Every deployment here started on cameras the site already owned. These are the outcomes the safety teams measured afterwards, each one attributed to the deployment it came from.
Each case study covers the risk the site was carrying, what the detection layer changed, and who acted on what it found.
Public housing estate construction, watched around the clock through the CCTV already on site.
Forklift and pedestrian conflict across dock and aisle cameras on the warehouse floor.
Deep shafts, tunnel headings and launch chambers across a multi-contractor civil programme, monitored on cameras already installed.
All case studies: HDB construction safety monitoring · Global logistics warehouse safety monitoring.
Each site measures what its own risk register cares about. These are the figures the deployments on this page reported.
A proof of concept on a handful of your existing cameras, tuned to the risks that matter on your site, live within 24 hours of camera access.
Every figure on this page came off a live site, not a benchmark reel. Each percentage compares six months before the detection layer went live against the six months after, on the same site. This is how the numbers were produced, so you can judge them the way your own risk register would.
Detection thresholds are tuned on real footage from the site before anything is counted. On the civil programme, underground scenes needed several weeks of real footage before the thresholds were usable, and full coverage took two months. Numbers reported before that tuning would not have meant anything.
The percentage changes compare a six month period before the detection layer went live against the six months that followed on the same site. A window that length matters on a construction programme, because a shorter one would just be measuring the weather and the build stage rather than the intervention.
The observation figures compare continuous camera coverage against the manual patrol routine the site was already running. That is where 8x more observations than manual patrols comes from on the HDB estate build: the same site, the same risks, a different amount of looking.
Each detection is stored as a timestamped clip, so an event exists as evidence rather than as an anecdote. The counts on this page are drawn from those records, which means they can be re-checked rather than taken on trust.
Detection accuracy of 85%+ is verified on live footage from the deployment itself, per risk type, rather than quoted from a model benchmark. Accuracy is re-checked as a site changes, because an estate build at handover does not look like the same site at structure stage.
Three deployments, three sectors, one layer of the platform. Being specific about the edge of the evidence is part of the evidence.
The detection layer, running on cameras the sites already owned, across estate construction, warehouse logistics and underground civil works. Continuous coverage, alerts in under a second, and evidence retained per event.
Autonomous patrols, self-closing corrective actions, continuous risk scoring and generated briefings build on top of detection. They are part of the platform, and they are not what these three deployments measured.
Camera placement, lighting and scene type change what is achievable. Detection trained on rooftop and facade footage read underground scenes badly until it was retuned on site. Expect a tuning period, not a switch.
The questions that come up most often when a safety team is comparing platforms.
No. Any camera with an RTSP stream connects, and standard deployments need no new hardware and no new cabling. All three deployments on this page ran on cameras the site already owned.
A standard site is live within 24 hours of camera access. Tuning to a specific site, particularly unusual scenes such as underground works, takes longer and is worth doing before anyone counts results.
Processing runs on site through edge or on-premise deployment. Footage does not need to leave the site to be analysed, and detection keeps running if the uplink drops.
85%+ verified detection accuracy, measured on live footage from the deployment itself and tuned per site, per zone and per shift rather than quoted from a benchmark.
WhatsApp, Microsoft Teams or email in under a second, routed through your escalation chain, with the video evidence attached so nobody has to hunt for the footage.
Yes. Run it on your own recorded footage first and see the detections on your own site before spending anything. How pricing works sets out the four things that drive cost.
Quoted with permission from safety leads on live deployments. Attributed by role at their request.
Invigilo gives us much better visibility on what is happening on site, so we can pick up safety issues earlier and act before they become bigger problems.Safety Manager
It helps our team see risks that would otherwise be easy to miss, and gives us the information we need to respond much faster.Safety Manager
With Invigilo, we have a much clearer picture of what is happening across the site and where we need to focus our attention.Safety Director
Invigilo has helped us move from reacting to incidents to spotting risks earlier and dealing with them before something happens.Safety Director