Safety analytics: how data turns reactive safety into proactive prevention

Here is a question worth asking at your next safety meeting: how much of your safety data tells you what already went wrong, and how much tells you what is about to?

For most organisations, the honest answer is heavily weighted toward the past. Incident rates. Lost-time injuries. Workers' compensation claims. DART cases. These are the metrics that fill safety reports, get presented to boards, and satisfy regulatory requirements. They are also, by definition, a record of failure. Every data point represents someone who was already hurt.

The National Safety Council's Campbell Institute has spent over a decade researching this problem. Across five foundational white papers on leading indicators (published between 2013 and 2025), their consistent finding is that sole focus on lagging metrics is not as effective in promoting continuous improvement as using leading indicators to anticipate and prevent injuries and incidents. Their most recent paper, The Challenge of Safety Metrics (2025), concluded that leadership engagement is the single most critical enabler of effective leading indicator programmes, and that data quality and usability matter more than quantity.

The shift from reactive to proactive safety is fundamentally a shift in what you measure, how you capture it, and what you do with it.

The problem with lagging indicators

Lagging indicators are not useless. They provide the baseline against which all improvement is measured, they satisfy compliance requirements, and they reveal patterns over time. The Bureau of Labor Statistics' Census of Fatal Occupational Injuries tracks 5,070 workplace fatalities in 2024, while the NSC reports 84 forklift-related deaths and over 25,000 DART cases in 2023-2024. This data matters.

But lagging indicators have three fundamental limitations. First, they require something bad to happen before they generate a data point. You cannot learn from an incident that has not occurred yet. Second, they have low frequency and high variability. A site can go months without a recordable injury, then have two in a week, making trend analysis unreliable. Third (and this is the critical one), they tell you nothing about the risks that currently exist on your site right now.

A warehouse with zero incidents this quarter might have 500 unreported near misses. A site with a declining injury rate might be accumulating risk through deteriorating behaviours that nobody is tracking. Lagging indicators cannot see this. They will only tell you about it after someone gets hurt.

Research published in the journal Safety Science found that lagging indicators are limited in their ability to predict future incidents because they are post-incident artefacts. They describe what happened, but they are not causally related to what will happen next.

What leading indicators actually look like

The Campbell Institute defines leading indicators as proactive, preventive, and predictive measures that monitor the effective performance of a safety management system. They categorise them into three types: systems-based indicators (relating to the management of EHS systems), operations-based indicators (relevant to organisational infrastructure), and behaviour-based indicators (measuring the actions of individuals or groups).

In practical terms, leading indicators for a warehouse or manufacturing operation might include the number of near misses reported per shift, the percentage of coaching conversations completed after safety events, exclusion zone breach rates by zone and time of day, speed violation events across the forklift fleet, corrective actions raised versus corrective actions closed, the ratio of safety observations to incidents, and training completion rates for new and existing workers.

The key difference is that every one of these can be measured before an incident occurs. They tell you about the health of your safety system right now, not just what it failed to prevent last month.

A 2019 study published in the American Journal of Industrial Medicine tested associations between leading and lagging indicators across 2,198 construction contractors, finding that higher safety management system scores (a leading indicator) related directly to lower injury rates. The evidence is clear: what you measure proactively predicts what happens reactively.

Why most organisations struggle with leading indicators

If leading indicators are so valuable, why do most organisations still rely primarily on lagging data? The Campbell Institute's 2025 research identified several persistent barriers.

The first is data collection. Leading indicators require ongoing, systematic capture of events and observations that traditional safety systems were not designed to collect. Near misses, speed events, and exclusion zone breaches happen continuously, but manual reporting captures only a tiny fraction. Estimates vary, but the gap between what actually happens and what gets reported is enormous.

The second is data quality. Even when organisations collect leading indicator data, it is often inconsistent, subjective, or incomplete. If your near-miss reports depend on individual workers choosing to fill out a form, the data will be shaped by reporting culture, shift patterns, and individual willingness more than by actual risk.

The third is the link between data and action. Many organisations that do collect leading indicators struggle to turn them into interventions. The data goes into a dashboard. Someone reviews it monthly. But the feedback loop from detection to analysis to action to improvement is too slow or too disconnected from daily operations to drive real change.

How computer vision AI solves the data problem

This is where the conversation shifts from theory to practice. The fundamental challenge with leading indicators has always been the gap between what happens and what gets captured. Computer vision AI closes that gap.

By analysing footage from existing CCTV cameras, computer vision AI continuously and automatically detects the events that leading indicators depend on: near misses, pedestrian-vehicle interactions, exclusion zone breaches, speed violations, and unsafe behaviours. It does this 24 hours a day, across every monitored zone, without relying on a human to observe and report.

