© Fraugster

02 September 2019

Clever AI helps online shops fight fraudsters.

For example an Israeli who books additional baggage with a German credit card. And this for a flight the next day. For e-commerce merchants, all the alarm bells are ringing. The transaction is blocked. Understandably, for the card's 3-D Secure procedure has discovered two suspicious indicators at once with the different nationalities and the short-term nature of the purchase. But as understandable as these happenings are, they prove to be flawed in practice. Chen Zamir knows this from his own experience, as he went through it himself. “I had to use my mother’s Israeli credit card,“ he smiles afterwards and adds: “Sometimes, there are 10 fraudulent indicators and one legitimate indicator with online purchases, but the transaction can still be legitimate. The indicators alone are not relevant. The story around the indicators makes the difference.”

Looking at Zamir's case, an Israeli passport - combined with a German credit card - is suspicious at first glance. On the other hand, the whole thing makes sense: an Israeli who books additional luggage from his home country for a flight to Germany with a German credit card - sounds credible, doesn't it? However, the traditional 3-D Secure procedure does not reach this conclusion. As a precaution, it blocked the transaction.

Chen Zamir, CTO and co-founder of Fraugster © Fraugster

Fraud is expensive – fear is more expensive

This story is not a special case. On the contrary: This story is not a special case. On the contrary: e-commerce merchants lose 1.5 per cent of their sales per year equal to 52 billion US dollars in 2019 due to fraudulent actions. But even if criminal activities are increasing twice as much as the entire e-commerce market itself, this is not the biggest problem for online merchants: even more expensive than fraud itself is fear of it. “Most companies, when they encounter fraud issues, these count as large attacks. A few months of revenues are down the drain,” says Chen Zamir, who spent five years with PayPal in risk management and previously served in Israeli Army intelligence. “They want to focus on it not happening again – they don’t care how much it costs, they don’t care if it is accurate. It is just frightening. The easiest thing they can do is to get a very secure, but maybe outdated system. I also think and we have to keep in mind that lots of technology advances so fast, but new technology is hard to keep up with.” It is hardly surprising to him that they often resort to outdated anti-fraud systems. After all, it would be difficult even for experts to keep up with the latest technology - and even more so with criminal elements. But this is exactly what conventional anti-fraud technologies are trying to do: to detect fraudsters they work with rule-based algorithms that do not sufficiently explain unusual but legitimate behaviour such as that of Zamir. Additionally, these tehnologies often create friction in the checkout process and lead to cart abandonment. These drop-offs and so-called "false positive declines", i.e. the false blocking of real customers, can cost merchants between 5:1 to 30:1 for every fraudulent dollar.

AI writes a story in milliseconds

What conventional anti-fraud systems fail to do is what the Berlin start-up “Fraugster“ has specialized in. “We did not approach it first with an algorithm and then a business case, but the opposite“, explains Chen Zamir, who founded the company in 2014 together with his German partner Max Laemmle. Instead of collecting huge amounts of data for months to train the machine and develop a model that has to be reworked every year, they developed a self-learning artificial intelligence. “We don‘t train models, we don’t have a model. We wanted to build an algorithm that looks at fresh data in real time. On the one hand massive and on the other hand really fast and agile,“ explains the Israeli with an adopted home in Berlin: “Fraud is a fast development, it changes quickly – it doesn‘t take six months but one night.“ Fraugster technology is also fast. Artificial intelligence detects fraudulent transactions in no less than 15 milliseconds. Max Laemmle sums up how this works in Forbes DACH: "Our engine collects data points - and then tells a story."

Individual gap filling

This is exactly what makes the technology so special: Fraugster not only looks at the indicators, but also enriches the transactions with data in order to close the gaps. Case by case. "Normally, a transaction has between 20 and 60 data points," explains Laemmle, who previously worked for the Sum Up and Better Payment payment services, "including name, home and billing address, IP address, and so on. But we want to know what the customer's behaviour is at the exact time of the transaction. That's why we add around 2,500 data points to each transaction." With this unique innovation, the 60-strong international team around the two founders has succeeded in topping the industry average not only in terms of speed, but also in terms of accuracy, where no one can currently keep up with the Berliners: merchants using Fraugster technology only lose two US dollars on every fraud dollar. If you ask Chen Zamir, there is however still room for improvement: “On the technical side, we want more speed, to identify more complex data and also work on accuracy. 2:1 is not 1:0. One thing that is nice about fraud is that the demand is always there. This is not a problem that you can solve or reduce to zero …. It is more about tackling bigger, more complex problems to increase accuracy.”

Safety net plus: full liability against fraud

However, Fraugster's service has long ceased to be limited to pure anti-fraud technology. While the product packages ”Flame“ and ”Fire“ serve the classic field of risk management software and offer not only SMEs but also larger companies an additional safety net for their in-house risk management teams, ”FraudFree“, launched in summer 2017, has a special additional benefit. Through its partner Munich Re, Fraugster assumes full liability  for every approve transaction and provides chargeback protection if the technology identified a fraudulent transaction as legitimate.  “Currently we are in the pilot phase for an improved version of FraudFree“, adds Chen Zamir, “our latest offering not only gives you chargeback protection but the technology also guarantees uplift in approval rates for data-driven e-commerce services. This is what we are working on at the moment. It is more expensive, but will guarantee an increase in revenues for our clients.“

E-commerce merchants will welcome this expansion of the core business to other data-driven services. According to Fraugster, the company already handles around $ 35 billion in payments per year. This brings the start-up company money: a fee of between 0.01 and 0.3 per cent is charged per transaction. "The fee always depends on the risk category of the retailer," Max Laemmle told Forbes Dach, "selling T-shirts is less risky because you need someone who wears the same size and has the same taste - and who is willing to pay a certain price. Consumer electronics products (such as iPhones) are therefore more risky than fashion. But also in fashion there is a "one-fits-all" category, such as handbags." So far Fraugster mainly has customers in Europe and cooperates with payment service providers such as the French companies Ingenico and Natixis and the Swiss company Six Payments. Chen Zamir sees international growth potential: “We are looking to both the US and Asia and want to lock down more regional partnerships there.“

Rosy future in Berlin

However, this rosy present and even rosier future at Fraugster is not noticeable in black figures. Not yet. But more important than profitability, according to co-founder Zamir, are "growth and how you operate a business," he says diplomatically. Investors seem to be satisfied with both: right from the start, Fraugster did not have to worry about its finances. In addition to venture capital provider Earlybird, the Austrian funds Speedinvest, Seedcamp and Rancilio Cube, HSB Ventures, a subsidiary of Munich Re, participated in the start-up. Chen Zamir can only guess why: in addition to the unique technology and early partnerships with payment service providers, he believes that the start-up scored particularly well with his team. “We were convinced from the beginning that there is a lot of raw talent in Berlin. Not so many experienced talents, but a lot of graduates from relevant universities“, the decision in favour of the location proved to be the right one, “there is also a great appeal for immigrants. I am one of these, and clearly this influx of highly-skilled immigrants is still growing. We were able to get great people early on, not least because our technology and our niche were and still are sexy." The employees are not alone with this opinion: Fraugster did not make the headlines until the end of 2018 with its latest financing round. Among others, CommerzVenture, an investment company of the German Commerzbank, has invested more than 14 million US dollars in the Berlin start-up. Fraugster will undoubtedly be heard for even longer with its stories.