Data-driven safety

In pursuit of operational efficiency, fleets collect a staggering amount of data from their vehicles, their drivers, their terminals and even their customers. They have invested heavily in information technology in attempts to cut costs through increased productivity, and to a great extent that investment has paid off. The most progressive fleets are learning that these new tools and the data collected

In pursuit of operational efficiency, fleets collect a staggering amount of data from their vehicles, their drivers, their terminals and even their customers. They have invested heavily in information technology in attempts to cut costs through increased productivity, and to a great extent that investment has paid off.

The most progressive fleets are learning that these new tools and the data collected by them can also be leveraged in other areas, unlocking additional value from IT investments and resources already in place. Among the many potential applications for mining a fleet's collected data, one in particular has already begun to show promising results. Risk management at the fleet level is ready to take a giant leap forward, making safety efforts both easier to manage and more effective.

Driven by data and the IT power to analyze it, this latest development in safety management is being called “a real breakthrough,” “a new frontier” and “the wave of the future” by safety experts with decades of experience in trucking. More importantly, fleets of all sizes can now realistically close the safety loop that starts and ultimately ends with the driver, a safer driver.


The first step sounds simple — gather all the safety-related files in one place where it can be organized and managed effectively, and where you can lay your hands on it when you need to document your safety management efforts.

“Fleets have to manage a storm of stuff,” says Steven Bryan, CEO of the new web-based risk management service Vigillo. With background checks, drug and alcohol testing, CDL status and other related files, “there are about a dozen documents just related to driver qualifications that have to be captured and managed consistently.”

Add accident and incident reports, a fleet's ongoing training requirements, and other safety programs that require documentation and the job of keeping on top of all that paperwork can be overwhelming.

That's where IT shines. “Once you define a process, the technology won't forget; it won't let [a requirement] be ignored,” says Bryan.

Vigillo approaches that task with an online “software organizer.” The system uses a fleet's own documents to build and monitor a “safety process.” Enforcement is both automated and consistent, with all employees receiving notification when it comes time to update a record or attend training. With all the records in one place, the system can also provide managers with a dashboard to let them quickly assess the fleet's overall safety status, as well as alert them to issues with individuals.

But just organizing safety-related data is only the first step. “We see it as an operating system for risk management,” says Bryan. “Data from all sorts of independent [fleet] sources will flow through our system for reporting and analysis. The analytics and metrics that come out of it will be a gold mine.”

Others are also looking at web-based online models for tracking and managing safety-related data already collected by fleets. J.J. Keller & Associates, for example, is now Beta testing a service it calls FleetMentor. Aimed at small and midsized fleets, it supplements those required documents with additional accident and claims data that can then be used to look for patterns, says Jacqueline Jurmu, design manager for the new service. Alerts and reminders for upcoming or past due driver-compliance events are a core part of its tools, which will be gradually expanded to include a range of safety auditing items, she says.

Documents provide a valuable baseline for risk management, but they're historical — a record of what's happened that's received well after the event. Mobile communications systems and related onboard computer (OBC) hardware have long provided the ability to add actual driver and vehicle operating information to the mix. But use of that on-the-road information has been limited by two factors: Wireless costs generally make it impractical to download all of the information in real-time; and if you wait until the truck returns to a terminal to download the data and sort it, you're dealing with issues that are days — if not weeks — old.


In the last few months, however, developers of fleet management systems have begun introducing products or modules that not only allow fleets to unlock the value of that mobicomm data, but also to use it in ways that make sense for their own particular operations.

One of the newest is the IES Aware module for Innovative Enterprise Software (IES) from Inovating Computing Corp. (ICC). Released initially for use with Qualcomm's OmniTracs and OmniVision mobile communication systems, IES Aware alerts fleets about driver behavior based on each fleet's own rules in “near real-time,” says Ernie Betancourt, president of ICC.

More than just a filter for data from the Qualcomm system, the new ICC product ties vehicle-generated information like speed and GPS location to other fleet data to identify violations of parameters set by the fleet. “It uses exception reporting [to issue alerts], and it can escalate the alerts as driver behavior dictates,” says Betancourt.

Currently, fleets can set their rules using five variables: restricted driving hours, speeding, number of hours driven in a defined period, continuous hours driving, and check calls. The marriage of vehicle and fleet data means the rules created to trigger alerts can be quite complex. For example, a fleet might choose to allow driving during its restricted hours if the driver has gotten a prescribed amount of sleep beforehand, Betancourt points out. Or tying vehicle speed and location to a route database means the fleet can set speeding triggers based on actual posted road speed limits rather than a general overall top speed limit.

