Through Machine Learning Risk management is becoming increasingly complex and demanding, which requires an intelligent strategy to deal with the numerous transformations that fraud types present. Offering a frictionless shopping experience is no longer a differentiator and has become a necessity to ensure security and a good shopping experience.
There are several ways to authenticate a purchase
from through Machine Learning models that approve more than 99% of transactions automatically to technological components that help solve the most difficult cases.
However, isolated components are not effective phone number database in defending against unknown schemes, adapting to new fraud patterns, or dealing with increasingly sophisticated fraudster techniques. When misused, they become poor allies in risk management, approving fraudulent transactions or generating enormous friction with the good consumer.
For example facial biometrics
Eonsidered infallible by many, when differences between a business name and a brand used alone, cannot block all fraud attempts. When poorly applied, it negatively impacts the user experience and incurs significant costs for the company. However, biometrics perform optimally when integrated into a multi-layered prevention system.
It is important that each purchase is carefully analyzed, as well as the use of components in the validation of each one. For this reason, ClearSale created the concept of Safety Authentication Score , which measures the intention of the person who makes the order, based on the data entered by the buyer and the purchase history.
The importance of context in every transaction
When purchasing online, there are two important parties involved: the person making the purchase and the credit card holder. These are not always the same people.
For example, suppose the following scenario: Antonio mobile list through Machine Learning registered with a store using his own data, except for the credit card, which he uses as José’s card. The e-commerce site does not use solutions that verify whether the credit card actually belongs to the person making the purchase, i.e., there is no certainty, at the time of purchase, whether Antonio is using a credit card that is not his.
To illustrate think about classifying the behavior
the person making the purchase in our example, as well-intentioned or ill-intentioned.
If Antonio is well-intentioned when making a purchase, even if he uses José’s card, he can validate the transaction without having to bother the cardholder: José is in agreement with the transaction and Antonio’s good intentions can confirm this. In this scenario, any additional validation made using Antonio’s data (whether it be a phone call, biometrics, text message, etc.) will legitimately confirm the purchase.
However, let’s suppose that Antonio is acting maliciously and that José is not aware of the purchase on his card. Systemically, looking at the transaction, nothing has changed. It’s the same data, same time, same product, etc. However, if validation were performed using the data provided in the order, the transaction would be approved by Antonio and José would have his card debited for something he did not purchase. The transaction, initially approved, would be disputed and the store would be responsible for the chargeback .
It is known that fraudsters are creative
That fraud occurs in many formats. As an example, we can report some situations already faced by ClearSale: a physical store, with a malicious salesperson, offered a certain product to a buyer at a lower price than the store advertised, justifying that, as a direct seller, he had this possibility.
The buyer, attracted by the new proposal, made the payment for the product directly to the seller, into his bank account, via PIX (there are also cases in which the seller generates a payment link).
The fraud occurred when the seller
malicious through Machine Learning from the beginning. Used the data of the good buyer linked to a fraudulent credit card. In an online purchase of the same product sold. On the website of the store he worked for. Opting for the express purchase option with in-store pickup.
The buyer, upon receiving the link to capture. Facial biometrics, collected their biometrics and confirmed the purchase. After all, they were actually in the store buying that product.
The sale was approved. The product was delivered to the buyer. Who left satisfied and unaware of the fraud. And the seller continued to deposit the amount received from the buyer directly into his account, as if the situation had been honest. The store, however. Suffered a loss, as the purchase made on its website was disputed.
In another similar situation. The malicious seller could register as a seller on a marketplace. Offer a certain product at a lower price than that practiced. On the market and, upon attracting the good consumer, receive payment for the supposed sale in his own account.
In order for the buyer to not file a complaint
The seller uses a fraudulent card to buy the same. Product on any e-commerce site and provides. The buyer’s address as the destination.
Similarly, if any validation were performed with the buyer. The chances of validation would be high and the seller would complete. The fraud without much difficulty.
These examples reinforce how it is necessary to look beyond the box. And it is exactly in these types of situations that the Safety. Authentication Score can act. As it indicates the probability that the person making. The purchase has malicious intentions.
