When looking at preventable accidents, time and time again we see a summation of events boiling up into a critical incident. A breadcrumb trail of risky behavior is left behind, quantified by the many data sources that fleets have and pull from.
Telematics, on-board cameras, MVR’s, medical history, years of service, observations, citations, and violations, the list seems never ending. Over recent years we as an industry have transitioned from not enough data to being inundated and overwhelmed with the massive amount of data we do have!
When taking a holistic look at the technology available today, it’s not unreasonable to think that we can collectively create a world with fewer accidents caused by operator error. We simply need to be able to do two critical things:
- Predict at-risk drivers with higher accuracy as behaviors start to change
- Prevent accidents by prescribing corrective action based on the data
Predicting
“The most important aspect of all is having each of your systems connect together and integrate into one central place or Safety Suite.”
To accurately predict at-risk drivers, you need to be able to track and connect driver behaviors both inside and outside of the cab. These behaviors are manifest through important indicators that span all of the different systems that fleets use. Because there is so much data available, it’s critical to filter out any noise and focus on the strong leading indicators both as individual metrics, but most importantly as correlated connections of behaviors and attributes.
Industry knowledge and experience is a good way to see these trends and can provide a good base to start from, however, that can only get you so far. Technologies such as Machine Learning (ML) can sort through all of the different data sources, weigh their importance based on correlations to actual outcomes, find connections between leading indicators, and provide incredibly accurate outputs of which drivers are most risky.
If you’d like to go more in depth on this my colleague, Jonathan Bikowski, put out a great post on ML applications in trucking. But in short, ML takes past outcomes like accidents or violations and digests millions of data points to learn the patterns of behavior mixed with driver demographics that lead to that outcome, so you can then prevent future incidents.
The beauty of technology like this is that it augments and supports our current efforts to understand driver behavior and provides us the reach to fully analyze all the information we have at our disposal. I’ve talked with so many fleets that pull data from Telematics or Onboard Cameras in hopes of consistently sorting through and filtering risky driver behavior, only to find there is so much data coming through that they have to file it away in an email folder and hope to get back to it in a few weeks. Not to mention, the struggle to look at Telematics events and then cross check that with HR incidents, violations, medical history, and claims data for that and every driver that gets an alert is near impossible.
Another important aspect of prediction is getting consistent, clean, and connected data. Consistent meaning that your data is always coming into the same place, logged in the same way, and displayed in a digestible and actionable manner. The data also needs to be clean, or free from errors or inconsistencies (i.e. “backup”, “Backup”, “back up”, “Bakup”, and “backing” all being potential inputs when logging a Backup Accident).
The most important aspect of all is having each of your systems connect together and integrate into one central place or Safety Suite. This helps keep things consistent, clean, and linked so you don’t have to go chasing data to get a clear understanding of the whole picture surrounding driver behavior.
Imagine a mirror made up of many smaller pieces, where each data source (i.e. Telematics, HR, Training, etc.) equates to one piece of the mirror. If all you did was look at one piece at a time, you would have a limited perspective of the image you’re looking at. Not only can you not see the whole picture, but you might create false assumptions based on your lack of vision. But with all the pieces connected and together, you clearly see what’s in front of you.
All of this should get you to a place where you can analyze the complete set of data and produce outputs of drivers who are at-risk, which is the ultimate goal! You can also compare this to historical data and start to see accuracy and improvements within your fleet.
Preventing
“Prediction and prevention will help improve driver retention and lower turnover”
The first thing you must do before you can prevent impending accidents or provide help and benefit to your fleet of drivers is TRUST THE DATA. All the technology and predictive power in the world is ineffective if you are not willing to make decisions based on that data.
I’ve spoken with Safety Managers who’ve received alerts and notifications of at-risk drivers and have done nothing because it was a well tenured, ‘very safe’ driver that ‘is just going through an off week’. But what they didn’t know was that this driver was going through a divorce which caused his behavior and focus to deteriorate into an accident.
This Safety Manager was right in saying that the driver is normally a great driver, but the data was there to show his behavior had changes. Predicting driver behavior isn’t nor should it be used solely to find and fire bad drivers. It should be used to identify which drivers need the most help and get them assistance or training before something critical does happen. When used correctly, prediction and prevention will help improve driver retention and lower turnover.
The next step would be to assign actions based on the driving factors or leading indicators. This could be trainings, ride-alongs, verbal or written letters, or simply just reaching out and speaking with them one-on-one. If you ever think a driver is on the fringe or has barely crossed your threshold for action, DO SOMETHING! Every time you interact with your drivers, you have the opportunity to catch a building problem or at the very least show that you’re looking out for them and keeping an eye on what they do.
Incorporating your 3rd party Learning Management System or your in-house training content is helpful and makes administering and tracking the actions more seamless. Being able to see an at-risk driver, assign the corrective action, have that push to the driver to take the training, then log and track the results in a central management system is definitely the goal.
The last key piece is to have a process and stick to it. Creating and presenting an escalation process is critical, but consistently administering it is imperative to getting buy-in from drivers and tracking the impact of the decisions you make. Another benefit of a Safety Suite that has all of your driver data in one place is the ability to see all the culminating factors of this driver’s risk so you can always administer the right step of your corrective action plan.
Final Thoughts
Throughout the industry, everyone’s goal is to have an accident free fleet and strong safety culture. The challenging part is providing these things with limited resources and difficult hurdles like recruiting and driver retention. With the technology that is already available in today’s market, we can make serious strides to achieving these goals.
Financial benefits and ROI are huge factors when evaluating the feasibility of implementing these technologies and processes. When done correctly and with the right vendors, you will see tremendous reductions in accident and claims costs, improved driver retention, and even insurance premium reductions.
One word of caution: Don’t get fooled by a simple and static dashboard disguised as a predictive Safety Suite. If you are looking to vet solutions and vendors, make sure there is a data management / safety operations platform at the base, full 3rd party systems integration supporting it, and ML or advanced predictive analytics driving it. Without these things you’ll never get the entire picture of your driver’s behavior and your fleet’s risk.