Feedzai Using Data Science And Machine Learning To Minimize Payment Fraud

Feedzai Using Data Science and Machine Learning To Minimize Payment Fraud

/* Story Top Left 2010 300×250, created 7/15/10 */
google_ad_slot = “8340327155”;

The US fight against payment fraud got a new player last week as Feedzai, a data science company based in Europe, launched in the US with a system that uses real-time, machine-based learning to prevent fraud.

Founded by aerospace engineers and data scientists, the company says it is the only fraud prevention solution available for omnichannel commerce. Fraud prevention has to work across all channels, because that’s the way consumers buy, said Nuno Sebastiao, the company’s CEO.

“The consumer /4/buy online, then at a physical store and then through a mobile. We need to collect as much as possible from all these channels. We look at the network of payments as a whole.”

Rather than working from rules and patterns, Feedzai uses machine learning, he added. So as a new channel comes online, Feedzai starts to ingest data from it and begins to make inferences.

Feedzai technology fuses machine learning and lets  analysts predict and prevent electronic payment loss in real time based on behavioral analysis. The software uses massive data, including mobile and social data streams, to create deep learning profiles for each customer, merchant, location or POS device, with up to a three-year history of data behind it.

“Banks can use it to look at transactions from a consumer perspective. A user hasn’t done mobile transactions and all of a sudden mobile transactions appear on his credit card. The bank wants to understand the way users consume on different channels. If you buy on mobile device, is the purchase near where you live?”

In Europe, where many countries have centralized payment systems, Feedzai is the fraud prevention engine in two of them with a combined population of 170 million, said Sebastiao. “We are engaged with issuing banks, payment networks that receive transactions when you swipe your credit card or use an online system, the credit card brands and the retailers. We are able to look at payment transactions from the issuing side, the retailer, the consumer and the brand. In the US our strategy — because we need a lot of data for our systems to shine — we’re talking with banks and merchants who handle large payment volume because that’s where the much of the data exists. In order to safeguard their brand and customers, which is  important due to their large market share, these types of clients want to prevent fraud from data theft, before damage occurs. Payment service providers and payment networks are also highly interested in Feedzai, and we’re handling data for some of them today. ”

He said the system is able to infer behavior to reduce the number of false alerts. If MasterCard sees you have bought an expensive airline ticket in Madison and your next purchase is at Heathrow, it will realize you have bought an international plane ticket and it won’t block your purchases in England.

“That contrasts with today when the first thing that happens when you travel internationally is that you get a call from your bank or your card is blocked. They have the information to know better, but they don’t have the technology to infer that. We see that transaction is not fraud, that is normal behavior. which is why issuing banks come to us.”

In an era when most transactions were made in person with a credit or debit card, older systems of fraud prevention worked adequately, especially when combined with the chip and PIN cards prevalent in Europe.

“Historically physical user card fraud has been managed, but today breaches are skyrocketing, and the existing systems that are in place were not built for these new channels. Today if you find a new pattern of fraudulent behavior it might take six months to deploy measures to prevent it with existing systems. You need systems which are continuously using machine-based learning to look for these signals.”

The threat seems modest now, but it has the potential to explode.

“Online sales are still hovering around six percent of all retail sales, and recent events have shown that fraud is not just an online problem. Our software has the ability to analyze data from any platform to detect breaches by as much as 10 days earlier than other solutions and expose more fraud cases, all with fewer false alarms.”
Douglas King, payments risk expert at the Atlanta Fed, noted in a blog post that Discover’s fraud rate on sales volume increased from 4.8 basis points in 2010 to 7.2 basis points in 2011, and reached 8.8 basis points in 2012. Discover reported $93 million in fraud losses for 2012, or roughly $8 million more than it spent on postage, he wrote.

Sebastiao admitted that for Discover, $93 million is still manageable, but that is up from $40 million in 2010.

“You look at the trend and look at the types of payments — online and mobile are where 70 percent of fraud is happening  — and their use is skyrocketing. If the losses grow to $200 million or $300 million, it becomes a real issue. For cards and stores, it is also a reputational issue. In recent breaches, the reputational damage was far worse than the actual physical losses.”

Feedzai started in Europe, and is also available in South America, Europe, and Africa. With its foray into the U.S. market, Feedzai will begin selling to payment networks, banks and retailers to prevent fraud in omnichannel commerce. The company is backed by SAP Ventures, big data investment firm Data Collective, and other international investors for a total of $4.3 million.

Source: Forbes

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *