Trucking is becoming accustomed to using data, and lots of it, to help guide business decisions or even to trigger automated responses to certain events. Is a delivery going to be late because the driver doesn’t have enough hours left to finish the trip? Is a tire underinflated? Is there cheaper fuel along the assigned route? Is one APU more cost-effective than the other?

Today a fleet can see it all, make decisions, and act day in and day out on the information flowing through the organization. So far, that information has been largely based upon proprietary data about the workings of the individual business and the environments in which it operates. What happens, however, if data from a huge number of carriers is combined into much larger data pools that can also be mined for actionable information?

According to some industry suppliers and watchers, that is exactly where trucking is now headed. The jury is still out, though, on what that might mean for the industry as a whole.

Not every company is prepared, at least not yet, to set aside privacy concerns and splash into the common data pool, either. After all, it takes a certain level of trust just to begin the process of confidentially sharing your data with a telematics provider.  

Telematics suppliers occupy a unique spot in the information technology ecosystem, according to Mike Scarbrough, CEO of Nextraq. Using data to drive decisions involves three basic steps, he notes, collection, transformation and communication. “Telematics providers are the operators of all this data, the transformers who turn raw data into information that other can use to guide important business decisions,” Scarbrough says.

“In order to do this, you need to have a business partner you really, really trust,” he adds, “because underneath the covers, we are making lots of small decisions about data [that can make a big difference to you in the end.] For example, suppose I tell you that one of your drivers has gone over the speed limit twenty times.  What will you do?  What if I tell you that the average for all drivers was one hundred times? Or two times?

“Adding context changes your view rather dramatically,” he continues. “I could give you just the raw data, the fact of one driver speeding twenty times, knowing that you may be misled to a poor decision, but none of us want people to make bad decisions.

“There is a natural human resistance to sharing information, such as on social networks,” Scarbrough notes. “Data helps to take that reluctance away. It lowers our resistance to sharing information. Everybody would like things to be less subjective and more fact-based in order to make better decisions. Everybody appreciates that.”

“There has been a remarkable growth in databases and in the technology to analyze that data,” says Corey Catten, CTO of inthinc. “With a small IT staff, we have access to thousands of processors for short bits of times. Parallel processing means that 250 servers can be working for 15 minutes [to analyze data] instead of using eight servers running all night to do the same thing.”

Think about what happens if you combine data from many carriers and many sources, Catten asks. Insurance companies today take all their actuarial data and give it to an organization that combines it, sanitizes it and gives it back to the insurance companies so that they can compare themselves to others. How would this model work for trucking? What would a fleet gain if they put their data into the pool?

“We are just in the data collection phase right now,” he adds. “We can say to a customer (without violating any privacy policies) ‘we know you can do better because others are doing better.’”

Big data benefits

PeopleNet is also exploring the benefits of looking at larger data sets for information that can benefit its customers, according to Rick Ochsendorf, senior vice president operations for PeopleNet. Like other telematics providers, the company processes enormous amounts of data everyday for fleets. “We are looking at tools that analyze all this data, creating different pools of data that fleets are willing to share,” he says. “Customers would like to know ‘How efficient am I along a certain lane? What do I have to do to improve?’”

“Big data is here,” Ochsendorf observes. “You have to be willing to share to get something back. In the past, there were so many variables baked into the numbers [that it was virtually impossible to get an apples-to-apples comparison].

“There is so much data, big data, at our fingertips, but you have to really parse it down to make sense,” he adds. “[Otherwise, you can misunderstand what you are seeing and make some bad decisions as a result.]”

Tom Fansler, president of Vusion, also sounds a cautionary note when it comes to talking about data and advanced analytics, as exciting as the potential benefits it can deliver are. “This is an interesting time,” he observes. “People use the term ‘analytics’ in all sorts of different ways.

“One of the current trends is data aggregation, but simply having silos of data connected does not necessarily make it actionable or useful because much of the data is just not designed for statistical rigor,” he explains. “You have to describe things in a similar manner even if they have not been described that way before to get beyond simple groupings to the depth required to make predictions you can act on.

“For example, Fansler offers, “There is a very strong relationship between improper lane changes and accidents, but you have to go one step further and ask how often lane changes were the cause of accidents.  You also need to determine how often drivers were cited for improper lane changes not tied to accidents before that data becomes useful as a predictor.

“If you are doing predictive analytics around driver behavior across several companies, the same thing applies, he continues. “You have data on a collection of driver behaviors, but there is also a collection of variables around the carriers that has to be taken into account. If I describe my fleet as a ‘regional hauler,’ is that the same as what other carriers [in the data pool] are saying when they describe themselves as ‘regional haulers’?  Or is there some other metric we can use to identify type of fleet that will sidestep the problem?

“Working to standardize data is a precursor to predictive analytics,” Fansler says. “Lots of companies talk about data integration, but what they are really doing is just grouping together things that have the same field name. How fleets describe sudden accelerations or stops is a good example. While companies may all track those events, what a sudden deceleration or stop really means depends upon how the recording device is configured to capture those events.

“The same is true for speeding,” he continues. “One fleets may call 65 mph speeding, while another sets the trigger at 68 mph. All those events show up in the data bucket called ‘over-speeding,’ but they are not all identical.  In each of these examples, you end up looking at apples and oranges as if they were the same. Ninety percent of analytics is data preparation.

“In order for analytics to work, you also need a partnership between the provider and the carrier,” Fansler adds. “The client has to be invested in the process or we are stuck. Companies that want [to use predictive analytics] have to be committed if they really want to dive into even just the data about their own equipment or their own drivers.”