From 600 near misses to zero serious incidents: a coaching success story

What does it actually look like when a logistics operation installs computer vision AI for safety and commits to coaching over punishment?

This is a question we get asked a lot. People understand the technology. They understand the concept. What they want to know is: what happens on the ground? How do the numbers move? What does the safety team actually do differently on a Tuesday morning?

Here's what the first 12 months typically look like for a multi-site logistics operation that deploys inviol and takes a coaching-first approach. The trajectory below reflects real patterns we see across our customer base. The specific numbers are representative, not from a single site, but they're consistent with the results our customers experience.

The starting point: good intentions, limited visibility

Most operations that come to us aren't in crisis. Their injury rate is close to the industry average. They have a safety team doing good work. They run toolbox talks. They have a near-miss reporting system that captures maybe 10 to 20 events per month across their sites.

From the outside, the numbers look fine.

But the safety manager usually has a nagging feeling. Ten to 20 near misses a month, across busy distribution centres running 24-hour shifts with dozens of forklifts? If Bird's safety triangle holds roughly true (1 serious injury to 10 minor injuries to 30 property damage events to 600 near misses, based on his 1969 analysis of 1.75 million accident reports), those 15 reports are barely scratching the surface.

The question isn't whether near misses are happening. It's how many are invisible.

What the data looks like in week one

When inviol's computer vision AI connects to existing CCTV cameras, the first week's data is typically startling.

Instead of 15 manually reported events per month, the system detects hundreds of safety events per week. Forklift-pedestrian proximity breaches. Speeding in high-traffic zones. Workers entering exclusion zones while vehicles are active. Pedestrians crossing intersections at the same time as forklifts.

The reaction from safety managers is almost always the same: "I knew we were underreporting. I didn't know it was by this much."

This isn't a reflection of a bad safety culture. It's a reflection of how manual reporting works. Workers are busy. Definitions are fuzzy. Forms add friction. Without visible follow-up, people stop reporting. The gap between what's happening and what's being captured is always enormous, regardless of how committed the safety team is.

The temptation to crack down (and why it backfires)

When you suddenly have visibility of hundreds of safety events per week, the natural instinct is alarm. And the natural follow-up is to tighten the rules, increase enforcement, and start disciplining the worst offenders.

We always advise against it, and the research explains why.

Behavioural safety research has shown for decades that punishment-based approaches tend to suppress reporting rather than change behaviour. When people are penalised for incidents, they stop admitting to them. The underlying behaviours continue; they just become invisible.

Amy Edmondson's research on psychological safety, first published in 1999 and expanded in her 2019 book The Fearless Organization, demonstrates the same principle. Teams where members feel safe to report errors and raise concerns without fear of punishment consistently outperform teams where those behaviours carry risk. In Edmondson's healthcare research, better teams reported higher error rates, not because they made more mistakes, but because they were more willing to talk about them.

With inviol, the data is captured whether workers report it or not, so the suppression problem doesn't apply to detection. But it absolutely applies to the culture you build around the data. If workers see the system as a surveillance tool that gets people in trouble, you lose the coaching opportunity entirely.

What happens next: coaching, not compliance

The operations that get the best results from inviol use the data for coaching, not punishment. Here's what that looks like in practice.

Every week, supervisors receive a summary of the highest-severity safety events from their shift, with blurred, privacy-protected footage. They use these in two ways.

First, individual coaching conversations. When a specific event involves a particular area or pattern, the supervisor sits down with the relevant team (not in a disciplinary meeting, but in a coaching conversation) and reviews the footage together. "Here's what happened at the B3 intersection on Tuesday night. Here's why it matters. What could we do differently?" The footage makes the conversation objective and grounded, not a matter of opinion or memory.

Second, group coaching during shift briefings and toolbox talks. The safety team selects one or two events each week that illustrate common risks and plays the footage for the whole shift. No names attached. No blame. Just a concrete example of a real event that everyone can learn from.

This approach aligns with what the research finds most effective. A longitudinal study by Krause et al. (1999) across 73 companies found that behaviour-based safety programmes using observation, feedback, and coaching achieved an average 26% reduction in incidents in the first year, increasing to 69% by the fifth year. DEKRA's 2022 analysis of 3.4 million observations across 354 sites confirmed that programmes with strong feedback and coaching components consistently reduce injuries.

The IMEC Technologies framework captures the distinction well: coaching, as a form of constructive feedback, generally yields the best results in changing behaviour. Punishment works only if delivered immediately and consistently, which is practically impossible. Coaching builds understanding, commitment, and self-regulation.

The typical 12-month trajectory

Here's how we see the numbers move for a multi-site operation that commits to coaching.

Weeks 1 to 4: The system surfaces the true scale of risk for the first time. High-severity events (forklift-pedestrian proximity breaches at speed, for example) are far more frequent than anyone expected. The safety team spends this period calibrating thresholds, understanding the data, and planning the coaching approach.

Weeks 5 to 12: As coaching conversations take hold, high-severity events start to drop. Workers begin self-correcting. Supervisors report that conversations about safety shift from "you need to be more careful" to "here's what the data shows about this intersection at shift changeover." The tone changes from enforcement to problem-solving.