Coming at it from another angle, EBE Technologies has also recently added near real-time alerts based on mobicomm data, but extends that information to its core SHIPS business process software to integrate it with driver documentation and compliance records.

“We're taking data from multiple systems, either hosted or on-site,” says Cindy Nelson, senior director of marketing and business development. “We rely on the OBC and mobicomm for driver performance information and fleet business rules to trigger actions based on driver activity.” With the company's roots in document imaging, the new module can also push required documents or electronic files out to drivers over the OBC.

“Fleets can leverage data from dispatch, accounting and mobicomm to automate the process of tracking and managing all driver information, triggering the appropriate workflows to keep drivers in compliance,” says Nelson. “It can also be used to correct unsafe driver practices in real-time and can interface with online safety and training tools, as well as auditing services. It can even take a driver out of the dispatch pool if a key document is missing and send violation letters with online signature capture to an OBC.”

Moving up a level, it provides a manager with a dashboard view into safety, showing fleet performance for specific events such as hard braking, excessive idling or hours-of-service violations. More importantly, the dashboard view “has a workflow behind it creating an audit trail that can impact safety ratings,” Nelson adds.

Using data already being collected, this approach creates a safety circle. “Mobicomm pushes data from the driver to the fleet, EBE processes it, the fleet accesses it via the Internet and integrates it with its business processes, and the appropriate actions are communicated back to the driver,” Nelson explains.


Data has been used for a number of years to create models of behavior patterns in an attempt to identify the riskiest drivers before they have a serious accident. Zurich, one of trucking's major insurance providers, launched a service it called Virtual Risk Manager in 2003 that records and weights a variety of safety infractions to do just that.

“We take moving violations from motor vehicle records, inspections from carrier profiles and crash information from claims history, and integrate it with a few other safety indicators to create predictive models,” says Jim York, assistant vp of technical services, Zurich Services Corp. Risk Engineering. Coaching for identified risky drivers and documentation for all fleet responses to address the identified issues round out the service.

“Today, we have six insured clients and over a quarter of a million drivers on the system,” says York. “The fleets that use it aggressively have seen a dramatic impact [on safety performance].”

Using such historical data “gives us a two-dimensional map [of predictive driver behavior],” says York. “It indicates hills and valleys with contour lines, but you don't really appreciate the elevations involved until you actually see them.”

But now there's an effort to “overlay real-time operational data to create a 3D map,” he says. “History is not always the greatest indicator of the future, so overlaying real-time operational data is a real breakthrough in predictive modeling.”

By real time, York means vehicle data captured by the OBC in the last 60 days; for example, hard braking or acceleration and speeding. Historical indicators in the Zurich system are a rolling three-year history of moving violations, accidents, roadside inspection violations and other data from various records.

Unlike exception reporting systems offering real-time alerts to dangerous driver activity, predictive modeling requires collecting operational data from all drivers. “The key [to accurate predictive modeling] is consistency and completeness,” says York. “You have to capture the rolling data consistently and capture the operational data for all drivers, not just those who raise red flags.”

Completeness, however, doesn't mean every bit of data collected by a fleet. “There are a number of [data sources] you could add — maintenance records, for example,” says York. “But we found with Virtual Risk Manager that adding more data reaches a point of diminishing returns.”

Zurich is working with a group of its fleet customers right now to develop its 3D map and expects to have prototype models within the next six to twelve months. “But early adopters are seeing anecdotal results that are very encouraging,” says York. “We wouldn't be making this kind of investment if we didn't expect success.”


While predictive modeling efforts by Zurich and others such as Fleet-Risk Advisors (see box) are built on real-time data already collected by the OBC, others are looking to augment that information by adding other capture devices and building broader databases that reach beyond a single fleet's activities.

IVOX Inc. has built what it calls “a device agnostic technology” to deliver predictive “driver risk scores,” which president Craig Lotz describes as a credit score for safe driving. While the system can use data from any onboard recording device, “our ability to be confident in a driver score is pegged to the amount of data we collect,” he says. “Our preference is for a device that records second-by-second GPS data and sub-second data from a three-axis accelerometer. Otherwise we can miss events that are important to driver behavior risk analysis.”

The company looks at speeding, aggressive driving patterns, time of day and other operational factors to create the score. Then they map the score to claims to look for common behaviors tied to specific types of accidents.

“We've built a database for all commercial vehicles and can segment for specific vehicle type to create specific scores for garbage trucks or pickups, for example,” Lotz says.

