Product

Following distance is the collision cause your telematics cannot see.

Rear-end collisions are the most common at-fault event in commercial fleets, and following too closely is the most preventable cause. Time headway is measurable from forward video alone. That is what we build first.

For fleets

What a safety manager gets.

Headway exposure, per driver

Not an event count — the share of driving time spent under two seconds of gap, normalised by distance. A driver who spends 30% of motorway time tailgating is a different risk from one who does it twice a month, and event counting cannot tell them apart.

Attributed harsh braking

Each hard brake labelled by what preceded it: closing on a lead vehicle, or an external cause. This turns the noisiest metric in fleet safety into one a driver can be coached on without arguing about it.

Coaching evidence

The clip, the headway trace and the moment of the event. Safety conversations fail when they are a number against a driver's recollection; they work when both parties are looking at the same video.

Your own calibration

Thresholds fitted to your collision record rather than a generic default. Given the sensitivity we publish on the Evidence page, a borrowed threshold is close to meaningless.

Build order

What exists, what is next, what is blocked.

Published honestly because it is the first thing a technical evaluator establishes anyway.

Validated on real data

Kinematic event layer

Harsh braking, acceleration and cornering from speed and curvature. Debounced, exposure-normalised, trip-boundary safe. This is the substrate — it is deliberately not the differentiator, because an IMU produces the same thing.

In build

Time headway

Seconds of gap to the lead vehicle, continuously, from forward video. The wedge: high value to fleets, unavailable from accelerometers, and directly linked to the most common claim type.

Planned

Event attribution & lane discipline

Joining each kinematic event to the scene that caused it, plus lane-keeping consistency and unsignalled weaving — behaviours associated with distraction and invisible to an accelerometer.

Blocked on data

Risk score calibration

Fitting weights against collision outcomes. A fleet's own accident record unblocks this; a carrier's claims data unblocks the insurance version of it. No amount of engineering substitutes for either.

Who this is for

  • Commercial fleets running forward-facing cameras who want collision causes, not event counts.
  • Telematics platforms and dashcam makers who want camera-derived signals without building the perception stack.
  • Insurance carriers running usage-based programmes that have reached the limit of accelerometer data.