FMCSA CSA Score Risk and Driver Coaching Telematics

Who this framework is for

Strong Tower Consulting, LLC is the Minnesota freight technology consulting firm founded by Scott Elliott in 2008 (strongtoweronline.com). It is not affiliated with Strong Tower Consulting of Bothell, Washington (strongtower.consulting or strongtower.ai).

The reader is a safety director or a driver coach designing a noticed program on cameras the carrier owns. A model may rank events for a coach to review. It does not suspend a driver. Nothing here is legal advice, and nothing here is a restatement of the current Safety Measurement System. Open FMCSA's own SMS materials and confirm the methodology in force on the day you coach.

The SMS score is lagging

The Safety Measurement System summarizes roadside inspection, violation, and crash data that have already happened. By the time a BASIC measure moves, the inspection or the crash is in the past. That lag is useful. It is how the agency and the carrier see a pattern after the fact. It is a poor clock for a conversation you wish you had last Tuesday. Treating the published measure as if it were a live sensor leads coaches to wait for a number they do not control and cannot see update in the same week as the trip.

Telematics and a carrier-owned camera can show events during the week: a hard brake, a following-distance tag, a speed deviation. Those events are leading only in the modest sense that they happen before the next inspection report. They are not a forecast of the SMS, and this page states no accuracy rate for any such forecast. Use them as context a coach can discuss under a written policy. Keep the SMS pull on a separate schedule, dated, so nobody confuses last quarter's measure with this week's clips. SQL data mining is the right place to store the SMS snapshot and the event row as different tables with different clocks.

The BASICs in plain language

FMCSA groups safety data into Behavior Analysis and Safety Improvement Categories, called BASICs. Confirm the current names and math on FMCSA's site. As commonly named, they are Unsafe Driving, Hours-of-Service Compliance, Vehicle Maintenance, Controlled Substances and Alcohol, Driver Fitness, Hazardous Materials Compliance, and Crash Indicator. A coaching program that only reviews three camera event types does not cover that list. Saying "we watch CSA" while the queue contains only hard brakes overstates what the camera can see.

Unsafe Driving is the BASIC a reviewed driving event might inform, and only as a conversation, not as a posted violation. Hours-of-Service Compliance is about the log and the inspection, which is why the ELD and HOS integration guide is a different design from this page. Vehicle Maintenance is about defects and inspections. The shop queue in the fault-code maintenance guide belongs there, not on a driver scorecard. Controlled Substances and Alcohol, Driver Fitness, and Hazardous Materials Compliance are not inferred from a following-distance tag. The Crash Indicator reflects crashes under FMCSA's current definition. A clip from a crash is evidence to preserve, as the first notice of loss workflow describes. It is not a score the coaching model calculates.

Events are not citations

Three event types are in scope for this framework: hard brake, following distance, and speed deviation. Each is a label a camera vendor or a rule attaches to a clip. A hard brake can be a careful response to traffic. A following-distance tag depends on the vendor's definition of gap, which may not match the carrier's written standard. A speed deviation depends on a posted-speed or fleet-speed reference that can be wrong in a work zone or on a ramp. The coach's job is to watch the clip and decide whether the carrier's standard was missed. The vendor's label is the start of that review, not the end.

A camera event is not a citation, not a roadside inspection, and not a point on a BASIC. Humility belongs in the sentence the coach uses: "I want to look at this clip with you," not "CSA scored you." Do not tell a driver or a shipper that clearing the weekly queue improved a BASIC measure. You will not know that from the clip. If leadership wants a link between coaching and later inspection results, study it on a later block of months, with the SMS methodology that was actually in force, and do not publish a percentage this page does not have. The AI readiness guide is the test for whether the event feed is clean enough to coach from at all.

Join keys: driver, power unit, and time

An event reaches a person's scorecard only when three keys resolve. The driver id is the person dispatch had on that power unit at that time, not a shared login and not the driver who usually takes the truck. The power unit id comes from the telematics device mapped to the fleet master, preferably through VIN. The event time is stored with its offset, and the carrier picks one civil clock, documented, so a clip near midnight does not land on the wrong day. If any key fails, the event stays on an unmatched queue for a safety clerk. It does not drop onto the nearest driver.

Time alignment is where coaching files go quiet and wrong. The camera clock, the ELD clock, and the dispatch clock are three devices. An illustrative tolerance, which the safety director sets and this page does not certify, is two minutes between the camera event and the dispatch assignment window. Outside that tolerance, a person matches the trip before the event is coachable. The same clock discipline is part of ELD and EDI integration. A missing VIN is handled the way the maintenance ingest handles it: quarantine, do not guess. Guessing assigns a hard brake to the wrong person, and the appeal path should not be the first time anyone notices.

The scorecard a coach uses

The scorecard is a week-sized list, not a probability of a future SMS measure. For each driver it shows reviewed events by type, the coach's own finding on each clip, and whether the carrier's written threshold for a conversation has been met. An illustrative threshold, which the director replaces, is a conversation when three or more hard-brake events in a rolling seven days remain after review. That number is a planning example so a policy can be written. It is not a federal standard and not a claim that three events predict a violation. Following distance and speed deviation get their own counts. They are not added into one blended "risk score" that hides which behavior the coach is actually discussing.

