Fleet telematics and driver cameras: how to coach drivers, not punish them
Use telematics and cameras to coach drivers fairly, protect privacy and measure real change
Picture a delivery driver who brakes hard when a car cuts into her lane. Minutes later, her supervisor gets an alert. What happens next decides whether your monitoring program makes the fleet safer or just makes drivers resentful.
Telematics systems record how a vehicle is driven, including speed, braking, belt use and, with cameras, what the driver was doing. The Occupational Safety and Health Administration (OSHA) says these systems are meant to flag risky driving so drivers can correct themselves and supervisors can coach [2]. OSHA also urges employers to explain to workers why the system is going in and how it works [2].
This guide takes a fleet telematics specialist’s and a human factors researcher’s view, with privacy and employment notes. It covers what the systems measure, how to set and check thresholds, coaching, rewards, access, privacy, data retention, gaming, escalation and evaluation. For the wider program, see our workplace motor vehicle safety program guide.
What the research says: coaching is the active ingredient
Comparison compiled by Safety Behind the Wheel Foundation
| Study | Setting | What changed | Caveat |
|---|---|---|---|
| Bell et al., NIOSH (2017) [1] | 315 vehicles in local trucking and oil and gas support, about 2 years | Warning lights alone: no significant change. Lights plus coaching with video: odds of risky driving about half the control group’s | Video was coded by the vendor |
| Hickman and Hanowski, VTTI (2011) [3] | 2 trucking carriers; 4 week baseline, then 13 weeks of feedback lights plus coaching | 37% and 52.2% fewer safety related events per 10,000 miles | Not randomized |
| Boyle et al., VTTI for FMCSA (2016) [4] | 156 trucks and motorcoaches, 317 drivers, 4 fleets | Most analyses showed lower critical event rates. One fleet’s crash rate fell 59.8%, from 12.5 to 5.0 per million miles | Results varied by fleet and coaching quality |
| Mase et al. (2020) [5] | 532 heavy trucks at 2 companies | With cameras plus coaching, harsh braking fell 16.8% and speeding 34.3%. Monitoring alone: 4.6% and 28.1% | Only 2 companies |
| Wouters and Bos (2000) [6] | 840 fleet vehicles, 270 with recorders | About 20% fewer crashes on average | Varied by sector and prior safety record |
Three lessons stand out:
- Being watched helps a little. Coaching helps more. Drivers who know they’re monitored tend to drive more carefully [5, 6]. The larger gains came when a person reviewed events with the driver [1, 5].
- Speed and consistency matter. In the Federal Motor Carrier Safety Administration (FMCSA) field test, one fleet that coached events within a week saw steady drops. In another fleet, event rates rose slightly during weeks without coaching [4].
- Fewer events isn’t the same as fewer crashes. Only 2 of the 5 studies above measured crashes. One of them is more than 25 years old [4, 6].
What each system measures, and where it goes wrong
Comparison compiled by Safety Behind the Wheel Foundation
| Signal | How it’s measured | Common errors | How to check |
|---|---|---|---|
| Speed | GPS speed compared with a speed limit stored in a digital map | Maps may miss local roads and lag behind limit changes [7] | Compare alerts with posted signs on the route |
| Harsh braking, acceleration and cornering | Accelerometer force, in g | Potholes and rough roads [4]; device mounting; vehicle weight and load | Watch the clip; look for clusters by vehicle or location |
| Seat belt | Belt latch signal or sensor | In the FMCSA test, one fleet logged 10,573 unbelted events, versus 2,443 in another, possibly from sensor trouble at low speeds [4] | Look for vehicles with unusual alert counts |
| Phone use, distraction, drowsiness | Driver facing camera with software that tracks the face, eyes, head and hands | Glasses, sunglasses, lighting and head angle can throw it off [9] | A person reviews every clip before coaching |
| Hours and location | Electronic logging device (ELD) and GPS | Wrong driver assigned to a trip [4] | Confirm who was driving |
Be realistic about camera analytics. In the studies covered by a 2022 research review, camera based drowsiness detection accuracy ranged from about 72% to over 99%. Many systems were tested in simulators or on video of people acting drowsy [9]. The Federal Trade Commission (FTC) warns that some facial recognition tools make more errors for certain groups of people [10].
A 2023 review in Human Factors found that drowsiness warnings improved sleepiness and driving performance in the studies it covered. But most research has focused on sensors and accuracy, not on how drivers respond. Long term effects need more study [12]. Cameras also show what happened, not why. Mase’s team found the causes of deliberate violations hard to capture on video [5].
A camera event is a clue, not a verdict.
Design in cab alerts with the driver in mind, too. Keep them brief, such as a light or a short tone. Explain what each one means. In the FMCSA test, satisfied drivers said the system made them more aware of their surroundings. Unhappy drivers said it distracted them from the road and didn’t always work [4].
