How were fraud attempts in online retailing be detected in the past - and what progress has been made due to machine learning? ? AI expert Martina Neumayr from IT and analytics service provider Experian reveals the answer in a guest article for the German IT magazine BigData-Insider. We think: It's worth reading.
"For a long time, companies had to rely on classic analytical methods and systems to separate the wheat from the chaff," writes Martina Neumayr in a guest article for the German IT magazine BigData-Insider. The expert works as Senior Vice President Credit Risk & Fraud Services at Experian DACH.
The systems at the time would have examined transaction data in e-commerce in a static way, according to precisely defined parameters. They would then trigger a warning message, for example, if the online order for a particularly expensive product was placed via a smartphone with a German SIM card from abroad. In such cases, a fraud manager often had to check whether it was a valid transaction or a fraud attempt. This not only cost time and money - but also delayed the processes, which in turn led to honest customers abandoning their purchases.
"Such a procedure is no longer appropriate in the face of increasing profitability pressure, massively rising transaction numbers and more demanding customer expectations," the expert postulates in her article - and then turns to the advantages of modern machine learning (ML) systems for transaction control and explains how they work and what they can be used for.
Contrary to what many myths about artificial intelligence (AI) suggest, in reality they are "by no means threatening", but rather "useful and supportive, modern analytical methods". And: "They can help companies increase their online sales and manage their businesses soundly by very concretely helping to better distinguish honest customers from fraudsters."
Curious? Read the full article here (in German):