Carrier411 Guidecarrier411.fun

Home › BASIC percentile scores

BASIC percentile scores

Percentiles rank a carrier against its peers. They are not grades, and they are not comparable across carriers of different sizes.

The seven categories

The Safety Measurement System sorts roadside inspection violations into seven behaviour categories, known as BASICs:

How a percentile is calculated

Violations are weighted by severity and by how recently they occurred, then normalised against exposure — roughly, how much the carrier is on the road and how often it is inspected. The result is compared against a peer group of carriers with a similar volume of inspections, and expressed as a percentile from 0 to 100.

A percentile of 75 does not mean the carrier failed three quarters of something. It means that within its peer group, three quarters of carriers had a better measure in that category.

Two carriers with identical raw violation counts can sit in completely different percentiles, because the peer group and the exposure denominator differ. Comparing raw counts between a twelve-truck fleet and a four-hundred-truck fleet tells you almost nothing.

Minimum data requirements

A category only produces a percentile once there is enough data to support one. Carriers below the threshold show no percentile in that BASIC even where violations exist. This is why a carrier with a handful of inspections often appears to have a nearly blank record — not because it is clean, but because the sample is too small to rank.

The corollary is that small carriers can swing dramatically. One inspection with several violations can move a low-volume carrier a long way up a percentile, and the same carrier may fall back just as sharply as the violation ages out of the window.

The twenty-four month window

Data rolls off after twenty-four months. Percentiles therefore change without any new event occurring, simply because older violations expire. A rising percentile means recent activity is worse than what it replaced; a falling one may mean improvement, or may mean an old cluster of violations has aged out.

Crash Indicator has a known limitation

The Crash Indicator counts reportable crashes without, in the public presentation, distinguishing who was at fault. A carrier struck from behind at a traffic light and a carrier that caused the same collision are recorded alike. A review process exists to flag not-at-fault crashes, but it is not automatic and not universally used.

This is worth stating plainly whenever crash data informs a decision.