How accurate is the risk tier? We measured it
We rebuilt the HaulProof risk tier for 379,944 US for-hire carriers as it would have looked on May 1, 2025, using only data dated before that day, then followed what happened to them over the next 12 months.
| Tier on May 1, 2025 | Carriers | Authority revoked | Out-of-service order | Liability insurance lapsed* |
|---|---|---|---|---|
| Low | 220,799 | 9.1% | 0.5% | 8.0% |
| Average | 117,661 | 17.2% | 4.9% | 14.4% |
| Medium | 27,922 | 28.0% | 5.0% | 17.5% |
| High | 13,562 | 67.3% | 10.9% | 48.9% |
| High vs Low | 7.4× | 20.7× | 6.1× |
Within 12 months of May 1, 2025. * Among the 362,063 carriers with liability (BIPD) insurance in force that day; not replaced within 30 days.
What it means
- Carriers rated High had their operating authority involuntarily revoked 67% of the time, against 9% for carriers rated Low, and received federal out-of-service orders 21 times as often (10.9% vs 0.5%).
- Many High-rated carriers already had a revocation or insurance-cancellation notice on file. Leaving those out, High was still about 4 times as likely as Low to be revoked.
- The tier ranks risk; it does not catch everything. Revocations are common, and about 70% of them happened to carriers rated Low or Average, simply because those carriers are the large majority. Use the tier to decide where to look harder, not instead of checking.
How we measured it
- Data: public FMCSA records (authority, insurance, inspections, crashes, out-of-service orders). Population: 379,944 US carriers holding active for-hire authority on May 1, 2025.
- The tier was rebuilt only from signals that can be dated (15 of the 26 weighted signals and 3 hard rules). Signals that exist only as today's value (USDOT status, chameleon links, address, truck count, MCS-150 date) were left out so the test couldn't peek at the future.
- A second start date (October 1, 2025) gives similar results. Canadian provincial ratings are not part of this test.
- FMCSA moved its revocation and insurance records to a new system in May 2026, so this test stops there. We re-run it monthly and whenever the tier's method changes.
Backtest run October 26, 2026. The probabilities shown on carrier records are not calibrated: read the tier as a ranking, not as a percentage forecast.