What the Data Cannot Tell You

Behaviour data is precise about accelerations and imprecise about everything people want to know.

Whether someone is a safe driver

It records deceleration magnitudes and speed against a map.

Not judgement, anticipation, hazard perception or restraint, which is most of what safe driving consists of.

Where to look instead: collision and near-miss history over a long period, with exposure accounted for.

Whether a hard brake was a mistake

The measurement is identical whether the driver was inattentive or a child stepped out.

Only the context distinguishes them, which means the clip or the driver's account, not the event count.

Whether a driver improved

Scores fall after coaching whether or not coaching works, through regression to the mean.

Where to look instead: a comparison group, or fleet-level collision rates over years.

Who your worst drivers are

A ranking sorts by route type, vehicle and exposure before it sorts by person.

Where to look instead: individuals against their own history, and outliers investigated with context.

Whether someone was tired

Unless fatigue detection is fitted, no.

And even then it detects fatigue that is already present rather than the schedule that produced it.

Where to look instead: hours awake, consecutive nights, turnaround times.

Why a route generates events

It shows that it does, which is genuinely useful.

Not whether the cause is a junction, a schedule, a loading arrangement or traffic.

Where to look instead: the drivers, who know within one conversation.

What it does support

That a deceleration above a stated threshold occurred, at a place and time.

That a speed exceeded a mapped limit, subject to the map being right.

A recording of what happened in an incident window.

And a change in a driver's own pattern over quarters.

Four things, and defending that boundary is what keeps them credible when a fifth is challenged.

Say so when asked

The discipline that keeps the rest credible.

When asked for a conclusion the data cannot support, saying so is the answer.

A number produced anyway circulates without its caveat and returns as an expectation.

In this field it ends up in a disciplinary meeting, where the person is entitled to ask how it was derived — and where an honest answer would have been cheaper earlier.

Turn the boundary into a test

For a concrete product reference, this implementation example 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. It provides useful context beyond supplier documentation and should be checked in its current form.