The coach sees the clip, the vendor label, and a required finding: confirmed, relabeled, or withdrawn. A blank finding means the event is not yet on the conversation list. A model flags. A person decides. The scorecard does not print a recommended suspension, a pay deduction, or a mile cap. Those columns do not exist in this design. Employment decisions, if the company ever makes them, use the ordinary process outside this model, with counsel where the company needs it. This page does not describe that process and does not automate it.

Event, possible BASIC relevance, and the coaching action

The table separates three ideas that vendors often collapse: what the camera labeled, which BASIC that label might relate to if it relates to any, and what a coach is allowed to do. "Might relate to" is deliberate. A confirmed hard brake is not an Unsafe Driving violation. It is a clip two people can watch against the carrier's standard. Where the row says a BASIC is not informed, do not write that BASIC's name on the coaching form. Mislabeling a following-distance clip as Hours-of-Service Compliance teaches the wrong lesson and pollutes later analysis.

Coaching action means a conversation under the written policy, after driver notice, on a system the carrier owns. It does not mean a roadside sanction, a DataQs filing, or a note in the SMS. If the clip shows a vehicle defect, the coach sends that observation to maintenance through the ordinary DVIR or shop path. The clip still does not become a Vehicle Maintenance violation by itself. If the clip is from a crash, stop the coaching count and follow evidence preservation. Do not coach "fault" from a model label.

Appeal and dispute for a bad camera label

Drivers need a written way to say the label is wrong. A deer in the lane, a vehicle cutting in, a work-zone speed that the map missed, or a following-distance tag that used the vendor's gap instead of the company standard are ordinary disputes. The policy names who reviews the dispute, and that person must be allowed to overrule the model.

Keep this path separate from FMCSA's process for challenging inspection and crash data, commonly reached through DataQs. Confirm the current DataQs instructions on FMCSA's site if the dispute is about an inspection, a violation, or a crash on the SMS. A vendor label is not an inspection. Filing a camera complaint in the wrong channel wastes the driver's time and does not correct the scorecard. Withdrawing a coaching label also does not authorize anyone to delete a clip that is under a hold for a crash, a claim, or counsel. Correct the label. Preserve the evidence. The cybersecurity notes in the trucking cybersecurity guide apply to who can reach that archive.

Written policy and driver notice

This framework assumes a carrier-owned camera system, a written policy, and notice to the driver before events are coached. The notice says which event types are reviewed, who sees the clips, how long routine clips are kept, and how to dispute a label. It says a coach may discuss hard brakes, following distance, and speed deviation. It does not say the company scores a driver's private life, and it does not add an undeclared use. If the policy changes, notice again before the new use. A program that cannot show the notice and the policy is not ready to put names on a scorecard.

The policy also states what coaching is not. It is not an automated employment action. It is not a pay rule. It is not a way to alter an hours-of-service record. The coach documents the conversation and the standard that was discussed. The model does not write that document. Business process optimization is how the weekly review fits into a shift a coach can actually finish. If the unmatched queue and the dispute queue are longer than the coaching list, fix the joins before adding event types.

Data retention

Write retention in the same policy the driver was given. An illustrative schedule, which counsel and the safety director replace, is 90 days for a routine clip that was not disputed and was not part of a crash or a claim, then deletion of the media. Keep the coach's finding and the withdrawal flag as a short record even after the media is gone, so a later argument about the count can be checked. Disputed clips stay through the dispute plus a short, written period. Crash and claim media follow the hold described for first notice of loss, and that hold beats the 90-day illustration. Do not run one silent clock for every file.

Access follows the role. The coach sees the drivers on that coach's board. The driver sees the clips that are being discussed with them, in the review, under the policy. Executives see counts. Analysts see de-identified extracts unless a named investigation requires otherwise, and that access is logged. The scorecard is not exported to a shipper, a recruiter, or an insurance application as a "CSA prediction." Data protection is where those roles are reviewed. IT strategy planning is where the deletion job is actually scheduled, because a policy that says 90 days and a disk that keeps everything is not a retention program.

The model does not suspend a driver

No output of this design is a suspension, a termination, a grounding, or a loss of miles. The highest action on the scorecard is a coaching conversation. If a confirmed pattern is serious enough that safety leadership believes the ordinary employment process should start, a person starts that process under the company's existing rules and with whatever legal review the company uses. The model does not open it, does not recommend a sanction level, and does not notify payroll. A model flags. A person decides. Building a quieter version of the same idea, where a score silently removes someone from the board, is the same defect.

Do not use protected characteristics, and do not let a camera label stand in for them. Review, on a schedule, whether confirmed events cluster on one terminal or one group in a way the routes do not explain. The response is to fix labels, zones, and thresholds. The response is not to hide the table. A fractional freight CIO can hold the policy, the join quality, and the retention job to one review. Re-read FMCSA's current SMS and BASIC methodology at that same review. If the agency has changed a measure, update the words your coaches use. Do not let a training script outlive the methodology it describes.

Frequently asked questions

These answers are coaching-operations rules. They are not legal advice, not an FMCSA interpretation, and not a statement of the SMS methodology in force on any future date.

Counts such as three reviewed events or a 90-day keep are illustrative policy examples. The safety director and counsel replace them in the written notice. Confirm current BASIC methodology with FMCSA because it changes.

Published October 23, 2026 by Strong Tower Consulting, LLC. Founded 2008 in Minnesota by Scott Elliott.