As with all risk management systems, reporting is the final link to improving safety performance. “Our reports are web-based,” says Lotz. The company can rank all fleet drivers by their scores for top-level managers or limit access to a branch or just show managers, and even provide access to individual drivers with pointers on how they can improve their scores.

“We plan to create a driver training effectiveness index, too, so fleets can see how a program affects specific driver behavior,” says Lotz.

Another approach to extending data capture for predictive modeling places a video event recorder in the truck cab. The DriveCam Inc. unit is constantly recording sound and sight in the cab and out through the windshield, using an accelerometer to save the recorded information just before and after an event is sensed. The data is then downloaded by WiFi when a vehicle returns to a terminal or sent during off hours over the Sprint cellular network.

While fleets have been using the video as coaching tools to show drivers their risky behaviors and suggest alternatives, the longer range goal is to augment that service with predictive analysis.

With 80,000 units in vehicles, DriveCam receives the stored event data from 22,000 vehicles for analysis, says Byron Cook, vp-data mining and analytics. “We're nearing six million events in our data warehouse, and all are scored for the trigger event and outcome, and segmented into 12 vehicle types,” he says.

The data warehouse has just allowed DriveCam to launch a Risk Center, identifying risky behavior patterns based solely on analysis of the video and audio files. “We've really just been digging into that data for the last few months, but we've built a [predictive] model that will get smarter with time,” says Cook.

So the elements are in place to use your fleet IT resources to bring consistent management to safety-related documents, to identify risky driver behavior as it occurs, and even predict who should receive targeted attention from your safety department. “But what do you do once you've identified the risky driver?” asks Bruce Weiss, exec. vp of Instructional Technologies Inc. (ITI). “The one thing you can't do is nothing,” Weiss says. “You have to correct the behavior with training.”


Taking advantage of a new generation of powerful OBCs and wireless communications, fleets should consider bringing that training to the driver. “Whether it's managing space, speed, fuel or so on, you can correct the behavior on the spot rather than waiting to bring them back into a terminal,” says Weiss, whose company has launched an in-cab version of its Pro-TREAD Internet-based training system.

While in-cab safety training might not be right for every fleet, Weiss sees it as an effective and efficient way to close the loop that started right there in the cab.

The process starts with using data from the truck to identify risky behavior. “Then it's sent over the mobile communications system to the data mining service,” Weiss says, where it's blended with other historical data to create a predictive model for targeting risk management efforts. In the final leg, ITI pushes out the appropriate training for the identified risk and notifies the risk management system once that training has been completed.

Running in tandem with development of the other elements in this scenario, ITI expects to have that final link ready for initial deployment by the third quarter, Weiss says.

Whether you see it as unlocking value from existing data or a true breakthrough in risk management, it's a vision that's looking more and more like the future of truck safety.

Predicting in the real world

Think that using existing business data to create a predictive safety management system is just another example of IT vaporware? C.R. England, the country's largest refrigerated truckload carrier, doesn't because it's already doing just that. For some time, the fleet had looked at developing a way to use its data to identify high-risk drivers, but it could never justify the investment in IT resources needed to carry off the project, says Chad England, vp of recruiting, training and safe driving. Instead of giving up, it decided to look for help from a third party, two actually: Qualcomm Enterprise Solutions and FleetRisk Advisors.

“We started implementation in April 2007, and by October we had our [predictive] performance monitor in place,” says England. The model is used to create predictive performance scores for each driver. The scores are used by five safety managers to set training priorities and other countermeasures intended to lower those scores for individual drivers and the fleet overall.

“The key to modeling is to prevent any bias from interfering with the data,” says England. “So we draw from every database we have — equipment, driver characteristics, violations, dispatch, miles, fuel stops. We're looking at over 3,000 pieces of data to make up a predictable model.”

On the recommendation of FleetRisk Advisors, C.R. England also added fatigue scoring from Circadian Technologies. A pioneering technology company in the field of fatigue management, Circadian analyzes hours-of-service logs to predict an individual driver's likelihood of having a fatigue-related accident, and suggests schedule changes to reduce that likelihood.

“Predictive modeling works with whatever data you have, but we've already identified 60 data types we want to start collecting for our next-generation models,” says England. “It's simple stuff. Our first model didn't include citation type, for example, because we just didn't have a data field for it. So we've added one.”

Although it's too early to attribute it to predictive modeling, C.R. England has already seen a significant drop in preventable accidents even though fleet miles are up. “I won't declare victory yet, but we have to find a way to target our efforts so we get the most impact out of them,” says England. “I am convinced, though, that it's the wave of the future for safety management.”

View more Fleet Owner news relating to trucking safety, trucking regulations and driver awareness.

TAGS: Safety
Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.