Methodology v6.6 · site revision i · 2026-07-27

How We Build Carrier Records from Federal FMCSA Data

Name: Safeny Carrier Safety Dataset
Creator: Safeny
License: https://creativecommons.org/publicdomain/zero/1.0/

Crash records don't determine fault; here's how we report them.

SHA a81bb40e… · built 2026-05-25T00:00:00Z · verify

Core Definitions

Defined terms (§1–§7)

§1BIPDFederally mandated liability coverage for interstate motor carriers

Bodily Injury and Property Damage insurance — the primary liability coverage federally required for interstate motor carriers carrying general freight. Minimum coverage levels specified by power-unit class and cargo type, with a $750,000 floor for general freight, rising to $1,000,000–$5,000,000 for oil and certain hazardous-materials classes per 49 CFR § 387.9.

Per 49 CFR § 387.5–.9 · Methodology v6.6 §4 (BIPD / self-insurance)

§2Self-Insurance (presumed)Presumed self-insurance, inferred from the FMCSA filing data pattern

FMCSA permits large motor carriers to self-insure their bodily-injury & property-damage (BIPD) exposure under 49 CFR § 387.309 instead of carrying a third-party policy. Safeny does NOT have per-carrier filing-form data, so self-insurance is PRESUMED from a data pattern — a large fleet (≥ 100 power units) with an active BIPD filing showing a $0 third-party liability limit. The authorization amount is privately maintained with FMCSA and is not visible in the public BIPD filing data; no specific filing form is asserted.

Per 49 CFR § 387.309 · Methodology v6.6 §4

§5EDSEvidence Density Score — categorical record-corroboration band (HIGH / MODERATE / LIMITED)

Evidence Density Score (EDS), a categorical band reflecting the corroboration depth of a carrier's federal regulatory record — how thoroughly its safety posture is documented across inspection history, crash-record consistency, and authority continuity (records discoverable under Fed. R. Civ. P. 26 / 34). Reported as HIGH (≥70), MODERATE (50–<70), or LIMITED (<50). The band currently informs factor-based ranking order; it is not displayed as a standalone label on carrier pages. Statistical observation only — not a prediction of liability, fault, settlement value, or any specific case outcome.

Per Methodology v6.6 §5 · Litigation record-corroboration framing

§7Jurisdiction TierCoarse A–D state grouping used only as neutral context and to cluster comparable carriers — not an outcome estimate

A coarse A–D grouping of U.S. states used solely as neutral jurisdictional context and to cluster comparable carriers (the 'Comparable fleet size and jurisdiction tier' panel). It reflects broad, publicly observed differences in states' civil-litigation environments. It is NOT a Safeny prediction or estimate of case value, recovery, settlement, or verdict; it is not a probability; and it is not an input to any carrier safety score, ranking, or the Evidence Density Score. Groupings: A — CA, IL, LA, FL, NY, NJ, PA; B — MO, MN, WA, MD, NM, NV; C — AZ, GA, OH, MI, WI, NC, VA, CO, OR and other national-baseline states; D — IN, KS, OK, AL, MS, SC, TN, WV, UT and Plains/Mountain states (any state not listed defaults to D). Within-state venue variance is not modeled — see Limitations §D.

Methodology v6.6 §6 (jurisdictional context only)

Public-safety analysis has always rested on one practice: the records the government collects are published, and anyone can examine them. This page documents how Safeny applies that practice to the federal government's own commercial-carrier safety records.

Every figure on this site is compiled from public FMCSA records. Any attorney, researcher, or carrier can reproduce the published figures from those records using the steps in Reproducibility. We document our data sources, definitions, and known limitations so that the compiled record can be examined under the same standard as any other public-data compilation.

Cohort Definitions

A cohort is the comparison group a carrier's percentile is measured against — a percentile is meaningless without one, because being in the 90th percentile of six-truck fleets and of thousand-truck fleets are different statements. Every SMS BASIC percentile rank partitions the carrier population by 5 × 4 = 20 cohort buckets: fleet-size × safety-event-count.

