How to spot forklift near misses in a warehouse

Incidents get written up. Near misses mostly don't, because by the time anyone thinks to mention it the truck has moved on and the moment is gone. That's the gap that matters most in a warehouse: the reversing truck a picker stepped clear of, the blind corner where the same shortcut gets taken every shift, all of it happening on camera and none of it ending up anywhere. Workplace injury and ill health costs UK employers £20.7 billion a year (HSE, 2022/23), and a good share of what leads up to the serious incidents is exactly this kind of close call that nobody happened to be watching for.

Why near misses go unrecorded

A near miss, almost by definition, resolves itself: nobody's hurt, the shift carries on, and the person who stepped clear has a dozen other things to do before they'd think to file a report. Reviewing hours of aisle footage on the off-chance of catching one is not a realistic use of anyone's time, so the pattern, the same corner, the same route, the same time of day, stays invisible until it isn't.

What the check actually flags

Forklift / Pedestrian Proximity is an alerting check that judges a window of recent frames as they arrive, so a close call reads as a sequence of movement rather than one still image: a forklift reversing without a spotter, a picker walking close behind it, someone cutting through the truck route as a shortcut. There are no tags on people and no sensors on the trucks. It works from what the camera already sees, which means it also works on recorded footage, so a week of yard or aisle video can be run through it after the fact to see what's actually been happening.

1. 01

   Point a camera at the aisle, dock or yard

   Wherever forklifts and pedestrians share space without a fixed barrier between them: an existing RTSP feed, an upload, or a phone camera on a temporary route all work.

2. 02

   Turn on Forklift / Pedestrian Proximity for that camera

   Checks run per camera, so it can watch the truck routes specifically without touching cameras elsewhere on site.

3. 03

   Run it over a week of existing footage first

   Recorded video goes through the same pipeline as live, so you can see what it would have flagged, and where, before deciding where to put a live camera.

4. 04

   Use the flagged clips for the toolbox talk

   An actual clip of the close call at the blind corner lands harder than a generic safety poster, and it points the conversation at a specific place and time rather than a general reminder.

A flag for review, not a safety system

This is worth being direct about: it's a flag for human review, not an autonomous safety system, and it doesn't replace site rules, pedestrian walkways, spotters or forklift operator training. Like any AI judgment, it can miss things and it can flag things that turn out to be nothing; every alert is a prompt for a person to look at a clip, not a verdict. It's genuinely useful evidence after a near-miss report comes in too, not only for catching things live: if someone mentions a close call from earlier in the shift, the footage is there to go back and look at.

What it costs

The Store prices Forklift / Pedestrian Proximity from your own usage before you buy, shown as the extra it would add per month at your camera count and hours watched. It sits on a subscription also priced from your own usage, floored at $9 a month, with the full derivation shown on screen at signup. The checks page has the full list, including PPE Compliance, which most warehouses running this check end up watching alongside it.

Does this replace pedestrian walkways or spotters?

No. It's a camera-based flag for human review, not a substitute for site rules, marked walkways, spotters or forklift operator training. It sits alongside those, it doesn't stand in for them.

Can it see near misses that already happened?

Yes. It runs on recorded footage the same way it runs live, so a week of existing aisle or yard video can be checked for the close calls nobody reported at the time.

Does it stop the forklift or warn the driver?

No. It has no connection to the truck itself. An alert reaches a screen, an email, or a webhook for a person to review; there's no in-cab warning or physical intervention.

Does it track individual workers?

No. There's no facial recognition and no tracking of anyone once they leave the frame. Alerts describe what's visible in the moment, an outfit, a position, not an identity.