Predict & Prevent Your Next Big Collision

predict and prevent

Predict & Prevent Your Next Big Collision Using This Surprisingly Simple Technique

 

It’s easy to get overwhelmed by the many ways things can go wrong in your fleet. Short of a crystal ball, the next best tool to manage risk is a surprisingly simple technique called frequency analysis. In a nutshell, frequency analysis tells us that if something risky happens a lot in our fleet, we should pay attention to it.

This article will show, step-by-step, how you can use this fast, simple technique to help predict and prevent collisions in your fleet. Best of all, you don’t need to be a data jockey to do this analysis and get results.

Let’s get started!

 

Step 1: The Who What When Where of Collision Risk

This is just as easy as it sounds. Using a safety telematics system such as the program offered by Lytx, you can generate simple lists based on volume, or frequency.

  • The “who” is a list of drivers in your fleet with the highest amount of risk points generated by how much they engage in risky driving behaviors.
  • The “what” is the types of risky behaviors most prevalent in your fleet.
  • The “when” is the time of day or year when the bulk of your risk happens.
  • The “where” is the list of hot spots, or where your risk tends to cluster.

prevent collisions

 

Step 2: Visualize Your Risk

By turning these lists into charts or maps, you can get a better sense of scale — and identify clues about where to prioritize your efforts. In the hypothetical charts below, for example, you can easily see that Driver A is more than five times riskier than Driver F. If you focus your efforts on coaching the top five drivers in your list, or even the top three, you can dramatically reduce your risk.

frequency analysis

 

Step 3: Apply the 80/20 Rule

The 80/20 rule simply states that 80 percent of your risk come from 20 percent of the source of what you are measuring. So, in the “who” column, 80 percent of your fleet’s risk tend to stem from 20 percent of your drivers. This is more of a rule of thumb than a hard rule, but it’s tremendously helpful when you have limited resources and need to know where to start and where you can get the best return for your efforts.

 

segmenting risk frequency

 

Step 4: Predict Collision Risk

By charting the frequency and severity of collisions (or other risky events) over a period of several years, you can start to pick out patterns. Below, for example, we see spikes every October going back eight consecutive years. This pattern gives you an important clue to start digging for root causes. Once you find out what’s behind these surges, you can plan ahead with safety campaigns, driver outreach, or seasonal re-routing to reduce risk and, ideally, prevent future collisions.

safety campaign planning

 

Step 5: Target Higher Risk Using Odds Ratios

Some behaviors are riskier than others, and you need a way to prioritize the things you know to be more hazardous. Odds ratios are your key. Simply put, odds ratios are the odds of an outcome (say, a collision) given a particular exposure (driver fails to keep an out, for example). Here’s how you fold in odds ratios into your analysis.

Let’s say you’ve found 70 instances of drivers smoking, which can be a distraction to the driver, and 50 counts of the driver failing to keep an out. In a simple frequency analysis, smoking would rank higher on your list of risks, because it occurs more often.

However, as a safety professional, you know from experience that failing to keep an out is more dangerous than smoking. So, you mentally shuffle your list to prioritize failing to keep an out over smoking, even though smoking happens more frequently in your fleet. You’ve intuitively weighted your list.

There’s another, easy way to adjust your list with more precision, if you have access to more comprehensive data. Lytx, for example, has generated a list of behaviors by collision odds based on more than 100 billion miles of commercial driving data from our clients. From this data, we know drowsy driving tops the list — with an odds ratio of roughly 1.50. This means the odds of getting into a collision is about 1.50 times higher if you are drowsy than if you weren’t drowsy.

If you multiply each behavior count by its odds ratio, you get a list of each weighted frequency value. For example, if you looked at 100 instances of drowsy driving and multiplied it by 1.6 your new, weighted frequency value would be 160. 

Summary of Key Takeaways

You’re now ready to use frequency analysis to help predict and prevent collisions and manage risk. Some things to keep in mind as you tackle your first risk analysis:

  • Keep it simple
  • Use the 80/20 rule to focus your efforts
  • Don’t forget about adding seasonality to your analysis to help with prediction
  • Use weights to prioritize riskier behaviors

No Previous Articles

Next
Sensor Fusion: Building the Bigger Picture of Risk
Sensor Fusion: Building the Bigger Picture of Risk

Discover how sensor fusion technology and artificial intelligence can be used to detect risky driving behav...