Setting thresholds: start with a pilot
A threshold is the line that turns data into an alert. Set it too low and coaches drown in noise. Drivers also learn to ignore alerts. Set it too high and you miss real risk.
- Give written notice first, then run a 4 week data only pilot. Drivers see their own data. Nobody is coached or disciplined.
- Calibrate by vehicle type. A loaded box truck can’t stop like a sedan, so the same g force means different things in each.
- Start conservative. Flag fewer, clearer events, then tighten once validation is working.
- Publish the final thresholds so drivers know where the lines are.
These starting points are examples for a pilot, not industry standards:
Comparison compiled by Safety Behind the Wheel Foundation
| Behavior | Example starting trigger | Adjust for | Check during the pilot |
|---|---|---|---|
| Speeding | 10 mph or more over the mapped limit for 20 seconds or more; a 2025 AAA Foundation trial used 10 mph over [8] | Lower in school and work zones; known map errors | Compare 20 random alerts with posted signs |
| Harsh braking | About 0.30 g in cars and light trucks, the hard braking level in that trial [8] | Use the vendor’s heavy vehicle setting for trucks and loaded vans | Share of events rejected on review |
| Harsh acceleration | About 0.27 g in light vehicles [8] | Vehicle type and load | Clusters by vehicle |
| Harsh cornering | Vendor default | Vehicle height and load | Repeat events at one intersection |
| Seat belt | Unbelted above about 10 to 15 mph for 30 seconds or more | Low speed yard moves; sensor faults | Vehicles with unusual counts |
| Phone or distraction (camera) | Vendor default, with human review of every clip | Glare, sunglasses, cab layout | False alerts by vehicle |
| Drowsiness (camera) | Vendor default, treated as a welfare and scheduling issue | Night and early morning shifts | Match alerts to schedules and hours |
By our calculation, 0.30 g means losing about 6.6 mph every second. Stopping from 30 mph in under about 4.5 seconds would trigger it.
Validate before anyone is coached
Unchecked alerts erode trust faster than anything else. Before an event reaches a coach, a trained reviewer should:
- Confirm who was driving.
- Watch the clip, if there is one, from a few seconds before the trigger.
- Check the posted limit for any speeding alert.
- Rule out false triggers, such as potholes, rail crossings or a faulty sensor.
- Sort the event: risky, justified or invalid.
Track the share of events rejected by type. If a large share of speeding alerts on one road are wrong, fix the map or the threshold, not the driver.
Coaching that changes behavior
Coaching works best when it’s timely, private, about specific behavior and led by the driver’s own review. In the FMCSA test, safety managers got no instruction on coaching methods or frequency [4]. Don’t repeat that gap: train supervisors and set a 7 day coaching target.
Avoid coaching by text, coaching in front of coworkers, or reading a score aloud without watching the clip. Never contact a driver about an alert while they’re driving.
Positive reinforcement that doesn't bury problems
Recognition works when it’s specific and earned. Thank a driver by name for a clean month on a tough route. Celebrate a team’s falling event rate and invite top drivers to mentor new hires.
Be careful with prizes. OSHA has warned that a no retaliation promise may not be enough to make workers feel free to report injuries when reporting costs them a large reward [11]. By contrast, OSHA says programs that reward reporting near misses or hazards are always allowed [11]. Reward the habits you want more of, such as finishing coaching goals, reporting near misses and steady improvement.
Feedback has limits, too. In the 2025 AAA Foundation trial, feedback plus a reward of up to $100 cut speeding by 11% to 13% and hard braking by 16% to 21%. Handheld phone use didn’t improve [8]. Phone risk needs a firm policy and dispatch practices, as our workplace distracted driving policy guide explains.
Who sees what: access, privacy and transparency
Limit access by role. A starting point:
Comparison compiled by Safety Behind the Wheel Foundation
| Role | Can see | Shouldn’t see |
|---|---|---|
| Driver | Own trips, events, clips and coaching notes | Other drivers’ data |
| Direct supervisor | Validated events and trends for their own team | Other teams; raw clips not yet reviewed |
| Safety reviewers | All events, for validation | HR files |
| Dispatch | Live location during shifts | Video and scores |
| HR | Cases at step 3 or higher on the ladder below | Routine events |
| Leaders | Fleet level dashboards | Named driver detail unless a case is escalated |
Transparency means no surprises, so give every driver a plain language notice and get a signed acknowledgment before go live. Hold a meeting where drivers can see the camera, the app and a sample clip.
Personal vehicles and off duty time need tighter limits. Track only work trips. Use an off duty or privacy mode in take home vehicles if your system has one. Federal ELD rules offer a model: during authorized personal use of a truck, location is recorded only to within about 10 miles [13]. Leave in cab audio off unless counsel approves it.
