Analytics in the detection of financial fraud has been vital in combating fraudulent activities on credit cards. The rise of the internet has given scammers and cloners a better opportunity than ever to spend your money after cloning your card or stealing your identity.
Through analytics credit card companies have been able to see buying patterns that are out of sync with a customers buying habits, which in theory makes it far easier to automatically detect fraud and stop the transaction before any money is spent.
It is done through modelling what is normally bought and where these items are purchased, then building models that allow a system to notice when these patterns are deviated from. For instance, if you were a keen cyclist and you spent a considerable amount on cycling equipment, the system would learn that these purchases would be out of sync with what you buy.
If you were then to use the same card to buy parts for a motorbike, this would then be flagged in the system and the transaction would potentially be blocked.
The basis of this is positive, it stops scammers from being able to buy anything that is out of line with the kind of things that you would normally be purchasing. Given that when a card is cloned, the purchase history is not, it would be very unlikely that the same things would be bought in similar places by any criminals.
However, this can cause what are known as ‘false positives’ on your card.
It has happened to me, when I know that my card has more than enough available credit to complete a purchase, but it is declined. You later find a fraud detection email in your inbox, which shows that the declined transaction was due to fraud prevention. However, this does not help with the embarrassment that comes when your card is declined in a shop.
This could happen for any number of reasons.
One of the most common is use of the card abroad. This is because it is outside of a regular buying pattern of where items are bought from. This can block the card and make it difficult or expensive to unblock it.
Also if something is bought that does not fit with a long running patterns created, this can cause a false positive. For instance if a card has a history of being used at grocery stores then is suddenly used to buy a car, this would be flagged as being out of pattern and the transaction may be cancelled.
If a present is being bought for somebody who has a different set of interests, this could cause the fraud prevention software to notice and cancel the transaction too.
Companies like FICO are looking at various ways to get around this issue at the moment. What is being found is that as more data is gathered from cards and card usage, models are becoming more accurate. This creates situations where fraud detection is more effective, but the possibility of false positives is increased alongside it.
The solution to both preventing fraud but avoiding false positives is not easy, it will take time to find a suitable model, but until these solutions are found, we may need to get used to the embarrassing moment when ‘declined’ appears on the card machine at the kiosk.