Methodology Note

Grading Owner-Operators

The single-power-unit credibility problem — what the data shows, and how we treat it.

44%
of gradeable carriers operate a single power unit
0.11
ranking power (Gini) on single-unit carriers — vs 0.64 on 100+ unit fleets
~1 in 6
carriers record any reportable crash in a year

A single-power-unit carrier presents about one power-unit-year of exposure in any twelve-month window. In a book where reportable crashes touch roughly one carrier in six per year, that is simply too little exposure to separate a genuinely safe operator from a lucky one. Credibility theory is unambiguous here: with one unit-year on the board, the credibility assigned to the carrier's own record is a few percent, and almost all of the weight falls to the size-cohort mean. A clean year for a one-truck operator earns the benefit of the doubt — it is not, by itself, evidence of safety.

The data bears this out. Measured out-of-time — one year of record used to rank carriers, the following year's crash burden used to judge the ranking — the model separates single-unit carriers with a Gini of about 0.11, roughly a sixth of the 0.64 it achieves on fleets of a hundred or more power units. Discrimination climbs monotonically with fleet size for exactly this reason: more trucks mean more exposure, and more exposure means a record the model can actually read. Single-unit carriers, taken together, go on to realize forward crash burden indistinguishable from the broad middle of the book. The signal that makes the score useful on real fleets is, for the most part, not present on owner-operators.

So we decline to overstate what we cannot measure. Single-power-unit carriers are capped at a “Strong” grade — they can earn a solid mark on a real inspection record, but never our top “Excellent” tier, which one unit-year of crash exposure cannot support — and every one is flagged as limited-confidence. The cap is one-sided: a single-unit carrier with a materially adverse record (a reportable crash, an out-of-service pattern) still grades down on its own merits, because the downside is the part the data does support. A carrier with no inspection or crash history at all is not graded; we mark it “Not Rated” rather than invent a number from thin air.

Because owner-operators make up nearly half of the carriers we can rate yet realize forward burden no different from the broad middle, they tend to crowd a portfolio view without informing it. We exclude them by default and let you switch them back on with a single toggle — so you can read the gradient among the fleets where it is real, and bring owner-operators back when one is the risk actually in front of you. Read a single-unit grade for what it is: a starting point set mostly by peers of similar size, not a verdict the carrier's own record has yet earned.

Beyond a single truck. The same credibility problem extends to every fleet under ten power units: their realized forward crash burden is statistically flat, so they are held out of the default ranked book — yet they stay fully searchable, so you can still pull a small hauler's record when vetting it for a brokered load. Why we surface the available data without fabricating a fine ranking is set out in the Small Fleets: Searchable, Not Rankable section of the FRED Scoring white paper.

Figures from the most recent out-of-time scoring run. See the full methodology for how grades are built.