructively, and that they understand the value of reporting events that didn't result in injury. A low or declining rate almost always indicates a trust problem, not an absence of near misses.

With computer vision AI, near-miss detection becomes objective and continuous. The system captures events regardless of whether anyone chooses to report them, which means you get a complete picture of event frequency across the facility. But the voluntary reporting rate still matters as a cultural indicator, because it tells you whether your workforce is actively participating in safety or passively relying on the technology.

Event density trends over time. Raw event counts aren't particularly useful on their own. What matters is the trend. Is the number of detected events per zone decreasing week over week? Is the heatmap showing a reduction in hot spots after you made a traffic flow change or ran a coaching session? A declining trend in event density, particularly in zones where you've actively intervened, is strong evidence that the culture is responding to safety inputs.

Coaching session frequency and quality. How often are supervisors running coaching conversations using real safety event data? And more importantly, are those conversations translating into measurable behaviour change? Tracking coaching completion rates alongside the event density trends in the coached zones gives you a direct link between cultural activity (coaching) and cultural outcome (fewer events). If coaching is happening but event rates aren't changing, the quality of the conversations needs attention.

Time to corrective action. When a safety issue is identified (whether through an event detection, a worker report, or an audit finding), how long does it take to address it? Organisations with strong safety cultures close the loop quickly. Extended delays between identification and action signal that safety isn't being prioritised when it competes with operational demands. Track the average time from event identification to completed corrective action, and watch the trend.

Safety walk and inspection completion rates. Are safety walks being completed on schedule? Are the findings being acted on? Completion rates that drop during busy periods reveal how safety is prioritised relative to production pressure, which is one of the most telling cultural indicators available.

Perception indicators: what people believe

Leading indicator data tells you what's happening operationally. Perception data tells you what your workforce thinks and feels about safety, which is the subjective dimension of culture.

Safety climate surveys. A well-designed safety perception survey measures specific cultural dimensions: leadership commitment, psychological safety (do people feel safe raising concerns?), just culture (are honest mistakes treated differently from reckless behaviour?), reporting confidence, and trust in the system's responsiveness. The NSC Safety Barometer and the dss+ Safety Perception Survey are established tools with large benchmarking databases.

The key is to survey regularly (at least annually, ideally every six to twelve months) and to track the trends across dimensions, not just the overall score. An organisation might have strong scores on leadership commitment but weak scores on reporting confidence, which tells you exactly where to focus your cultural investment.

A Yale study using the NSC Safety Barometer found that each one-percentile increase in the overall survey score was associated with a decrease of approximately 13.6 work-related injuries per 10,000 employees. That's a meaningful correlation between perception and outcome.

Qualitative feedback. Surveys capture breadth. Qualitative methods (focus groups, skip-level conversations, open-ended feedback) capture depth. They reveal the stories behind the scores: why people don't report near misses, what happens when someone raises a concern, how leaders respond when safety and production conflict. These narratives are harder to quantify, but they often contain the most actionable insights.

Behavioural indicators: what people actually do

The third layer of measurement bridges the gap between what the data shows and what people say. It captures what people actually do in practice, which is the truest expression of culture.

How leaders respond to safety events. When a near miss is detected, does the supervisor use it as a coaching moment or ignore it? When a worker raises a concern, does the response come within hours or weeks? When safety and production conflict, which one wins? These leadership behaviours are observable, and they shape culture more powerfully than any policy document.

Whether workers intervene when they see risk. In a strong safety culture, workers don't just avoid risk themselves. They speak up when they see others at risk. Tracking whether workers actively intervene (stopping a task, flagging a hazard, pulling a colleague aside) is a direct measure of cultural maturity. It's difficult to quantify through traditional reporting, but qualitative observations and coaching conversations can surface it.

How the organisation responds to failure. Does a safety incident trigger a learning conversation or a blame exercise? Are investigation findings shared openly or buried in a file? Does the organisation treat failures as opportunities to improve the system, or as evidence that someone didn't follow the rules? The response to failure is one of the most visible and influential expressions of safety culture, and it's directly observable.

Putting it together: a practical measurement framework

The strongest safety culture measurement approaches combine all three layers: operational data (leading indicators), perception data (surveys and qualitative feedback), and behavioural data (observable leadership and worker actions). Here's a practical cadence.

Weekly: Review event density trends and coaching completion rates at the site level. This gives you a real-time pulse on whether the cultural inputs (coaching, operational changes) are translating into cultural outcomes (fewer events).

Monthly: Review heatmap comparisons and time-to-corrective-action trends in the operations meeting. Identify which zones improved, which didn't, and what operational or cultural factors explain the difference.

Quarterly: Review aggregate leading indicator trends against the baseline. Compare site-to-site performance for multi-site organisations. This is the data that goes into your board report alongside (not instead of) the lagging indicators.

Annually (or bi-annually): Run a safety perception survey. Compare results to the previous survey and to industry benchmarks. Correlate the perception data with the leading indicator trends to identify where beliefs and behaviours are aligned and where they diverge.

Continuously: Observe leadership behaviours through coaching quality reviews, skip-level conversations, and qualitative feedback. These observations don't need a formal schedule, but they do need consistent attention.

The metric that matters most

If I had to pick a single metric that best reflects safety culture, it wouldn't be an injury rate. It would be the trend in leading indicator event density across your monitored zones, overlaid with the coaching activity that's driving it.

A declining event density trend, in zones where coaching is happening regularly, where corrective actions are being closed quickly, and where workers report feeling heard and supported, is the strongest evidence you can get that your safety culture is genuinely improving. Not because someone got lucky and nobody got hurt this quarter. Because the system is working.

That's what measuring safety culture actually looks like. And if you'd like to see what continuous leading indicator data can tell you about yours, book a demo and we'll walk you through it.

Frequently Asked Questions

Why shouldn't I rely on injury rates to measure safety culture?

Injury rates (LTIFR, TRIFR) are lagging indicators that reflect outcomes, not culture. They change too slowly to connect to specific interventions, they're statistically unreliable at individual facility level due to small sample sizes, and they can incentivise underreporting when tied to bonuses or performance reviews. They should be tracked, but they shouldn't be your primary measure of cultural health.

What are the best leading indicators for safety culture?

The most informative leading indicators include near-miss and safety event reporting rates (both voluntary and AI-detected), event density trends over time by zone, coaching session frequency and completion rates, time from event identification to corrective action closure, and safety walk completion rates. Tracking these alongside the event density trends in coached zones gives you a direct link between cultural activity and measurable outcomes.

How often should I survey my workforce on safety culture?

At least annually, ideally every six to twelve months. More frequent surveying allows you to track trends and measure the impact of specific cultural interventions. The survey should measure specific dimensions (leadership commitment, psychological safety, reporting confidence, just culture) rather than just producing an overall score, so you can identify exactly where to focus your efforts.

How do I measure whether coaching is improving safety culture?

Track coaching session frequency (how often supervisors are running coaching conversations using real event data) alongside event density trends in the coached zones. If coaching is happening regularly and event density is declining in those zones, you have strong evidence that coaching is driving behavioural change. If coaching is happening but event density isn't changing, the quality of the conversations needs attention.

What is the single best metric for safety culture?

The trend in leading indicator event density across your monitored zones, correlated with coaching activity and corrective action closure times. A declining event density trend in zones where coaching is active and corrective actions are being closed quickly is the strongest evidence of genuine cultural improvement, because it reflects the system working continuously rather than depending on the absence of injuries.