What the Sensors Actually Measure
Every driver score is built from a handful of physical measurements. Knowing which, and how they are derived, explains most of what goes wrong later.
Measurement · Explainer
A driver behaviour system produces a number. That number comes from a small set of measurements, and each has a known failure mode.
The measurements
Acceleration in three axes, from an accelerometer in the device or the vehicle.
Speed, either read from the vehicle's own systems or derived from successive positions.
Position and heading.
Engine and vehicle data on connected units: throttle position, brake pedal status, cruise control, seatbelt status.
Video, on camera-equipped systems, which is a different category and has its own notes.
Everything else is derived from these.
Derived against measured
A device reading speed from the vehicle bus is measuring it.
A device calculating speed from position over time is estimating it, and the error grows as the sampling interval grows.
Harsh braking derived from an accelerometer is a measurement of the device's movement, which is the vehicle's movement only if the device is mounted rigidly.
This distinction runs through everything here. Ask which, per parameter, before relying on any of it.
Why mounting matters
A dongle in a loose OBD socket moves independently of the vehicle.
Which produces acceleration events the vehicle never experienced, especially over potholes.
A hardwired unit with a fixed mount does not have this problem and costs an installation.
Fleets reporting implausible event rates on some vehicles and not others are usually seeing a mounting difference rather than a driving difference.
Calibration
Devices need to learn which way is forward.
Most do this automatically over the first journeys, and get it wrong if the first journeys are unusual.
A miscalibrated device reports braking as cornering and produces a driver whose profile makes no sense.
Check the first week's events per vehicle and re-calibrate the outliers before anyone sees a score.
What none of them measure
Attention, intention or skill.
Whether a manoeuvre was necessary.
What the traffic was doing.
The whole subject of this collection is the distance between the measurements above and the conclusion people want to draw, and most of the errors are in that gap.
What this collection covers
How events are triggered and what thresholds do.
How scores are built, and the statistics that make them misleading.
Video review, coaching and what the evidence says about effectiveness.
Fatigue and distraction, which are different problems from the ones scoring addresses.
And where the programme is a safety intervention rather than a surveillance one, which is a design choice rather than a claim.
Who this is for
Three audiences with different reasons to be here.
Whoever has been asked to deploy one of these and needs to know what it will actually detect.
Whoever will review the clips and hold the conversations, which is the job nobody budgets and the one the programme depends on.
And the drivers, whose cooperation determines whether the data describes anything — which is why their view of the system gets its own notes rather than a paragraph.
Turn the boundary into a test
For a concrete product reference, time tracking software for transport teams can help turn this boundary into a test. Verify current behaviour with representative roles and record what reaches the destination.
Independent reference
For a thematic point of reference, see the National Highway Traffic Safety Administration. Use the current material as an outside check before turning the principle into a system rule.