By Kurt Thearling, Vice President, Analytics, WEX
Analyzing data is built into our brains. The cavemen analyzed data to understand animal migration and hunting patterns to help them survive. Sailors analyzed data to make charting improvements on their ocean voyages. Astronauts use data to connect to places mankind had only dreamed of going.
In a sense, we’re still exploring and learning new things about how data can be collected and used. We often bring together new data, creating endless combinations and layers to answer a plethora of questions. By combining data in new ways, we may even answer questions that we didn’t know we had.
But, let’s start at the beginning. For a fuel payments company like WEX, that means right at the pump. What’s usually available is information about gallons purchased, price per gallon (PPG), fuel grade, and the location of the gas station. With the data, we can start considering some interesting questions.
Saving money by changing driver behavior
For most fleet managers, efficiency is key. They want to understand how to lower costs, often by identifying and reducing waste. A quick money-saving example would be to see what, where and how drivers are purchasing expensive fuel. A fleet manager can identify drivers who purchase premium when regular unleaded would work just as well. By comparing prices to nearby stations and the amount the driver spends per fill-up, the fleet manager will understand the full savings potential and can reach out directly to the largest offenders and correct their behavior. We’ve found that messaging drivers upon identifying a premium fuel purchase pattern can result in immediate 80% reduction in the number of drivers buying premium. That simple change in behavior translates into significant cost savings for fleets.
But, occasionally the driver makes the argument that they didn’t purchase premium. Sometimes the data is wrong, so when you want to hold a driver accountable you need to be sure of the accuracy of your data. In our experience, up to 6% of fuel transactions are coded incorrectly by the merchant. What do you do about that? That’s where additional layers of data can help. Layering historical transaction information can help determine if the reported fuel grade is correct. Does the purchase price look correct, relative to nearby purchases? There’s usually a significant difference in the PPG for premium and regular as well as diesel and gasoline. The purchase history for that vehicle can also help. If the driver purchased only diesel for a particular vehicle in the past, it’s very likely that the questionable purchase was for diesel. Putting all of this together allows us to correct nearly all mis-coded transactions, giving a fleet manager confidence when addressing negative fuel purchase behavior.
Reducing fraud and theft
Data can also assist fleet managers from losing money through fraud and theft. There are a number of ways that data can be used to pinpoint where and when fuel theft is happening.
One of the most definitive ways to identify theft is by matching the vehicles’ GPS location to the location of the gas station where a transaction takes place. If the vehicle is not actually near the pump at the time of fueling, it’s a pretty good sign that something is amiss.
But, even if you don’t have the detailed location information, there are ways to flag suspicious transactions. By bringing in vehicle information (make, model, year), often by decoding the VIN number, you can tell a lot about the vehicle. This translates directly into what manufacturers say the fuel mileage should be. Coupled with the size of the fuel tank, you can estimate how many miles a vehicle can travel on a tank of gas. If the vehicle is refueled more frequently than would be required by the distance traveled, that gasoline might be going somewhere else — maybe into the driver’s personal vehicle.
[AI and machine learning for fleets
With some fuel cards, drivers are required to enter their odometer reading every time they pay at the pump. This is an easy way to track overcharges when the mileage doesn’t match the fuel purchased. Occasionally, drivers enter incorrect odometer readings. There are techniques that use artificial intelligence (AI) to spot incorrect odometer readings and correct them to match the true mileage. By using GPS and satellite imagery, AI could be used to determine if the supposed fuel location is a gas station. AI and machine learning could look at the fuel purchase data and call out vehicles that are less efficient relative to others in the fleet. They could also use the data to look for situations where certain vehicles in the fleet would be better suited for use than others, saving the fleet even more money.
Odometer cleansing product sensing
At WEX, we have over 30 years of experience turning data into real-world solutions. Artificial Intelligence is just one of the many ways data can be captured and translated into meaningful information that can help fleets make more informed decisions – or correct negative behavior. Working in tandem with experienced professionals, there are so many possibilities for fleet solutions. From better data reporting to better tracking to implementations that haven’t even been conceived of yet, it’s a truly exciting time to be in the business.
After all, every transaction has a story. It’s our job to tell it.
