Driver Fatigue Q&A: Predicting Risk Before the Shift Starts

When we hosted The Future of Fleet Safety and Its Impact on Operational Efficiency for the NAFMP Solutions Series, the audience sent in 47 questions. We expected questions about technology. We got those, but the sharper ones were about trust: who sees the score, what happens to a driver who gets flagged, and what to do when the ELD says a driver is legal but the data says he is tired.

DOWNLOAD NOW: THE EHS BUYER'S GUIDE TO READI

Those questions get to the real issue. Most fleets do not have a visibility problem anymore. Cameras and telematics show exactly what happened. What fleets lack is a way to act before it happens, without turning safety into surveillance.

Below, Tomas Kuras (VP of Revenue) and Marco Parlato (Director of Product Management) answer the questions we can speak to with evidence: from our predictive fatigue platform, Readi, and from a controlled pilot at a large U.S. logistics fleet where the drivers using Readi had 42% fewer fatigue-linked in-cab telematics events than a matched control group.

Where a question belongs to a physician, regulator or lawyer, we have left it for them.

Part 1: How predictive fatigue technology works

What kind of technology can help manage fatigue? Are AI cameras helpful?

Yes, AI cameras help, and they answer a different question than prediction does. Fatigue technology falls into three groups by when it acts:

In-cab cameras detect head nods, eye closure and lane drift. That is valuable, but by the time those signs show, the driver is already impaired at highway speed. A 2–4 second microsleep at 65 mph covers roughly 190–380 feet with no one in control.

Prediction moves the intervention point upstream. Readi gives dispatchers an hour-by-hour fatigue forecast for each driver's full upcoming shift, so they can act before the engine starts. It is designed to work alongside cameras, not replace them: prediction is the first line, the camera is the last line.

How does the "Readi watch" identify fatigued drivers?

There is no watch required. Readi does not depend on wearables or new hardware. It estimates each driver's sleep from work and rest patterns already in your systems, then runs that through the SAFTE™ biomathematical model to forecast alertness across the shift.

That was a deliberate design choice. Wearables create rollout cost, charging and compliance problems, and privacy concerns that slow adoption.

For a carrier with no wearables and no app budget, how much works from the schedule alone? What data do you use?

In our logistics pilot, fatigue forecasts were generated from existing ELD data, with no wearables and no new hardware. That is the setup that produced the 42% result.

The inputs are:

• Schedule and duty-status data pulled from your ELD or scheduling system: when a driver was on and off duty, and when the next shift starts.
• A one-time questionnaire that sets a baseline for each driver's typical sleep.
• Circadian timing: the time of day each hour of the shift falls on, which matters as much as how long someone has worked.

From there, our SleepML model estimates when and how much each driver likely slept. SAFTE™, a model developed for the U.S. Army and used in DOT and FAA research, turns that sleep history into an hour-by-hour alertness forecast.

How does the model account for individual driver differences?

Each driver gets a personalized sleep profile. Readi matches individual work and rest patterns against a database of more than 10 million anonymized sleep records, then refines the profile as the driver's own history builds.

The forecast reflects that driver's recent sleep pattern, typically the prior ~10 days, not a fleet average. Two drivers on the same route and the same shift can get very different forecasts.

Can this integrate with hours-of-service software?

Yes. Readi pulls duty-status data automatically from existing ELD platforms such as Geotab, Motive and Samsara. There is no manual entry and no new login for drivers.

The pairing matters. HOS data tells you how long someone has been on duty. Readi adds when they are likely to be at risk within the shift.

Do you see smaller fleets using this?

The principles apply to any fleet: sleep opportunity and time of day drive fatigue regardless of fleet size. The software delivers the most value where there is a dispatcher or supervisor who can act on a forecast, such as adjusting a route, moving a break or having a check-in conversation.

Our deployments today skew toward mid-size and large fleets with dispatch teams and dynamic schedules. Smaller carriers can start with the countermeasure side: build a non-punitive fatigue reporting culture and plan schedules around circadian low points.

Part 2: Accuracy, privacy and the driver experience

How accurate is it? What about false positives?

We cannot speak to the false-positive rates of other devices on the market. We can tell you how ours is validated and what that means in practice.

• Sleep estimation: SleepML, trained on more than 1,800 workers, estimates sleep with over 90% sensitivity and under a 17% false-positive rate, with a mean error under 10 minutes.
• Fatigue modeling: SAFTE™ is a peer-reviewed biomathematical model developed for the U.S. Army and used in DOT and FAA research.
• Real-world correlation: across our fleet data, when a driver's forecast ReadiScore falls below 70, harsh braking events run 8.5× higher and speeding events 4× higher than for alert drivers.

One framing note: Readi is not a pass/fail test at the check-in window. It forecasts risk across every hour of the shift, so a dispatcher sees when risk will peak, not just a yes/no at departure. That changes the response from "pull the driver" to "plan the break before 3 a.m."

How do you avoid flagging well-rested drivers who are just waking up before the shift?

