How These Programmes Fail

Driver monitoring fails in recognisable ways, and almost none of them is the hardware.

Default thresholds left in place

Symptom: implausible event volumes, or none at all, varying by vehicle type.

Correction: a month of silent logging and thresholds set per type from your own distribution.

Scores published before calibration

Symptom: a driver who looks reckless because their device is loose in the socket.

Correction: event rate by vehicle checked first; the vehicle-swap test where the data is ambiguous.

Ranking across route types

Symptom: the urban depot permanently at the bottom, and everyone knowing why.

Correction: compare within type, or against a driver's own history.

Coaching that is really discipline

Symptom: drivers treating a coaching invitation as a warning, and saying as little as possible.

Correction: separate the people, define the serious threshold narrowly, and publish it.

No feedback after review

Symptom: drivers hear from the system only when something is wrong.

Correction: tell them the outcome including dismissals, which is the single practice that distinguishes safety from surveillance.

A bonus attached to the score

Symptom: near-miss reports fall, disputes rise, and the measure gets optimised.

Correction: remove it. Recognise collision-free periods and reporting instead.

Cab camera enabled because it was included

Symptom: a consultation you did not expect, or a resignation you did not expect.

Correction: the proportionality sentence, finished honestly, before anything is switched on.

Structural causes coached individually

Symptom: the same finding across several drivers, addressed one at a time, forever.

Correction: treat a repeated cause as a route, schedule or vehicle problem.

Nobody owns the review

Symptom: clips accumulating, reviewed in batches weeks later, dismissed without watching.

Correction: a named reviewer with allocated time, and a time-to-review measure.

The common thread

Almost every failure is a configuration or a process decision rather than a technology problem, which is why changing provider fixes none of them.

The recovery sequence

For a programme already resented.

Publish the false trigger rate.

Remove any bonus link.

Start closing the loop on every review, including dismissals.

Fix one structural finding visibly and say the data found it.

One quarter of that, and a programme people work around becomes one they use.

A practical configuration prompt

During configuration, use this documented connection 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. This established source is a useful independent reference for the issue discussed here.