• Monica Eaton, Founder at Chargebacks911


• 11.05.2023 11:00 am
#ai #fraud

This year, a staggering 60% of all chargeback claims are estimated to be fraudulent, according to our research, and up to 75% according to Visa. As a result, merchants and the anti-fraud industry must prepare to identify what this activity looks like in real-time by collecting and analysing relevant data. While a great deal of industries benefit from the automated collation and analysis of huge amounts of information, known as ‘big data,’ this route presents a challenge for chargeback management. Merchants receive limited information about chargeback claims based on response codes from payment cards. The unfortunate result is that a great deal of manual work becomes necessary to reconcile this data. 

While technological innovations have served to reduce the frequency of chargebacks, merchants still lack sufficient data on chargeback attempts and the originating dispute enquiries. In this article, we’ll explore the emerging avenues available to improve the quantity and quality of the data received by merchants on chargebacks and how this can be used to improve chargeback management. 

Big data explained 

Businesses and organisations recognise the value of big data, but finding ways to process and analyse it can be a challenge due to its complexity and size. As a result, AI, specifically generative AI, is set to be the major technology story of 2023, with many people already seeing tools made possible by the technology being introduced in their workplaces. ChatGPT, a large language model developed by OpenAI, is a notable example of technology that has successfully leveraged big data to drive innovation. 

Using more than 40 gigabytes of data resources, including 175 billion syntax parameters, the tool has been trained to understand and generate responses in a language that feels natural to a human reader. Its applications range from customer service chatbots to language translation services and virtual assistants. While similar technology is already being implemented in certain areas, its widespread adoption seems inevitable as more sectors discover additional applications for the technology. We have recently seen similar technology to ChatGPT used by Bing to replace traditional web searches with mixed results, but, like self-driving cars, it is simply a matter of time before this technology becomes more widespread.

AI: the new frontier of fraud prevention  

Chargeback fraud occurs when a customer disputes a valid charge made on their credit card, insisting they did not make the purchase (oftentimes innocently unrecognised, sometimes referred to as first-party misuse) or voicing another issue related to the product or service.  If the dispute is upheld, the merchant must refund the money, as well as pay any associated costs, and is charged an additional fee by their payment processor. Regardless of whether or not the merchant proves their innocence, and even if their customer admits making a mistake, a negative chargeback statistic will forever stand as an irreversible black mark, and at best, the merchant may qualify for a reversal of the transaction amount that was originally debited along with the chargeback fee.

Chargeback fraud poses a significant threat to businesses, with the National Retail Federation in the U.S. estimating that retailers lose approximately $50 billion each year to fraud, a significant portion of which is attributable to chargeback fraud. 

The rise in online shopping has exacerbated this problem globally, creating pressures in streamlining the dispute process that have trickled down to expose gaps that can be easily exploited. 

While machine-learning technology promises to be a powerful solution to this growing issue, it is critical to understand its limitations. Platforms like ChatGPT and Large Language Models are not the kind of artificial intelligence popular in science fiction movies and are unable to "think" independently. The technology is able to draw upon a huge number of sources and pull enormous amounts of information together, but naturally it isn’t able to come to any conclusions on its own. 

While machine-learning tools can produce perfect text by copying existing text rather than “thinking” about the substance of the question, they are prone to producing errors. While a copywriter can work around these mistakes, AI solutions are not acceptable when it comes to fields like fraud prevention. Within the binary world of whether an action was fraudulent or not, machine-made platitudes will not cut the mustard and unfounded accusations will only serve to damage a merchant’s relationship with their customer.  

Consequently, businesses need technology solutions that are built to support merchants in their quest to tackle the issue of chargeback fraud rather than relying on general-purpose AI. Such solutions are already on the market and their models are based on tailored datasets aligned with the growing demands in the dispute and chargeback arena.

A multitude of advantages can come to businesses who leverage these solutions—from the ability to analyse vast amounts of data and improve accuracy over time, to supporting their needs for real-time monitoring, providing feedback to assist their fraud engines in reducing false positives, and help streamline complicated workflows that increase efficiency through task automation.

When choosing the right supplier, merchants should consider the vendor’s time in business, connectivity options, and overall scope.  Chargebacks are a problem in every country, and span across every card type and method. Defining scale and scope is paramount to help optimise benefits and results for the immediate and distant future.