Glossary

Event trigger — the threshold, duration and speed floor that together decide whether a manoeuvre becomes a recorded event. Three numbers, usually vendor defaults, that determine the entire output.

Exposure — the denominator that turns event counts into a rate. Distance is the usual choice and the wrong one for urban work, where opportunities to brake scale with drops rather than miles.

False trigger — an event recorded that was not a driving event: a pothole, a kerb, a loose device. The proportion of these is the single best indicator of whether a video programme will survive.

Incident hold — a mechanism preserving data for a specific vehicle and window beyond routine deletion. Triggered at report, not when a claim arrives.

Near miss — a collision that did not happen. The cheapest safety learning available, and the thing a scoring system suppresses by penalising the avoidance manoeuvre.

Regression to the mean — the reason drivers coached for a bad month improve the next, whether or not the coaching did anything. The reason a comparison group is necessary.

Serious threshold — the narrow, published list of conduct that goes straight to a disciplinary process rather than to coaching. Defined in advance or defined under pressure.

Silent logging — running the system with alerts off while thresholds are set from the resulting distribution. One month, and it is the step that decides whether the programme is usable.

Terms used loosely by vendors

"Risk score" is a weighted sum divided by exposure. Ask for the formula; if it is proprietary, it cannot be explained to the person it describes.

"AI-powered detection" usually means a classifier over the same accelerometer and video data. Ask what the false positive rate is on your vehicles.

"Proven to reduce collisions" is generally a before-and-after comparison with no control group.

"Driver engagement" frequently means gamification, whose durable effects are weakly evidenced.

How to read these notes

Start with measurement if a system is already producing numbers nobody trusts.

Start with video if cameras are being proposed.

Start with the risk section if the actual goal is fewer collisions, because the strongest measures there are not behavioural.

Start with the legal section before deploying anything.

Nothing here is legal advice, and the notes say so where it matters.

On the absence of numbers

How to treat any figure in this field.

No effectiveness percentages are quoted, because the published ones are almost all before-and-after comparisons without control groups.

No threshold values, because the right ones come from your own distribution.

No false positive rates, because they depend on your vehicles and your roads.

Every number worth having here is one you produce, and the notes say how to produce each.

A practical configuration prompt

During configuration, use review the data-flow page 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. Consult the source directly because technical, operational and legal details change.