Building a fleet safety program that insurers recognise

The policy, MVR cadence, training records and review board that make a safety program real, and what an insurer or defence counsel will ask to see.

Municipal / public fleets Small-business fleets Reviewed July 2026

Most fleets have a safety program in the sense that people talk about safety. Far fewer have one in the sense that matters: that a stranger can open a file and reconstruct, for any driver and any date, what the rules were, that the driver was told them, that you checked whether they followed them, and what you did when they did not.

That reconstruction is the product. It is what an underwriter assesses and what your defence counsel needs. A program that exists only as practice is worth very little at the moment it counts.

The written policy

Written policies are step 2 of ten in the joint OSHA/NHTSA/NETS employer guidelines, behind senior management commitment. One document, given to everyone who drives: seat belts, phone and distraction rules, alcohol and drugs, fatigue and maximum shift length, personal use, who may drive which classes, and what to do at a crash scene. State each rule as a behaviour — “no handheld use while the vehicle is in motion” beats “drive attentively”.

Two things make it load-bearing. Every driver signs it and you keep the signed copy, which turns “we have a phone policy” into “this driver acknowledged the phone policy on this date”. And you apply the stated consequences consistently — inconsistent enforcement is worse than no written rule, because it hands the other side a document you demonstrably ignored.

Driver selection and MVR monitoring

Pull the motor vehicle record before the offer, and set disqualifying criteria in writing beforehand: how many moving violations in what window, which offences are automatic bars, how long a DUI disqualifies. Deciding after you have seen a specific person’s record invites the accusation that the standard is elastic.

For fleets in scope of the federal rules the cadence is not yours to choose. 49 CFR 391.25 requires an MVR inquiry and documented review at least every 12 months, with the record and a dated note naming the reviewer in the driver qualification file. Part 382.701 requires a pre-employment full query of the Drug and Alcohol Clearinghouse and at least annual queries thereafter; the annual one may be a limited query, reporting only whether a record exists.

The limited query saves paperwork, not money — FMCSA charges the same for both. The difference is consent: a limited query runs on written consent you collect and keep, while a full query needs the driver registered and consenting electronically to that request. If a limited query reports a record exists, you have 24 hours to run the full query or pull the driver off safety-sensitive work. Unsure whether this reaches you? Start with do the DOT rules apply to my fleet.

Require self-reporting too: drivers tell you within a fixed number of days of any citation, suspension or crash in any vehicle, including their own. Put it in the signed agreement.

Training and telematics

New-hire training covers the policy, a road test for the vehicle class, and the specific hazards of your operation — for a municipality usually backing, work zones and plow operation rather than generic defensive driving. Periodic training works better short and frequent than as an annual session everyone forgets. What matters for the record is date, topic, materials and a signed roster; a roster with no topic attached is nearly useless later.

Telematics supplies objective facts and surfaces patterns before they become crashes. Deployed badly it destroys the reporting culture the rest of the program depends on — if drivers believe it exists to catch them they stop reporting the near misses you most need to hear about. Tell drivers what is collected, who sees it, how long it is kept and what it will not be used for, before installation. Do not run it as a leaderboard with discipline attached to raw event counts, which are heavily confounded by route. Give drivers a way to contest an event, and treat deployment as a bargaining matter early if you have a collective agreement.

Incident reporting

The OSHA/NHTSA/NETS guidelines say all crashes, regardless of severity, should reach the supervisor as soon as feasible. Extending that to near misses and no-damage contacts is a choice on top of the guidance, and it is the one worth making.

This step quietly fails because a driver who scuffs a bumper in a yard concludes saying nothing is easier. Fix it with a form short enough to complete at the scene, a policy making the failure to report the disciplinable event rather than the incident, and a review process whose visible output is training. DOT-regulated carriers must additionally keep an accident register under 49 CFR 390.15(b) for three years per accident.

The accident review board

A small standing group that reviews every reportable incident and makes one determination: preventable or non-preventable. Give it a written charter — the charter is what makes determinations look like a process rather than a verdict. It should specify membership (fleet, the operating department, safety or risk, and a driver representative — the driver seat is why drivers accept outcomes), what it reviews, a fixed timeline from report to determination, the evidence considered, the definition of preventable quoted in full, and rules for appeal and recusal.

What “preventable” means. Lift the regulatory definition: 49 CFR 385.3 defines a preventable accident as one “that could have been averted but for an act, or failure to act, by the motor carrier or the driver.” It does not govern your internal board, but adopting it costs nothing and gives you a citation. It is deliberately wider than legal fault — a driver rear-ended at a stop can be blameless in law and still draw a preventable determination if following distance contributed. The National Safety Council’s guide is the other common source; if you use it, quote NSC’s current wording rather than a version repeated on a blog, because the phrasings in circulation differ in ways that matter once someone is arguing about one.

