Identify Anomalous Product Postings

Business Problem

Online marketplaces connect the world. They allow consumers to purchase products from sellers across the world. However, at its immense scale of operations, online marketplaces have also become a breeding ground for criminal activity where fraudsters illegally deceive innocent users into schemes such as purchasing fake goods or giving up their credit card information. Criminal activity within the marketplace trickles into identity theft and money laundering outside of it. According to a study by criminologists at Surrey University, losses from fraud in online marketplaces alone adds up to $860 billion a year.

The difficulty today in preventing fraud is that most solutions rely on rules based systems that fail to adapt to the ever changing behavior of criminals. Rules that simply prevent obvious cases of fraud are quick to become outdated in the face of criminals who will go the extra mile to cheat the system. As a result, criminals are able to get away with their repeated crimes over and over again, often without ever being caught.

Intelligent Solution

AI helps online marketplaces identify and prevent both known and unknown behaviors of fraud. While supervised machine learning will predict whether new accounts or product postings are fraudulent based on patterns learned from historical cases of fraud, unsupervised machine learning will flag anomalous accounts or product postings that often also include unknown behaviors of fraud. By parsing out the noise in historical data to identify anomalies across accounts and product features, AI allows a marketplace to continuously adapt to new forms of fraud that threaten the integrity of their offerings. With this information, a marketplace is able to automatically flag accounts or product postings that have a high risk of fraud, ultimately preventing fraudulent sellers from conducting financial transactions until a subsequent manual investigation is completed. For human investigators, AI informs them of the top reasons why an account or product was flagged as potentially fraudulent, which helps investigators triage the extensive list of cases by prioritizing those with the highest risk of crime.