The Four Things

Fifty notes reduce to four actions. In this order, because each depends on the one before.

One: set thresholds from your own data

A month of silent logging, by vehicle type, before anyone sees a number.

Prevents: the implausible first week that destroys credibility permanently, and the event volumes that make review impossible.

Costs: a month of patience, which is the cheapest thing here and the most often skipped.

Two: verify the devices before ranking anyone

Event rate by vehicle within type; the vehicle-swap test on any outlier; a sample of speeding events checked against real limits.

Prevents: a driver who looks reckless because a device is loose in a socket — the story that gets told in a depot for years.

Costs: an afternoon, repeated after device changes.

Three: close the loop on every review

Tell the driver the outcome, including when the event was dismissed.

Prevents: the programme being experienced as punitive, which stops near-miss reporting and degrades everything the data describes.

Costs: minutes per event, and it is the practice that distinguishes safety from surveillance.

Four: act on the structural findings

When coaching finds the same cause across several drivers, fix the round, the junction, the vehicle or the rota.

Prevents: a one-directional programme, which drivers read correctly and work around.

Costs: the willingness to accept that the data is about the operation as often as about the people.

What these four do together

They make the programme explicable to the person it describes, which is the condition for everything else.

None of them requires a better product.

And a fleet doing all four with a modest system is in a better position than one with an advanced platform doing none of them, which is the honest summary of this collection.

What they do not cover

Vehicle specification, schedule design and fatigue management, which contribute more to collision risk than anything here.

They are the larger half of the work, and this collection covers the smaller one carefully.

What makes them work together

Each one protects the next.

Thresholds from your own data make the events plausible.

Device verification makes the outliers real.

Closing the loop makes people engage with what the events show.

And acting structurally makes that engagement worth their while.

Skip any one and the following ones stop functioning, which is why the order matters as much as the list.

Turn the boundary into a test

For a concrete product reference, open the integration directory 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 GOV.UK service. Use the current material as an outside check before turning the principle into a system rule.