Because the forecast is built from sleep history and time of day, not from how someone looks or performs at check-in, grogginess in the first minutes after waking does not drive the score. A driver who slept well and starts a daytime shift forecasts as low risk even if he is still on his first coffee.

This is a real weakness of point-in-time tests: they measure a moment, and the moment right after waking is a bad one to measure.

Does it use identifying or protected personal information?

Readi works from schedule and duty-status data plus a one-time sleep questionnaire. There are no cameras, no biometrics and no in-cab monitoring. Supervisors see a fatigue forecast so they can act; they do not see how a driver spends their time off duty. No one in the organization can see an individual’s sleep hours.

We would encourage any carrier to ask every safety vendor the same question: what data leaves the cab, who can see it and what it can be used for.

What safeguards keep scores from being used punitively?

This was the most important design question in the pilot, and the answer is part policy and part product.

• Volunteer-first enrollment. The pilot had zero mandates.
• Non-punitive countermeasures. A low forecast triggers support: a break moved earlier, a schedule adjustment, a check-in conversation. It never triggers discipline.
• Written fatigue policy. Fatigue is physiological, not a performance issue. Fleets should update policies and work instructions to say so before rollout.
• Framed as care. Leadership presented Readi as looking out for drivers, not watching them.

How well do drivers embrace the technology?

In the pilot, well. Supervisors acted on 1,722 high-risk shifts before departure, and countermeasure utilization reached up to 81%. One supervisor summed up driver feedback at the debrief: drivers liked that the company was looking out for their safety, not just grading how they drive.

Acceptance comes down to what the tool asks of drivers. Readi asks almost nothing: no device to wear, no app to open every shift, no camera pointed at them.

Part 3: Putting fatigue data to work in your safety program

How do we turn fatigue reporting from a perceived violation into a safety metric?

Stop treating fatigue as something a driver admits to and start treating it as something the operation forecasts. When the schedule, not the driver, is the source of the flag, reporting stops being a confession.

In practice that means three shifts:

1. Measure exposure, not admissions. Track how many forecast high-risk hours your schedules create each week. That is a number operations can own.
2. Reward the response. Count countermeasures delivered, not drivers flagged. The pilot tracked utilization daily and shared wins openly, including leaderboards.
3. Close the loop with event data. Show drivers and supervisors that the events your cameras capture fall when fatigue risk is managed. Visible results built trust faster than any memo.

How can predictive fatigue data fit into a carrier's safety management system?

As a leading indicator alongside the lagging ones you already track. Most SMS dashboards are built on events, citations and claims, all of which describe what already happened. A fatigue forecast adds the one input that arrives before the shift.

It also adds documentation. Every forecast and every countermeasure is time-stamped, which gives you an auditable record showing the fleet identified risk and acted on it.

HOS says a driver is within limits, but the data shows fatigue risk. Which do we trust?

Both, because they measure different things. HOS caps hours on duty. It cannot see whether a driver slept four hours or eight before a legal shift, or whether that shift crosses the 2–6 a.m. circadian low.

A legal driver can be a fatigued driver. The practical answer is to treat HOS as the floor and the fatigue forecast as the trigger for a conversation: check in with the driver, move a break into the forecast risk window, or adjust the route.

When should a dispatcher stop a driver from starting a shift?

That threshold is a company policy decision, and it should be written down before you roll out any tool. In our experience, standing a driver down should be the rare end of a ladder, not the default response:

1. Check-in conversation before departure
2. Break timing moved into the forecast high-risk window
3. Route or load adjustment to shorten exposure
4. Schedule swap or later start
5. Stand-down, guided by policy and the driver's own report

   Because Readi forecasts the whole shift, most interventions can happen early and lightly. Pairing the forecast with the driver's self-report keeps the decision human.

   Which leading indicators should we watch, and how do we know our fatigue barriers work?

   Track a short set, weekly:

   • Forecast high-risk shifts and hours per week, by terminal and lane
   • Countermeasure utilization: the share of flagged shifts that received an action
   • Fatigue-linked telematics events per driving hour (harsh braking, forward collision risk, lane drift)
   • Comparison against a control group or pre-rollout baseline

   The last point is how you prove the barrier works. In our pilot, the Readi group's in-cab events per driving hour fell 15% while a matched control group's rose 27% over the same period, a 42% gap.

   How did you create buy-in with drivers and management?

   Four things made the difference:

   • Executive sponsorship from safety and operations from day one. Health & Safety and Operations shared ownership.
   • Answering "what's in it for me?" for drivers directly, before rollout, not after complaints.
   • Dispelling privacy myths proactively: what data is used, who sees it and what it will never be used for.
   • Simple countermeasure processes with flexibility for weekends and holidays, plus regular feedback loops on how the pilot was going.

   No technology replaces the conversation between a supervisor and a driver. The best it can do is make that conversation earlier and better informed.

   Part 4: Hours of service and the science of fatigue

   Why doesn't the 11/14-hour rule better align with circadian rhythms?