Fleet size (power_units)

micro 1–6 trucks
small 7–20 trucks
mid 21–100 trucks
large 100+ trucks
unknown NULL or 0

Safety event count (inspections_24mo)

0 no inspections in last 24 months
1–2 1 or 2 inspections
3–9 3 to 9 inspections
10+ 10 or more inspections

Both axes define the cohort partition for SMS BASIC percentile rank (since FMCSA's public files publish raw MEASURE values only — see Limitations). Carriers in the 0-events bucket receive cohort-baseline percentiles regardless of their MEASURE because there's no observable behavior to rank.

Data Sources

All inputs are public FMCSA records. Snapshot timestamps below auto-populate from the ingestion audit log; each refresh updates the corresponding row.

source	file	snapshot	rows in table
Motor Carrier Census	SMS_Input_-_Motor_Carrier_Census_Information	2026-07-16	2,108,302
Roadside Inspections	SMS_Input_-_Inspection	2026-07-16	5,866,963
Inspection Violations	SMS_Input_-_Violation	2026-07-16	6,278,384
Crash Reports	SMS_Input_-_Crash	2026-07-16	270,319
BASIC Measures	SMS_AB_PassProperty + SMS_C_PassProperty	2026-05-12	2,047,391
L&I Insurance Filings	actpendins_allwithhistory (catalog.data.gov, daily)	2026-07-16	370,246

Bulk FMCSA Source SMS Data files refresh monthly. The L&I bulk file (ActPendInsur) refreshes daily on catalog.data.gov. We re-ingest on the documented cadence; new monthly snapshots are loaded within ~7 days of FMCSA publication. National Out-of-Service (OOS) rate averages are computed across the full FMCSA roadside inspection corpus (5M+ inspections). Carriers with fewer than 5 inspections in the trailing 24 months get a small-sample notice instead of an above/below-average comparison — at that few inspections a single out-of-service order swings the rate too far to compare meaningfully. As of 2026-07-27 the suppression covers the comparison bar as well as the sentence: a bar drawn beside a national average makes the comparison whether or not a caption states it. The underlying fact — how many out-of-service orders across how many inspections — is still published, because it carries its own sample size. Full suppression list: Record Interpretation Modules.

Night-Driving Pattern Detection

Per-carrier nocturnal-clustering identification is performed under the Safeny Night-Driving Pattern Detection Framework, an in-house implementation. The night-driving pattern test is a two-sided one-sample binomial test against the national dark-condition baseline, computed via normal approximation (Abramowitz & Stegun erf form). Each carrier's ~2-year crash record is classified by Light_Condition_Desc and tested against the national dark-condition crash share. The page surfaces the share and a one-sample binomial p-value when the carrier has ≥5 documented crashes in the window; below 5, the section reports insufficient sample.

Attribution. This night-driving pattern-detection framework is a Safeny-developed module. The framework name is internal nomenclature, not a citation to an external paper; the underlying binomial test (via normal approximation) is a standard statistical procedure.

Data source

FMCSA SMS Crash File 2026-04 Snapshot, field Light_Condition_Desc. Approximately 259k national records; 99.6% non-null on the field. Source values fall into eight categories: Daylight, Dawn, Dusk, Dark - Lighted, Dark - Not Lighted, Dark - Unknown Roadway Lighting, Unknown, Other.

Night definition

Night aggregates the three Dark-* categories: Dark - Not Lighted, Dark - Lighted, Dark - Unknown Roadway Lighting. National combined share: 25.7% (the binomial baseline used in the per-carrier test).

Excluded from night: Dawn (2.4% national) and Dusk (1.4%). The source file records these as their own categories, distinct from the three Dark-* values: they describe a brightness transition rather than sustained darkness. They are reported separately and are not collapsed into the night aggregate.

Baseline

The 25.7% baseline is computed directly from the source distribution in the same dataset we use for per-carrier measurement. The dataset is both the measurement source and the comparison reference — internal consistency by construction. No external estimate is imported.