Data must not become a pressure tool. FMCSA bars motor carriers from using ELD data to harass drivers into driving fatigued or breaking hours of service rules [13].
Data retention: keep what you need, then delete it
Old data adds risk without adding safety, so set a retention schedule with counsel. Here’s an example of how one might look:
Comparison compiled by Safety Behind the Wheel Foundation
| Data | Example retention | Why |
|---|---|---|
| Routine trip and GPS data | 90 days, then keep only totals | Enough to spot patterns |
| Rejected or unreviewed clips | 30 days | No safety value once ruled out |
| Validated events and coaching notes | 12 months | Supports the escalation ladder |
| Crash, injury and claim records | As counsel directs | May be needed in a claim or lawsuit |
| Hours of service records (covered carriers) | At least 6 months, the federal minimum [13] | Required by rule |
| Face templates or other biometric data | Avoid collecting; if used, follow state law | Strict state rules [10] |
How drivers and managers game the numbers
Economists call it Goodhart’s law: when a number becomes the goal, people find ways to hit the number without changing what it was meant to measure.
Watch for these signs:
- Devices offline or cameras blocked more often in some vehicles or routes.
- Event counts falling while near misses, crashes or claims don’t.
- Very few harsh braking events, but video showing close following.
- Supervisors rejecting an unusually high share of their own team’s events.
To prevent gaming, don’t rank drivers publicly or tie pay to raw scores. Judge supervisors on coaching quality, not low event counts. Audit a sample of rejected events and track device uptime. Ban tampering in your policy. Federal rules already forbid disabling or tampering with ELDs [13].
An escalation ladder for repeated risk
Most drivers improve after one or two conversations. A clear ladder handles the few who don’t, fairly and consistently. The triggers below are examples.
Comparison compiled by Safety Behind the Wheel Foundation
| Step | Example trigger | Response | Who’s involved |
|---|---|---|---|
| 0. Self review | Any validated event | Driver sees it within 2 days and can add context | Driver |
| 1. Coaching | 3 validated events of one type in 30 days, or a rate well above peers in the same vehicle type | Private coaching within 7 days; one written goal | Supervisor or coach |
| 2. Improvement plan | Pattern continues 30 days after coaching | Written plan, ride along or targeted retraining; check ins every 2 weeks | Supervisor and safety lead |
| 3. Formal review | No progress after the plan, or a repeat of high risk behavior such as unbelted driving or handheld phone use | Documented review; temporary limits such as no night routes | Safety lead and HR |
| 4. Driving privileges | Deliberate tampering or continued high risk driving | Suspend or end driving duties under your written policy | Management, HR and counsel |
Reset a driver to step 0 after about 90 days without a pattern. Apply the same steps to managers. Honest near miss reports never move anyone up the ladder. For retraining options, see our employee driver training and assessment guide.
Is it working? How to evaluate the program
Set your baseline during the pilot, then track these measures every month:
- Validated events per 1,000 miles, by type. Say a 40 vehicle fleet logs 120 validated harsh braking events over 200,000 miles. That’s 0.6 per 1,000 miles (our calculation).
- Coaching speed: the share of events coached within 7 days.
- Rejection rate: the share of alerts thrown out on review.
- Crashes per million miles, preventable crashes and claim costs.
- Driver views: a short, anonymous fairness survey twice a year.
Read results with care. Drivers picked for coaching after a bad month often improve partly by chance, a pattern called regression to the mean. A staggered rollout gives you a comparison group. Crash counts swing a lot in small fleets, so look at 12 months or more. Our guide to measuring whether a safety effort worked explains baselines and comparison groups.
Frequently asked questions
Is it legal for my employer to track the company vehicle I drive?
It depends on your state and your employer’s policy, since notice, consent and use rules vary. New York, for example, requires written notice for some electronic monitoring [14]. Ask HR for the written policy. Ask a lawyer in your state about a specific situation.
Do driver facing cameras record all the time?
It depends on the system and its settings. Many save short clips around a triggered event, such as hard braking, as in several of the studies above [3, 5]. Your employer’s policy should say when the camera records and who can view the footage.
Can telematics data be used to discipline drivers?
That depends on company policy, any union contract and state law. A fair program disciplines only after events are validated and coaching hasn’t worked. Research suggests coaching, not alerts alone, drives most of the improvement [1, 4].
What is a good harsh braking threshold for fleet vehicles?
There’s no single standard. A 2025 AAA Foundation trial used 0.30 g for hard braking [8]. Heavy and loaded vehicles may need different settings, so calibrate during a pilot.
How long should we keep dash cam footage?
Keep rejected clips briefly, validated events long enough for coaching, and crash footage as counsel directs. Covered motor carriers must keep hours of service records at least 6 months [13].
