How to Read CSA BASIC Percentiles Like an Underwriter

The underwriter's method in one paragraph

An underwriter never asks "is the number high?" They ask three questions in sequence: which BASIC is elevated (a behavioral category like Unsafe Driving predicts losses far better than a paperwork category like Driver Fitness), how much data sits behind the number (a percentile built on nine inspections is noise; one built on four hundred is signal), and which direction it's moving (a 55 trending down over three refresh cycles beats a 40 trending up). Only after those three does the raw value matter.

Quick orientation for anyone new to the system: FMCSA's Safety Measurement System distills roadside inspection and crash data into seven behavioral categories — Unsafe Driving, Hours-of-Service (previously called Fatigued Driving), Vehicle Maintenance, Driver Fitness, Controlled Substances/Alcohol, Hazmat Compliance, and the Crash Indicator. Each gets a 0–100 percentile where 0 is best: a 72 means the carrier ranks worse than 72% of comparable operations in that one category. The rankings refresh monthly, and two of the seven (Crash Indicator and Hazmat) are withheld from public display under the FAST Act, though they still exist and still drive enforcement.

Intervention thresholds, BASIC by BASIC

FMCSA doesn't act on every elevated number. Each BASIC has an intervention threshold; crossing it can trigger warning letters, prioritized roadside targeting, and investigations. Crucially, the thresholds are not uniform: the agency sets them lower (stricter) for the categories most predictive of crashes, and lower again for hazmat and passenger operations. That asymmetry is itself information — it tells you which categories the regulator considers most dangerous:

How a practitioner uses this table when vetting:

• An elevated number in the 65-threshold rows (the crash-predictive trio) is materially worse news than the same number in an 80-threshold row. A 68 in Unsafe Driving has crossed the line; a 68 in Vehicle Maintenance hasn't.
• A carrier sitting just under threshold in two or three BASICs simultaneously is often a worse bet than one alert in a single category — systemic near-misses describe the culture, not one bad event.
• For hazmat or bus operators, re-read every number against their stricter column before drawing conclusions.

Why a percentile is not a crash probability

The single most common misreading of this data is treating an 80th-percentile carrier as "80% likely to have a problem." The number is a rank among peers, and four structural facts break any direct line from rank to risk:

None of this makes the data useless — FMCSA's research shows above-threshold carriers do crash more as a population. It means the percentile is a screening instrument, not a verdict: strong enough to rank and prioritize, too noisy to condemn a specific carrier by itself. Which is exactly how underwriters treat it — and how it feeds our scoring, below.

What "no score" actually tells you

Search enough carriers and you'll notice most BASICs display no percentile at all. The SMS only computes a rank once a carrier clears minimum data thresholds — broadly, several relevant inspections with at least one violation in that category inside the two-year window. Below that floor, there's nothing to rank.

Reading the blank correctly matters as much as reading the number:

• Blank plus meaningful inspection history can be genuinely good news — the carrier has been stopped repeatedly without accumulating scoreable violations in that category.
• Blank plus almost no inspections is an information vacuum, not a clean bill of health. New authorities live here, and this is precisely the population where fraud and fly-by-night operators concentrate — a reason the DOT check workflow weighs authority age so heavily.
• Blank because the entity is a broker is a category error — no trucks, no inspections, no safety data. Broker vetting runs on bond and licensing signals instead; see the MC authority guide.

Separate concept, often confused: the safety rating (Satisfactory / Conditional / Unsatisfactory) comes from an on-site compliance review, not from SMS math. Most carriers have never been reviewed, so "unrated" there is the default state — while a Conditional rating is a loud, current red flag that our scoring penalizes hard.

How MC Look Up folds safety data into its 0–100 score

Percentiles are one input to our carrier score, not the output. The safety layer of the model (Layer 2A, worth up to 35 points of deduction) deliberately leans on raw inspection outcomes — which don't suffer from the peer-group artifacts above — and uses BASIC data where it's most reliable:

• Vehicle out-of-service rate vs. the national baseline. Nationally, roughly 23.4% of vehicle inspections end with equipment placed out of service. We grade each carrier's rate against that: modest deductions begin above the average and scale up steeply — a fleet failing more than half its vehicle inspections takes the maximum penalty in this component.
• Driver out-of-service rate vs. the national baseline. The national driver OOS average is about 6.7%. Rates above it deduct progressively, with rates past 25–40% treated as evidence of systemic hours or licensing problems. (Passenger carriers are graded against their own stricter baselines: 18% vehicle, 5% driver.)
• Violations per inspection. Averaging more than two violations every time an officer opens the hood indicates depth of neglect that a binary OOS rate can miss; past three per inspection the penalty escalates, past five it maxes.
• Acute/Critical BASIC flags. Where FMCSA has marked a BASIC with an acute or critical violation indicator, each flagged category deducts points directly — these flags reflect investigator findings, so they bypass the statistical caveats entirely.
• Conditional safety rating. An unresolved Conditional rating from a compliance review takes a heavy fixed deduction on top of everything else.

Confidence weighting: the anti-noise mechanism

Every statistical deduction above is multiplied by a confidence factor derived from inspection volume. A terrible OOS rate on four inspections is discounted because it might be one unlucky day; the same rate on two hundred inspections lands at full force. This mirrors the underwriter's sample-size discipline from the top of this page — and it's why two carriers with identical rates can score differently. Carriers with almost no data get an explicit "insufficient inspections" note on the profile instead of a false clean score.

The safety layer then combines with crash history, insurance, authority, and pattern-intelligence layers — including chameleon re-registration screening — to produce the final APPROVE / REVIEW / REJECT verdict. The full methodology, with every deduction itemized, is documented in the help center, and you can browse scored carriers by state in the directory.

Frequently asked questions

Keep going: BASIC scores are one input — see how they fit a complete vet in our walkthroughs of the DOT number lookup and the MC number lookup.