   HOS rules regulate quantity: how many hours a driver may be on duty and how much rest separates shifts. Fatigue depends just as much on timing. An 11-hour shift starting at 6 a.m. and one starting at 10 p.m. are equally legal, but they carry very different risk, because the second runs straight through the body's 2–6 a.m. low point.

   Rules have to be simple enough to enforce across millions of drivers, so they cannot be personalized. That is the gap a fatigue forecast fills: it applies circadian timing and each driver's recent sleep to a schedule that is already legal.

   Does a 24-hour or extended shift change the picture?

   Considerably. Alertness declines with time awake, and the decline is not linear: it compounds once wakefulness runs past the normal day, and it compounds again when the extended period crosses the overnight low. Sleep debt from the days before adds to it.

   This is also why risk can climb during peak season even when every driver stays within HOS limits. In our pilot, the control group's fatigue-linked events rose 27% through a demanding peak period.

   How does fatigue relate to hours behind the wheel? Does excessive wait time matter?

   Hours driving is only one part of it. The bigger driver is hours awake. Two hours waiting at a dock still count toward time awake and still erode alertness, even though no miles move. A driver who waits through the afternoon and then drives into the evening may start the drive already well into a fatigue curve.

   That is one reason detention time deserves a place in fatigue conversations, not only in billing ones.

   Would adding an extra hour to reach home or a hotel help?

   We will leave the regulatory question to regulators. From a fatigue-science standpoint, the effect depends on what the hour enables. If it lets a driver get to a place where they will sleep better, it can improve the next shift. If it extends a shift that is already running into the circadian low, it adds risk at the worst possible time. The answer is driver- and time-specific, which is what a forecast is built to show.

   How does fatigue lead to crashes?

   Fatigue degrades the skills safe driving depends on before a driver ever nods off:

   • Cognitive tunneling: visual scanning narrows, so drivers stop anticipating traffic, terrain
     • and road layout.
     • Late reactions: slower perception leads to sudden, reactive maneuvers, which show up as harsh braking events in telematics data.
     • Microsleeps: brief lapses of 2–4 seconds, often without the driver noticing, during which a truck at 65 mph travels the length of a football field or more.

     This is why fatigue shows up in your camera and telematics data long before it shows up in a crash report. At a ReadiScore below 70, harsh braking runs 8.5× higher than for alert drivers.

     What are the main myths about fatigue? How do we explain them to drivers?

     The one we hear most: "I can push through it." You cannot postpone fatigue. Caffeine, cold air and loud music mask the feeling briefly; they do not repay sleep debt. The only fix is sleep.

     Other common myths:

     • "I'll know when I'm too tired." People are poor judges of their own impairment, and microsleeps often go unnoticed.
     • "Experienced drivers handle it better." Experience improves skill, not biology.
     • "Legal means rested." HOS compliance says nothing about how well a driver slept.

     What works best when explaining fatigue science to drivers is making it personal and non-judgmental. Show drivers their own forecast pattern, talk about fatigue as physiology rather than performance, and let supervisors open the conversation with "How did you sleep?" rather than "Why were you braking hard?"

     Part 5: The pilot, and how success was measured

     How was the success of the pilot measured?

     Against a matched control group, using event data the fleet already trusted. The design is what makes the result defensible:

     Element	Detail
     Fleet	Large U.S. logistics fleet, high-volume feeder operations
     Readi group	95 drivers
     Control group	85 matched drivers ("twins"), paired on route type, schedule and tenure
     Conditions	Same operating period and geography, including peak-season pressure
     Inputs	Existing ELD data only; no wearables, no new hardware
     Metric	Fatigue-linked in-cab telematics events per driving hour
     Analysis	Performed independently by the fleet's existing video telematics provider

      

   • The result: events per driving hour fell 15% in the Readi group and rose 27% in the control group over the same window, a 42% difference. The categories that dropped were the ones camera programs already work to reduce: harsh braking and acceleration, forward collision risk and lane drifting.

     The pilot also had operational success criteria that mattered as much as the number: forecasts generated for every participant, a working ELD integration, countermeasures approved company-wide, and positive driver and dispatcher feedback.

     The lesson for any fleet evaluating a safety technology: insist on a control group and measure with data you already collect. Otherwise seasonal swings can make any tool look good, or bad.

     Closing: the questions that were really about timing

     Read the 47 questions together and one theme stands out. Fleets are not short on data about what went wrong. They are short on a way to act before it does, in a form drivers will accept.

     Three takeaways from the session:

     1. From detection to prediction. The next step in fleet safety is intervening before the shift, not detecting more during it.
     2. Leading indicators in practice. An hour-by-hour view of the upcoming shift reduces the events your cameras already track, making those investments work harder.
     3. Trust decides adoption. Tools built to support drivers, not surveil them, are the ones that get used.

     Want to see what this looks like for your fleet? Ask for a 15-minute sample forecast walkthrough: an hour-by-hour fatigue forecast across a real shift, in a fleet like yours. Or ask us for the pilot detail: methodology, twin-group analysis and operational findings. [Link: fatiguescience.com]

     Missed the live session? Watch the on-demand recording