Beta · trial-v1
Trial record reference score
A 0–100 summary of a few official public records, published as a trial so it can be checked before wider use. It is not an FMCSA rating, not a safety guarantee, not a prediction and not a load/no-load decision. VAX does not rank or sort carriers by it.
Inputs and weights
Operating authority · 25
100 when the current official authority file lists an Active authority; 0 only for an explicit revocation (shown as a separate red stop). Pending, inactive, withdrawn or no row → not determinable. A missing MC is never read as “no authority” (exempt, intrastate and private carriers differ).
BI&PD on file · 15
100 when an Active authority shows BI&PD on file at or above its required amount. Anything else → not determinable; a missing record is never read as “no insurance”.
Driver OOS rate · 30
Driver-OOS inspections ÷ driver-examining inspections (levels I, II, III, VI), 24 months. Needs at least 10.
Vehicle OOS rate · 30
Vehicle-OOS inspections ÷ vehicle-examining inspections (levels I, II, V, VI), 24 months. Needs at least 10.
OOS rate → points (fixed, not calibrated)
| OOS rate | Points |
|---|---|
| 0% | 100 |
| 10% | 90 |
| 20% | 75 |
| 30% | 55 |
| 40% | 30 |
| 50% | 10 |
| > 50% | 10 |
Points are interpolated linearly between these knots, so there is no cliff at a threshold (e.g. 15% → 82.5). These knots are VAX's v1 choice — not a national benchmark and not yet validated.
Calibration status & sample coverage
Not calibrated
Generated 2026-10-08 20:40 PT (America/Los_Angeles) from real records (latest background snapshot; earlier snapshot kept if a run fails) — not a national benchmark. Same rules as the score: 24-month window from 2024-10-09, driver levels I/II/III/VI and vehicle levels I/II/V/VI, at least 10 inspections per denominator, inspection source metadata updated 2026-10-08 PT (within 90 days).
- Population: Systematic sample: USDOT ending in 0 among carriers with an imported Active authority row; 11187 carriers currently selected.
- Today's scan pass 4 (started 2026-10-08 00:40 PT): 1280 carriers scanned so far — statistics use only these. Scan times 2026-10-08 17:40 PT – 2026-10-08 20:40 PT. 6791 sample rows are kept in the cache, but older rows from earlier passes are not counted. Pass not complete yet.
- Excluded: 0 failed/truncated, 0 stale source.
- Partial statistics — the authority baseline and/or the sample scan has not completed a full pass.
| Measure | Driver | Vehicle |
|---|---|---|
| Carriers with ≥10 inspections | 2 | 2 |
| Zero inspections (not determinable) | 1106 | 1160 |
| 1–9 inspections (small sample) | 172 | 118 |
| Median denominator (eligible) | 18 | 14.5 |
| OOS rate P25 / median / P75 (eligible) | 36.5% / 39.6% / 42.7% | 26.8% / 37.9% / 48.9% |
| OOS 0%–<10% | 0 | 0 |
| OOS 10%–<20% | 0 | 1 |
| OOS 20%–<30% | 0 | 0 |
| OOS 30%–<40% | 1 | 0 |
| OOS 40%–<50% | 1 | 0 |
| OOS 50%+ | 0 | 1 |
Both denominators eligible: 2 of 1280 (99.8% not determinable on safety). 0 carriers excluded for failed or truncated reads. 1000 carriers with both denominators is only the trigger for an engineering review — not statistical sufficiency and not a sign calibration is done. Even after it is reached, the sample still needs representativeness, bias and stability review before anything changes. Until then the v1 knots stay unchanged, no national percentiles are published and no comparison with an official benchmark is shown, because none comparable was obtained.
When no total is shown
- Any input is not determinable: the page shows “Insufficient data” plus coverage (weights that were determinable). Missing parts are never filled with 100 or 0.
- Fewer than 10 driver or vehicle inspections, a truncated pull, or a source last updated 90+ days ago (or with no date).
- An explicit official revocation: shown as a separate red stop that a score can never hide.
Not scored
Crash counts are shown but not scored — public crash records carry no fault or mileage exposure. Registration changes are review information only. New companies get no penalty for having little history; they simply show insufficient data. The census registration status code is not used.
Technical fields (for review)
- authority: inys-ebih op_auth_status = "Active"; yu5v-wbh6 latest reason ~ /revoked/ AND status = "Inactive" → hard stop
- insurance: inys-ebih bipd_file ≥ min_cov_amount (min_cov_amount > 0)
- driverOos: fx4q-ay7w insp_level_id ∈ {1,2,3,6}; driver_oos_total > 0; n ≥ 10
- vehicleOos: fx4q-ay7w insp_level_id ∈ {1,2,5,6}; vehicle_oos_total > 0; n ≥ 10
- stale: rowsUpdatedAt age ≥ 90 d or missing; truncated: ≥ 1000 rows
- total = round(Σ points × weight / 100) only when all four are determinable and no hard stop