Significance test

One-sample binomial against the 25.7% baseline, two-sided. The p-value is computed via normal approximation with Postgres's erf() (Abramowitz & Stegun erf-form of the standard normal CDF). Sufficient fidelity for decision-support; not a publication-grade significance test, particularly for the smallest qualifying samples.

z \;=\; \frac{\hat{p} \;-\; p_0}{\sqrt{\dfrac{p_0\,(1-p_0)}{n}}}, \quad p_0 = 0.257

p\text{-value} \;=\; 1 \;-\; \operatorname{erf}\!\Bigl(\tfrac{|z|}{\sqrt{2}}\Bigr)

Sample-size gate

Carriers with fewer than 5 documented crashes in the ~2-year window get an insufficient sample message instead of a share + p-value pair. The 5-crash floor is conservative — at smaller sample sizes the normal approximation degrades and an apparent elevation is more likely noise than signal.

Regulatory connection

49 CFR § 392.14 requires extreme caution when hazardous conditions (including reduced visibility) adversely affect visibility or traction. The federal Hours-of- Service regime at 49 CFR § 395 limits consecutive driving hours and mandates rest periods; sustained-darkness over-representation in a carrier's crash record can correlate with fatigue-window operations.

Limitations

Long-haul carriers may have legitimately higher dark-driving exposure because their operations type concentrates over overnight hauls. The exposure-type confound is not corrected for here. The page displays deviation and significance together precisely so the reader can judge whether elevation reflects an operational pattern or a safety pattern — this is an investigative signal, not a fault determination.

Self-Insurance Detection (presumed)

FMCSA permits large motor carriers to self-insure their bodily injury & property damage (BIPD) exposure rather than carry a third-party policy. The regulatory hook is 49 CFR § 387.309 (Self-insurance authorization for motor carriers and freight forwarders). Self-insurers must demonstrate financial capacity to FMCSA on an ongoing basis. Safeny does not have per-carrier filing-form data, so no specific form is asserted.

Why the public dataset looks like "$0 coverage"

In the FMCSA L&I bulk feed, self-insured BIPD filings appear with a third-party liability_limit of $0. The carrier is not without coverage — they have regulatory authorization to satisfy claims from their own balance sheet — but the bond amount underwriting that authorization is not exposed in the public dataset.

Detection rule

A carrier is presumed self-insured when all three conditions hold against the most recent ingest:

Fleet size power_units ≥ 100 — floor that distinguishes large self-insurers from the residual multi-subsidiary filing pattern.
An active BIPD filing on record (coverage_type starts with BIPD, is_active = true).
The filing's liability_limit is $0 or NULL.

The rule is implemented as a Postgres function public.apply_self_insurance_detection() invoked after each monthly recompute so the classification is refreshed on each ingest cycle.

What changes on the carrier page

confidence_level = 4 (the regulatory authorization is itself a high-confidence signal).
underinsured_flag = FALSE — self-insurance is not under-coverage.
Coverage Gap is not computed; the public dataset does not expose the bond amount underwriting the authorization.
Insurance section label reads Self-insured (presumed).

Sample size + limitations

Approximately 92 carriers match the detection rule on the current snapshot, including FedEx Corp, Walmart Transportation, Old Dominion, Schneider National, Estes Express, XPO Logistics, Penske Logistics, and the FedEx Freight Inc operating subsidiary. The count refreshes on each ingest cycle as filings change.

The (presumed) qualifier is intentional: we infer the regulatory status from the data pattern (zero-limit BIPD on a large fleet), not from a direct FMCSA self-insurance roster call. A small residual set of multi-subsidiary carriers (parent corporation files BIPD under a different USDOT) may still produce the zero-limit pattern at fleet sizes below the 100-PU floor; those carriers stay outside the self-insured set and continue to surface the multi-subsidiary narrative from v6.4.1.

Evidence Density Score (EDS)

The Evidence Density Score reflects the corroboration depth of a carrier's federal regulatory record — the extent to which its safety posture is documented across multiple, mutually-reinforcing primary FMCSA sources. It draws on three record dimensions: inspection history, crash-record consistency, and authority continuity. EDS is a measure of how thoroughly the federal record corroborates a carrier's safety profile; it is not a prediction of liability, fault, settlement value, or any specific case outcome.

