Thresholds and What They Decide

Every harsh braking event in your data exists because somebody chose a number. Almost always that somebody was the vendor.

How a trigger works

A threshold in gravitational units, typically somewhere between a quarter and half a g for braking.

Held for a minimum duration, so a single spike does not register.

And frequently a speed floor, so manoeuvring at walking pace is excluded.

Three parameters. Change any and the event count changes by a factor.

Why defaults are wrong for most fleets

Defaults are tuned for a passenger car.

A laden three-and-a-half tonne van decelerates differently and its suspension transmits differently to the device.

An unladen version of the same van is different again, which means the same driver on the same round scores differently depending on the load.

And a bus, a tractor unit and a refrigerated body are three more cases.

What happens if you do not adjust them

Event volumes that make the system unusable: hundreds a week, which nobody reviews.

Or so few that it detects nothing.

Both look like a fleet characteristic and are a settings choice.

Which is why the first month should be silent logging, with the distribution examined before anyone is shown a number.

Setting them from your own data

Log for a month with alerts off.

Plot the distribution of decelerations by vehicle type.

Set the threshold where the tail begins — the point beyond which events are genuinely unusual for that vehicle on that work.

Different thresholds per vehicle type, which some products support and many do not. That is a procurement question, not a detail.

The comparability problem this creates

Once thresholds differ by vehicle type, event counts are not comparable across types.

Which is correct and inconvenient, and means any score has to be normalised or confined within a type.

A fleet comparing a van driver against a truck driver on raw event counts is comparing the thresholds, and it will be challenged by the first person who understands that.

The check

Event rate per hundred miles, by vehicle type, by driver.

A vehicle type far from the others is a threshold problem.

A single vehicle far from its type is a mounting or calibration problem.

Only what remains is about driving, and it is a smaller residue than the raw counts suggest.

Log silently for a month first

The single most important sequencing decision here.

Alerts into a log, not to people.

Plot the distribution by vehicle type, then set thresholds where the tail begins.

A first week of implausible event volumes destroys the programme's credibility permanently, and it is entirely avoidable by waiting.

A practical configuration prompt

During configuration, use employee productivity reporting for mobile workforces to prompt questions about purpose, ownership and output. Treat the page as a starting point and document each assumption.

Independent reference

For a thematic point of reference, see the Federal Motor Carrier Safety Administration. It provides useful context beyond supplier documentation and should be checked in its current form.