Like most people in Florida, last week I spent the latter part of the week preparing for hurricane Dorian. Water and gasoline are two major requirements for disaster planning. Thankfully Dorian just passed over Orlando. But in my preparations, something happened that reveals the failure of automated processes to control fraud. Sam’s credit card algorithms failed.
We use Sam’s gasoline. We also use the credit/membership combination card. The combo card is a fully functional credit card in addition to the membership card for the store. On Friday, like most in Orlando, I decided to top off the gasoline tanks in our cars.
Orlando is about 50 miles to the west of the east coast and about 90 miles from the Gulf of Mexico. Hurricanes threaten Orlando from three sides, the east, the west and the south. To escape a hurricane, if needed, one must drive north-west. It is about 300 miles towards the Florida border and about 500 miles to Atlanta. Thus, most cars must fill up at least once before leaving Florida.
In an emergency, gasoline becomes scarce as most Floridians are also heading away from the danger.
As a result, it is always a good idea to top off your car’s gasoline tanks and add spare tanks to deal with the worst-case scenario. Staying in place may not always be a good thing especially with a long-term electrical outage.
On Friday I headed out to our nearest Sam’s to top off the tanks.
I wasn’t surprised to find long lines at the Sam’s gas pumps. Rationalizing that topping off the tank with about three to five gallons it would take, it made much more sense just to drive to the next gas station to top off instead of waiting and wasting the gas I already had.
I wasn’t the only one with the same idea. The individual next to my pump experienced the same thing, our Sam’s cards would not work at the pump. No error or decline codes, the pumps just would not turn on. Since it happened to both of us, we assumed the gas station was having problems with their credit card processing machines.
I drove off to the next station that had smaller lines. When I reached the pump, my Sam’s card was declined. So, I just used another credit card to top off the car.
A few minutes later I received a fraud alert advising me of a potential misuse of our Sam’s credit cards. It asked me to confirm two recent purchases for one dollar each at two different gas stations. I immediately realized what happened.
Sam’s likely uses algorithms to deter fraud on their credit cards. Seeing that we normally only use the Sam’s credit cards at Sam’s, my attempts to use it to fill up the car with gasoline at another gas station triggered a fraud alert.
Since we primarily use Sam’s for its gasoline, we keep the Sam’s credit cards in the car. It is easier for me to keep our budget by using one credit card for gasoline purchases. And, since I am a creature of habit, I just grabbed the “gasoline” card to fill up the gas tank, although I could have used any other card.
But my attempted use of the Sam’s card at another gas station was not fraudulent, although an algorithm determined it was. The predictive behavior of the algorithm did not consider that hurricane Dorian was headed my way.
As a result, it failed.
It wasn’t a serious problem as we have other forms of payment available to use.
But the algorithm’s failure to update its model to deal with a disaster reveals the failure of relying too much on automated predictive behavior models to make business decisions. Normally, predictive algorithms work as required. They are good at reducing overhead costs. But they remain computer code that cannot deal with real-world events.