Record dimensions

Inspection history — the volume, recency, and BASIC-domain coverage of roadside inspection and violation records, which establish how consistently the carrier's operations are documented in the federal record.
Crash-record consistency — the recency-weighted presence of fatal- and severe-injury crash records in the FMCSA Crash file, with recent records weighted more heavily than historical ones (records remain within the discoverable retention horizon for roughly a year).
Authority continuity — operating-authority and insurance-filing continuity, which indicates how stable and traceable the carrier's regulatory footprint is over time.

Categorical bands

EDS is reported as a categorical band, not a numeric score. The band currently informs factor-based ranking order; it is not displayed as a standalone label on carrier pages.

Band	Range	Interpretation
HIGH	≥ 70	Deeply corroborated federal record — dense, recent, and mutually-consistent across inspection, crash, and authority dimensions.
MODERATE	50 – <70	Standard federal documentary footprint present, with partial corroboration across record dimensions.
LIMITED	< 50	Sparse or thin federal record — limited inspection, crash, or authority history available to corroborate the safety profile.

Stated limitations

The component weights are heuristic priors, not empirical maximum-likelihood estimates; each is published so an independent expert can reconstruct the band assignment under Fed. R. Evid. 702 testability standards.
Record corroboration depth is independent of insurance posture — a self-insured carrier (presumed) with a dense, recent federal record can still fall in the HIGH band, since discoverable-record density does not depend on the bond instrument.
The crash-record recency weighting assumes uniform decay across crash severities. A severity-conditional weighting (multi-fatality incidents decaying more slowly) is left for Phase 2 calibration.

Record Interpretation Modules

A record-interpretation module is a single sentence the carrier page renders from one federal field, under a stated render gate, with a stated refusal. Each module below reports a count or a date and declines to characterize it. The gate is the condition under which the sentence appears at all; where the gate is not met, nothing is rendered in its place.

Module	Renders when	Source	Unit
Crash geography	≥ 6 crashes on record AND crashes in ≥ 2 states	MCMIS crash ledger (crash_state)	count of crashes
Fatal-record answer	fatal_count_5y ≥ 1	FMCSA fatal crash count (5-year) + the crash ledger	count, plus the month of the most recent fatal crash
Inspection category counts	≥ 1 inspection recording that category in the trailing 24 months	FMCSA violation records, pre-aggregated per carrier	inspections (distinct inspection IDs) — not violations
Operating-authority history	involuntary_revocation_count ≥ 1	FMCSA authority events, involuntary revocations only	count of separate involuntary revocations
Insurance continuity	a lapse in the filing history within 5 years	FMCSA L&I filing history	presence or absence of a filing gap
Federal minimum match	the active BIPD limit equals $750,000 exactly	FMCSA L&I active filings	exact equality, not a rounded multiple

What each module declines to say

Crash geography: No “most”, “concentrated”, or “disproportionate”. It reports how many of the crashes on record occurred in the largest single state and stops. A carrier whose crashes are all in one state gets no sentence — “6 of 6 in TX” restates the fleet's address, not its record.
Fatal-record answer: The two figures span different windows: the count is the federal 5-year total, the ledger reaches back about 26 months. When a fatal crash falls inside the federal window but outside the ledger, the module drops the date clause and reports the count alone rather than nominating a “most recent” it cannot see.
Inspection category counts: No denominator. The inspection total FMCSA publishes in its SMS snapshot and the inspection count we load are drawn from different files and differ by roughly 5–10% in either direction, so pairing them as a ratio would publish a wrong ratio. At most one category is reported; stacking three counts makes a list, and the list itself reads as a judgement.
Operating-authority history: It never substitutes the total revocation count, which includes voluntary surrender — stating a voluntary surrender as an involuntary revocation would be a false statement about a named company. It reports a count, never a “pattern”, and says nothing about why authority was revoked or what the carrier's status is today.
Insurance continuity: No adequacy judgement and no connection to any crash. A gap in a filing history is a fact about filings; whether coverage was sufficient for anything is not a question this record answers.
Federal minimum match: It does not round. A $760,000 limit is not “at the federal floor”, and an earlier implementation that rounded to one decimal said it was. Limits below the general-freight minimum are reported as filed and are not asserted to be deficient — lighter vehicles and intrastate operations carry different federal minimums.