Findings feed training, not discipline. If the only output is a disciplinary letter, drivers contest everything, report less, and the board produces no safety improvement. What earns its existence is a cause code and a corrective action tracked in aggregate. When backing incidents are a third of your preventable determinations, that is a training and spotter-policy decision, not eight personnel matters.

What to keep, and for how long

Record	Typical retention
Signed policy acknowledgement	Employment + 3 years
Pre-hire MVR and hiring decision	Employment + 3 years
Annual MVR and dated review note	3 years from execution — 391.51(d) lets you pull these from the DQ file
Clearinghouse query records	3 years — 382.701(e)
Training rosters with topic and date	Employment + 3 years
Incident reports and board determinations	5 years, or per counsel
Accident register	3 years per accident — 390.15(b)

The split in the first two MVR rows catches people. The pre-hire MVR stays in the driver qualification file for employment plus three years; the annual MVR and its review note are on 391.51(d)‘s removable list, so their clock runs three years from execution.

About the insurance claims

Do not put a premium percentage in a budget memo

Figures like “a documented safety program cuts premiums by 15 to 25 percent” circulate widely and nearly all trace back to vendor marketing with no disclosed sample or methodology. Ask your broker for the specific credit in writing instead — it is underwriter discretion, and the only number that means anything is the one your carrier commits to for your fleet. In a public risk pool or self-insured, premium is the wrong lever entirely: contribution tracks loss experience over a multi-year window, so the program pays by reducing claims, two or three years later.

Use avoided loss instead. NIOSH publishes $75,000 per nonfatal work-related crash injury and $751,000 per death — per injury, not per crash, and itself a modelled NETS estimate rather than a government survey.

Doing this at scale

All of this runs on a shared drive and a spreadsheet — a policy folder of signed PDFs, an MVR tracker with due dates, a training log, and a numbered incident file per event. That works, and plenty of well-run fleets do exactly this. It tends to break down on the cadence rather than the volume. The tracker only works if somebody opens it every month and chases the MVRs that came due, and that is the first task dropped in a short-staffed office. The failure is silent: nothing looks wrong until an incident file gets pulled and the review note for that driver is eighteen months old.

At that point fleets usually move to fleet management software that keeps asset history, work orders, and cost data in one place so these calculations run continuously instead of once a year. There are several credible options — evaluate them against your own requirements, not a feature list.

Disclosure: this site is supported by Axion, which sells software in that category.

Sources

OSHA — Motor vehicle safety for employers Primary
OSHA / NHTSA / NETS — Guidelines for Employers to Reduce Motor Vehicle Crashes Primary — joint federal publication, advisory rather than a standard or regulation by its own disclaimer; the ten-step structure holds up, but its cost figures come from 2003 NHTSA data and are badly out of date
NIOSH — Motor vehicle safety at work Primary — where the employer cost figures below are published. NIOSH is citing its reference 2, NETS [2021], Cost of Motor Vehicle Crashes to Employers — 2019 — a modelled estimate from an employer non-profit, not a government survey
49 CFR 385.3 — regulatory definition of a preventable accident Primary
49 CFR 391.25 — annual inquiry and review of driving record Primary
49 CFR 391.51 — driver qualification files and retention Primary
49 CFR 390.15(b) — accident register Primary — the register requirement sits in paragraph (b) of a section titled "Assistance in investigations and special studies"
49 CFR 382.701 — Drug and Alcohol Clearinghouse queries Primary
FMCSA Drug and Alcohol Clearinghouse Primary
FMCSA Clearinghouse — Queries and Consent Requests factsheet Primary — sets out the consent mechanics that actually separate a limited query from a full one
FMCSA Clearinghouse — Query Plan factsheet Primary — confirms both query types are charged at the same rate, and that a limited query followed by a full query on the same driver is charged only once
FMCSA — Crash Preventability Determination Program Primary
FMCSA — Crash Preventability Determination Program, 89 FR 96269 (4 December 2024) Primary — the final notice listing the 21 eligible crash types, applying to crashes on or after 1 December 2024
California DMV — Employer Pull Notice program Primary — one state's continuous licence-monitoring service, cited as a worked example
ANSI/ASSP Z15.1-2024 — Safe Practices for Motor Vehicle Operations Primary — consensus standard from a standards body; the document itself is paid
National Safety Council — Collision Preventability Review — NSC is a non-profit safety organisation; this page describes a paid independent review service, and NSC sells the preventability guide separately