The result is a fundamentally different data set. Instead of a handful of manually reported near misses per month, you have thousands of automatically captured events. Instead of subjective observations, you have consistent, timestamp-verified data tied to specific locations and video evidence. Instead of a monthly snapshot, you have a continuous feed.

inviol's reporting and heatmap features turn this data into the leading indicators that safety teams need. Heatmaps show where risk concentrates by zone and time. Trend lines show whether interventions are working. Shift comparisons reveal performance differences. Multi-site dashboards enable benchmarking across locations. The data is structured around the questions that actually drive decisions: where is the risk, when does it peak, is it getting better or worse, and what should we do about it?

From data to action: the coaching loop

Data without action is just noise. The Campbell Institute's research is emphatic on this point: the organisations that succeed with leading indicators are those that connect measurement to intervention through clear, consistent processes.

inviol's coaching and training platform creates this connection. When the data shows a pattern (a particular intersection generating repeated near misses during the morning shift, for example), the system provides the specific video evidence that makes the pattern concrete. A supervisor can sit down with the relevant team, review the clips (faces blurred for privacy), discuss what is happening and why, and agree on changes.

This is the feedback loop that turns analytics into prevention. Detection feeds data. Data reveals patterns. Patterns trigger coaching conversations. Coaching changes behaviour. Changed behaviour reduces risk. And the system measures whether the risk actually decreased, completing the loop.

inviol customers typically see a 67% reduction in risk through this approach. That reduction is not because the technology detected more hazards (although it does). It is because the coaching loop turns detection into lasting behaviour change.

Measuring what matters

The shift from reactive to proactive safety is not about abandoning lagging indicators. You still need to track injuries, lost time, and workers' compensation costs. Regulators require it. Boards expect it. And the long-term trend in these metrics is the ultimate measure of whether your programme is working.

But if lagging indicators are all you measure, you are flying blind between incidents. Leading indicators, powered by continuous data from computer vision AI, give you visibility into the risks that exist right now. They tell you where to focus, whether your interventions are working, and what to coach next.

In New Zealand, the Health and Safety at Work Act 2015 requires PCBUs to identify, assess, and manage risks. In Australia, Safe Work Australia's data shows 80% of fatalities concentrated in six industries, intensifying the expectation for proactive risk management. Leading indicator programmes supported by continuous monitoring provide exactly the kind of evidence that demonstrates a business is genuinely managing risk, not just recording outcomes.

If you want to see what your safety data could look like with continuous leading indicators across your operation, book a demo and see inviol's analytics and coaching platform in action.

Frequently Asked Questions

What is the difference between leading and lagging safety indicators?

Lagging indicators measure outcomes after they occur, such as injury rates, lost-time incidents, and workers' compensation claims. Leading indicators are proactive, predictive measures that monitor the performance of safety systems before incidents happen. Examples include near-miss rates, coaching conversations completed, exclusion zone breaches, and corrective actions closed. Both are important, but leading indicators enable prevention rather than just measurement.

Why are lagging indicators insufficient on their own?

Lagging indicators require something bad to happen before generating a data point. They have low frequency (making trend analysis unreliable), high variability, and tell you nothing about risks currently present on your site. A site with zero incidents may still have hundreds of unreported near misses and deteriorating behaviours that only show up in the data after someone is injured.

How does computer vision AI improve safety analytics?

Computer vision AI uses existing CCTV cameras to continuously and automatically detect safety events like near misses, exclusion zone breaches, speed violations, and pedestrian-vehicle interactions. This closes the gap between what actually happens and what gets reported, providing a continuous, consistent, and objective data set that makes leading indicators reliable and actionable.

What leading indicators should a warehouse track?

Key leading indicators for warehouse operations include near-miss rates by zone and shift, exclusion zone breach frequency, speed violation events, coaching conversations completed, corrective actions raised versus closed, safety observation ratios, and training completion rates. The specific indicators should be chosen based on your highest-risk activities and reviewed regularly to ensure they remain meaningful.

What does the Campbell Institute recommend for leading indicator programmes?

The Campbell Institute (part of the National Safety Council) has published five foundational white papers on leading indicators since 2013. Their key findings include that leadership engagement is the most critical enabler, data quality matters more than quantity, cross-functional collaboration and site-level ownership drive adoption, and non-punitive reporting cultures support sustained use. Their most recent paper (2025) emphasises that building the cultural and structural foundation to support leading indicators is what turns potential into performance.