Suppression rules

A suppression rule is a condition under which Safeny publishes less than the data would allow. Each one exists because a figure that is correct can still support a claim that is not, and the rules below name the point at which the figure stops supporting the claim.

Fewer than 5 inspections in the trailing 24 months: Every above/below-national-average assertion is suppressed, and — since 2026-07-27 — so is the comparison bar itself. The plain fact sentence (M out-of-service orders across N inspections) stays.; A bar drawn next to a national average is a comparison whether or not a sentence says so; erasing the caption leaves the picture making the claim. At one to four inspections a single out-of-service order moves the rate too far for that comparison to be supported, even when every number in it is correct. The fact sentence survives because it carries its own sample size.
The supporting window does not cover the figure: The clause that depends on the narrower window is dropped; the figure from the wider window is still reported.; A date we cannot see is not a date we may estimate. Reporting a count without its date is incomplete; reporting a date we inferred is wrong.
Two sources would have to be divided: The ratio is not published; the numerator is published alone.; Both counts are federal and both are correct about what they measure, but they are not measuring the same set. A ratio built across them is a number about nothing.
A count column that has not been derived yet: The module renders nothing.; The safe state of a claim about a named company is silence. A fallback to a broader count — the total that includes voluntary surrender — would produce exactly the false statement the module exists to avoid.

What Safeny does not compute

The following are not produced by this methodology at any stage. Some were withdrawn after publication; each is listed with the reason it is absent, because an unexplained absence reads as an oversight.

Fault or liability for any crash: Federal records document crashes; they don't assign fault — fault is established by investigation, police reports, and litigation.
Settlement value or case value: The prior settlement-range model was withdrawn rather than published with a manufactured error rate; no replacement is computed.
Attorney-value ranking as a published figure: The ranking is retained internally for corpus operations and is not surfaced as a carrier-facing score.
Insurance adequacy, or a coverage gap: The public dataset does not disclose the exposure a limit would have to cover, so no gap is computable from it.
A carrier's current operating status from its revocation history: Authority history is reported as history. Whether a carrier is operating today is a question for the live federal record, not for a count of past events.
Reconciled fatality counts: Fatality counts come from one federal source and are not reconciled against any other federal count. Measured against a second federal count the two disagree on roughly 29.5% of carriers, which is a reason to name one source and stay with it, not to average them.
Predictions of any kind: No outcome, no recovery, no future event. Every published figure is a description of a record that already exists.

Module gates and suppression thresholds are versioned with this methodology artifact (SHA a81bb40e…). A gate change is a methodology change and carries a version entry.

Reproducibility

An independent reviewer using only public FMCSA data can recompute every published figure within rounding error. All calibration SQL is maintained in version control; reference SQL for any published computation is available on written request (research@safeny.com). All formulas in the sections above are implemented directly in the deterministic stages below. The section documents data sources, computational environment, and fixed numerical parameters in the form intended to support Fed. R. Evid. 702 / Daubert testability.

Primary data sources

FMCSA SAFER — Motor Carrier Census, monthly snapshot. Source: safer.fmcsa.dot.gov (public).
FMCSA SMS — Safety Measurement System monthly files (Crash, Inspection, Violation, SMS_AB_PassProperty, SMS_C_PassProperty). Source: ai.fmcsa.dot.gov/SMS/Tools/Downloads.aspx (public).
FMCSA L&I — Licensing & Insurance ActPendInsur, daily delta. Filename: actpendins_allwithhistory.txt. Source: catalog.data.gov (public).

Computational environment

Language. ETL in Python 3.11 (numpy, polars, psycopg). Statistical computation for the published figures — the night-driving binomial test (normal approximation via erf) and EDS composition — is implemented in PostgreSQL SQL functions, maintained in version control.
Storage. PostgreSQL 15 (Supabase). All calibration SQL is checked into supabase/migrations/.
Orchestration. Staged execution script pipeline/scripts/recompute_staged.py runs the full pipeline as 16 deterministic stages; each stage is independently re-runnable and produces identical output for identical input.
Determinism. No random-seeded estimators are used. All shrinkage and significance-test estimators are deterministic given input data plus the fixed parameters below.