Months 3 to 6: Heatmap data reveals that a small number of locations account for a disproportionate share of events. Operations adjusts traffic flow, repositions signage, or changes delivery schedules at those spots. These changes often improve throughput as well as safety, since forklifts are no longer slowing down to avoid pedestrians in congested zones. This is one of the unexpected benefits customers discover: the same data that reduces risk often reveals operational inefficiencies.

Months 6 to 12: High-severity events stabilise at a fraction of their original level. The safety team shifts focus from detection volume to trend analysis, using reporting dashboards to compare shifts, sites, and time periods. Coaching becomes embedded in daily operations rather than being a separate initiative.

The results we typically see

Across our customer base, inviol deployments achieve an average 67% reduction in safety risk events. Over three years, customers see an average 42% reduction in recorded incidents. Forklift-on-pedestrian and vehicle-on-plant events, which are among the highest SIF-potential (serious injury and fatality potential, as the Campbell Institute defines it) events in any warehouse, typically see reductions of 61% or more.

Operations that fully commit to coaching (using the data in weekly briefings, running individual coaching conversations, and acting on heatmap insights) consistently outperform those that treat the system as a passive monitoring tool.

Five things that make the difference

Looking across our deployments, the same factors come up again and again.

The data is objective. When a supervisor shows blurred footage of a specific event, there's no argument about whether it happened. The conversation starts from shared reality, not conflicting accounts. This removes one of the biggest barriers to effective safety coaching.

The volume reveals patterns. Fifteen manually reported near misses tell you nothing about where, when, or why events are happening. Hundreds of events per week reveal that shift changeover is the highest-risk period, that two intersections are responsible for most events, and that speeding is concentrated on night shifts. Those patterns drive targeted interventions that actually work.

Coaching creates psychological safety. Because the data is used for learning rather than punishment, workers start engaging with it. Supervisors report that team members begin pointing out risks during briefings. Edmondson's research shows that psychological safety protects against burnout and turnover, even under pressure. For the operations we work with, it becomes the foundation of a stronger safety culture.

Operational improvements come free. The heatmap insights don't just reduce risk, they improve efficiency. Intersection redesigns reduce forklift travel time. Adjusted delivery truck arrival times ease pedestrian congestion at certain doors. Customers regularly tell us they discovered that existing processes or layouts were increasing both risk and inefficiency, and the data let them fix both at once.

It scales naturally. Because the coaching framework is consistent across sites, the approach scales without needing different interventions for each location. inviol only needs a selection of cameras focused on the highest-risk areas, not every camera on site, which keeps deployment fast and focused.

What this means for compliance

Under the Health and Safety at Work Act 2015, PCBUs in New Zealand have a duty to identify and manage workplace risks so far as is reasonably practicable. In Australia, similar obligations apply under WHS legislation. In the U.S., OSHA's framework emphasises proactive hazard identification.

A system that automatically detects safety events, feeds them into documented coaching workflows, and tracks risk reduction over time provides strong evidence that your organisation is meeting its duty to proactively identify and manage risk. That evidence matters when a regulator asks what you're doing to prevent serious harm.

If this sounds like your operation

Most of the sites we work with start from the same place: a reasonable injury rate, a small handful of near-miss reports, and a safety team that suspects there's much more happening than what they can see.

If that's you, book a demo and we'll show you what a coaching-first approach looks like with real data from your kind of operation. The visibility comes first. The behaviour change follows.

Frequently Asked Questions

How does coaching reduce safety incidents more effectively than punishment?

Behavioural safety research shows that punishment-based approaches suppress reporting rather than change behaviour. Workers stop admitting to near misses when they fear consequences, making risk invisible. Coaching builds understanding and self-regulation instead. A five-year study across 73 companies found that behaviour-based programmes using observation, feedback, and coaching achieved an average 69% reduction in incidents by year five.

How many near misses does inviol typically detect compared to manual reporting?

Most sites detect hundreds of safety events per week through inviol's computer vision AI, compared to 10 to 20 manually reported near misses per month. This is consistent with Bird's safety triangle, which predicts a ratio of 600 near misses for every serious injury. Manual systems capture only a fraction of that volume because of time pressure, reporting friction, and unclear definitions.

What is psychological safety and why does it matter for workplace safety?

Psychological safety, as defined by Harvard professor Amy Edmondson, is a shared belief that a team is safe for interpersonal risk-taking. In workplace safety, this means workers feel able to report near misses, raise concerns, and discuss errors without fear of punishment. Research consistently shows that psychologically safe teams report more issues and achieve better outcomes, because problems are surfaced and addressed rather than hidden.

How quickly do organisations typically see results?

Most inviol customers begin to see measurable reductions in high-severity events within the first 4 to 8 weeks. The initial drop comes from workers becoming aware that events are being detected and from early coaching conversations. Sustained reduction builds over the following months as coaching becomes embedded in daily operations. Full maturity typically takes 3 to 6 months.

Can inviol's coaching data help with regulatory compliance?

Yes. Under New Zealand's HSWA 2015, PCBUs must identify and manage workplace risks so far as is reasonably practicable. A system that automatically detects safety events, feeds them into documented coaching workflows, and tracks risk reduction provides strong evidence that your organisation is proactively meeting its duty of care. Similar obligations apply under Australian WHS legislation and OSHA in the United States.