The data is everywhere.
It’s in our old-school paperwork and office software, and it flows in at unprecedented rates from our smart devices. But with so much information sourced from a multitude of locations, it’s easy to feel overwhelmed—and difficult to piece it together into a cohesive picture of how to improve safety in transportation.
“You’ve got all this paper, and spreadsheets and internal files, where all of your data lives, and it’s the devil trying to wrangle it all together,” said Hayden Cardiff, chief executive officer of Idelic. “And then on the flipside you’ve got all your third-party systems that you work with, your ELDs, your cameras, the FMCSA portal, HR systems, training systems … so the issue becomes, ‘I’ve got all this data, what do I do with it?’ Because if you can’t get it into one place, it’s hard to start to aggregate and really get anything meaningful out of it.”
Cardiff’s suggestion during Tank Truck Week 2019 was to find a way, perhaps using his software-as-a-service company’s Safety Suite solution, to pool all that data, then utilize machine learning techniques to generate actionable insights and analytics on drivers, terminals and overall operations beyond anything previously possible.
Terry Kolacki, director of environmental, health, safety and sustainability for Superior Bulk Logistics, says this approach to aggregating data, identifying trends and proactively preventing accidents, is working wonders for his company.
“You’re able to see your riskiest drivers, you’re able to predict what type of training to give to them, if they take their training, and if they keep having the same problems on the training that’s administered,” Kolacki testified. “It’s a great way to put logging violations together with out-of-service inspections, 14-hour violations, accidents, speeding, hard-braking events—these are all classified with a driver with their score.
“So it’s very easy to see who your problem drivers are very quickly, and know who you need to coach and train.”
Taking action
Cardiff discussed data and trucking, using data effectively, the benefits of increased visibility, artificial intelligence in accident prevention, data and safety culture, and how to get started during his presentation on AI in transportation.
“What we really want you to do is come out of here with some actionable steps you can take back to your fleet and actually start to do something with,” Cardiff said. “I don’t want this to be high-level pie in the sky.”
The intent behind aggregating data and using predictive analytics is to increase visibility, which, in turn helps managers better understand safety data, identify at-risk drivers, and target those drivers with coaching and training designed to prevent accidents, instead of waiting until after a preventable accident to lay into them.
“You have all these disparate data sources, but when you pull them together, you start to paint a much clearer picture of what’s going on,” Cardiff said.
Many types of remote connections are out there, including electronic data interchange (EDI), file transfer protocol (FTP), WebRTC real-time web communication, or “web calls,” and application protocol interface (API), which is a direct, integrated connection between a third-party vendor and a company’s internal systems.
Company’s also can attempt to pull in data themselves, but however it’s done, the first and most important step in gaining actionable insights from machine learning is to bring the data together in one manageable location.
For Superior Bulk Logistics, that means using Idelic’s platform to simplify reporting from terminals across the country, including combining data found in accident and worker’s compensation forms, cargo liability forms, spill forms and more.
“This system allows all the terminals to have this site for their terminal, and go into it and put in an accident,” Kolacki said. “And there are fields that are populated, where they need to put that information in, then it goes to our claims department. There’s information in there where our claims will go into it, there’s information for HR, so everything is in one place and you’re not scanning documents in and creating more emails and work for everyone.”
With the data in one place, managers can implement machine learning, which takes huge data sets, identifies patterns, compares those patterns to actual results, learns from those results, and then uses that knowledge to predict future outcomes. And as new data is feed into the system, it continues to tweak results and refine predictions.
In the context of safety, and predicting who is most at-risk of an accident, machine learning considers leading indicators like hard braking, speeding, accident claims, injuries, work comp, medical information, years of service, pay and more.
“Say I have an accident,” Cardiff said. “Let’s look at all of the data preceding that accident, and that is a profile of a driver. So then I look at hundreds of thousands of drivers across multiple decades who’ve driven billions of miles, and now I feed it into a model to say, ‘Alright, go give it arbitrary weights, spit out a prediction and see where we get.’”
By doing this, we can understand patterns that previously were indiscernible. Most people would say speeding is riskier than taking a sick day, but comparing speeding to hard braking is trickier, and when more conditions are thrown in, like speeding by 25 mph or more compared to hard braking, it’s nearly impossible to determine severity. In that case, Cardiff maintained, most managers would say, “Well, it depends,” or “I need more information.”
Machine learning systems have all the information—and the inhuman ability to quickly and efficiently analyze and interpret it. And predictive analytics isn’t used only for preventing accidents. It provides the same insights into improving preventative maintenance, route optimization and overall operational processes.
“We all have data,” Cardiff said. “We can’t shield ourselves from it. Now we have to use it in some sort of positive fashion.”
Scrap the scorecard
Cardiff says using artificial intelligence is better than traditional methods, like a company scorecard, to weight risks and predict outcomes. While scorecards are better than flipping a coin, they’re only marginally so. “Scorecarding is helpful,” Cardiff insisted. “It’s going in the right direction. It’s at least putting us in the right frame of mind. But it’s not predictive. It’s not leading. It’s definitely lagging as far as an indicator.”
With machine learning, statistical insights into trends are better able to predict which drivers will get into accidents by up to 75%, Cardiff said. And by getting out in front of problems, fleets can realize very real ROI savings. But that’s not the approach’s only benefit. Along with helping streamline operations, and save time, predictive analytics systems can improve culture by helping a company reward and retain its best employees.
With driver turnover, they’re either fired for poor performance or choose to leave for better pay or working conditions. If a fleet can predict an accident and, through training, prevent it from happening, they won’t have to fire that driver, which Cardiff calls “retention by prevention.” And with deeper insights, companies can create and administer more effective rewards programs that ensure drivers never want to leave.
“All of this goes into creating a culture where your drivers feel like they’re being watched after and not over, and that is really important,” Cardiff said. “Because when they feel like you’re there for them, and not there to find fault in them, that changes everything.”
Cardiff said his company often hears from fleet managers that awards programs are a good idea, but they’re cumbersome to maintain, require more time to pull together and manage than anyone really has, and even with all that effort, it’s difficult to ensure the awards go to the right people. With AI, the burden is removed.
“It’s key to let those guys know (they’re appreciated),” Kolacki said. “We have a lot of award programs that we’re putting together for million miles, how many safe miles they’ve driven without a preventable accident. We have terminal awards for not having preventable accidents for a certain time period. And it used to be a spreadsheet where we were putting in the mileage for each location, and monitoring it, and this gives us the ease of having all that data in there.”