The bottom line
Recommendation from Safety Behind the Wheel Foundation
Treat every alert as a question, not a verdict. Validate events, coach privately within a week and reward progress. Tell drivers exactly what you collect and why. Then measure whether risky driving and crashes actually fall.
Related articles
- Workplace motor vehicle safety program: a step by step employer guide
- Distracted driving policy at work: a guide for employers and drivers
- Employee driver training: how to train and road test workers who drive
- Work crash and near miss investigation: a step by step employer guide
- Teen driver monitoring: how to use telematics to coach, not spy
Sources and further reading
- 1Bell JL, Taylor MA, Chen GX, Kirk RD, Leatherman ER (NIOSH). Evaluation of an in-vehicle monitoring system (IVMS) to reduce risky driving behaviors in commercial drivers: Comparison of in-cab warning lights and supervisory coaching with videos of driving behavior. Journal of Safety Research 60:125 to 136. 2017. sciencedirect.com
- 2Occupational Safety and Health Administration (OSHA). Motor Vehicle Safety: Employers. No date shown; accessed October 2026. osha.gov/motor-vehicle-safety/employers
- 3Hickman JS, Hanowski RJ (Virginia Tech Transportation Institute). Use of a video monitoring approach to reduce at-risk driving behaviors in commercial vehicle operations. Transportation Research Part F 14(3):189 to 198. 2011. sciencedirect.com
- 4Boyle LN, Guo EH, Hammond RL, Hanowski RJ, Soccolich S. Performance Assessment of an Onboard Monitoring System for Commercial Motor Vehicle Drivers: A Field Operational Test. FMCSA-RRR-15-019. Federal Motor Carrier Safety Administration. November 2016. rosap.ntl.bts.gov/view/dot/31605
- 5Mase JM, Majid S, Mesgarpour M, Torres MT, Figueredo GP, Chapman P. Evaluating the impact of Heavy Goods Vehicle driver monitoring and coaching to reduce risky behaviour. Accident Analysis & Prevention 146:105754. 2020. sciencedirect.com
- 6Wouters PIJ, Bos JMJ. Traffic accident reduction by monitoring driver behaviour with in-car data recorders. Accident Analysis & Prevention 32(5):643 to 650. September 2000. sciencedirect.com
- 7Insurance Institute for Highway Safety (IIHS). Speed (research area, including intelligent speed assistance). Updated July 2026. iihs.org/research-areas/speed
- 8AAA Foundation for Traffic Safety (Ebert JP, McDonald CC, Xiong RA, et al.). A Randomized Field Trial of Smartphone-Based Feedback Designed to Encourage Safe Driving. April 2025. aaafoundation.org/research/…
- 9Albadawi Y, Takruri M, Awad M. A Review of Recent Developments in Driver Drowsiness Detection Systems. Sensors 22(5):2069. 2022. mdpi.com/1424-8220/22/5/2069
- 10Federal Trade Commission (FTC). Policy Statement of the Federal Trade Commission on Biometric Information and Section 5 of the Federal Trade Commission Act. May 18, 2023. ftc.gov (PDF)
- 11Occupational Safety and Health Administration (OSHA). Clarification of OSHA’s Position on Workplace Safety Incentive Programs and Post-Incident Drug Testing Under 29 C.F.R. § 1904.35(b)(1)(iv). Standard interpretation. October 11, 2018. osha.gov/laws-regs/standardinterpretations/2018-10-11
- 12Ayas S, Donmez B, Tang X. Drowsiness Mitigation Through Driver State Monitoring Systems: A Scoping Review. Human Factors. Published online November 20, 2023. doi.org/10.1177/00187208231208523
- 13Electronic Code of Federal Regulations. 49 CFR 390.36 (Harassment of drivers prohibited), 395.8 (Driver’s record of duty status) and 395.26 (ELD data automatically recorded). Up to date as of September 25 to October 1, 2026. ecfr.gov
- 14New York State Senate. New York Civil Rights Law, Section 52-C (second section with that number): Employers engaged in electronic monitoring; prior notice required. Most recent revision shown May 13, 2022. nysenate.gov/legislation/laws/CVR
- 15California Privacy Protection Agency. Frequently Asked Questions (California Consumer Privacy Act). No date shown; accessed October 2026. cppa.ca.gov/faq.html
This article reflects the safety priorities of fleet telematics and human factors research, with attention to privacy and employment law. The thresholds, script, checklists, tables and ladder are original educational examples, not standards. No product or vendor is endorsed. It is general education, not legal, HR or professional advice. Laws, research and technology change, so verify with official sources and qualified counsel. Reviewed October 2026.
Keep learning
Safe Driving Resource Center
Guides, self-checks and articles for drivers, parents of new drivers, and families recovering after a crash.