Fixed numerical parameters

Symbol	Value	Used in
t_½ (EDS recency)	180 days	EDS R(τ) recency kernel half-life.
p₀ (night baseline)	0.257	National dark-condition crash share (binomial null).
N_min (sample gate)	5	Minimum ~2-year crash count to compute a per-carrier night-pattern p-value.

Recompute procedure (deterministic stages)

Download FMCSA SMS source for the target month from ai.fmcsa.dot.gov/SMS/Tools/Downloads.aspx. Six files: Motor Carrier Census, Crash, Inspection, Violation, SMS_AB_PassProperty, SMS_C_PassProperty.
Download L&I active filings actpendins_allwithhistory.txt from catalog.data.gov.
Compute SMS percentile ranks per cohort. Public SMS files publish raw MEASURE values; apply PERCENT_RANK partitioned by (fleet_cohort, safety_event_count_bucket).
Compose EDS: evaluateF(c, t) with the 180-day half-life kernel, then multiply through the seven M_* factors. Apply theM_feedback ≤ 5.0 clamp.

Reference SQL for the published computations is in our public migration files (supabase/migrations/). The literature underlying each numerical choice above is cited in the References section.

References

The estimators and statistical procedures used throughout this methodology are drawn from the following primary literature. Cited in APA 7 format.

Abramowitz, M., & Stegun, I. A. (Eds.). (1972). Handbook of mathematical functions with formulas, graphs, and mathematical tables (10th printing). National Bureau of Standards.§7.1.26 erf rational approximation; used inside the normal-CDF computation underlying the binomial significance test in the night-driving pattern section.

Limitations of analysis

Disclosure under Fed. R. Evid. 702

We disclose the following limitations of carrier-level analysis, consistent with Daubert evidentiary standards for expert opinion (Fed. R. Evid. 702). Each limitation is scoped to its current resolution status; Phase 2 work items are tracked in the methodology PDF appendix.

Percentile back-fill — FMCSA SMS Output ships raw MEASURE + AC flag, not percentiles. We backfill via cohort PERCENT_RANK; results may differ from FMCSA's official percentile calculation by ±2–3 points.
EDS band sensitivity — Sensitivity (measured): overall band retention is 99.7–100% under ±10–25% component-weight perturbation; HIGH-band membership (N≈100) is boundary-sensitive, with retention as low as 17.4% under ±25% perturbation.
Insurance stacking simplification — Multi-subsidiary BIPD JOIN miss affects 208 legit-large carriers (Ryder, Penske, Walmart Transportation, et al.) where coverage is filed under parent MC#. Phase 2 resolution via FMCSA family-tree API pending.
Sub-state venue variance — State tier classification assumes within-state homogeneity. County and circuit variance within a state is not modeled; consult local counsel for venue-specific tendencies.
Carrier-level scope only — No individual driver history, specific shift conditions, equipment-specific maintenance trail, or internal company policies. These factors are often determinative of case merit and require attorney evaluation of case-specific records.

SHA a81bb40e…

Disclaimer for Ranking Data

Safeny rankings are mathematical sorts of FMCSA public records, computed by applying the methodology described above. The act of ranking does not constitute a determination, judgment, or assertion of fault, negligence, liability, or culpability against any named carrier or person.

Carriers identified in ranking data may dispute factual claims via the Safeny Carrier Governance protocol.

The information in ranking data is provided as a research aid; court admissibility, evidentiary weight, and legal interpretation are responsibilities of qualified legal counsel.

Disclaimers

Not legal advice. The compiled FMCSA figures on this site are decision-support information for case selection. They are not, and do not substitute for, legal advice. Consult a licensed attorney for case-specific evaluation.
Statistical record only. These figures describe a carrier's federal safety record from public records; individual case outcomes vary widely with facts, jurisdiction, evidence admissibility, and counsel quality. We do not predict any specific case's verdict.

Full version history and erratum log: /cite/integrity

Verify this methodology snapshot: /v/6.6