Jason Tan, Sift Science: "Machine learning digests considerably more data than a human ever could"

Which are the main industries that your company opens its solutions to? Per industry, do you see a trend in fraud solution preference?

Sift Science serves customers across a number of different verticals and business models, including ecommerce and retail, travel, online marketplaces, on-demand businesses (like food delivery and ride services), payment gateways, and more.

Regardless of industry, we are seeing that companies are no longer satisfied with the traditional rules-based fraud detection approach. A few years ago, businesses depended on simple and stagnant if-then rules to cull through every transaction. That’s just not good enough anymore. Companies need a fraud detection tool that works at the speed of modern business – instantaneously. Think about how Amazon is changing the way that we shop. Buyers now expect shipping within days or hours; the old rules aren’t cutting it, and fraud solution providers are investing in smarter, more adaptable technologies like machine learning and data analytics to meet customer needs.

How is machine learning helping revolutionise the fraud ecosystem and provide security to merchants?

The traditional approach to fighting fraud is based around rules. But rules don’t adapt or learn – whereas fraudsters are constantly evolving their approach.

Our customers are seeing a wide variety of fraud abuse – money laundering, fake listings, promotion and referral abuse, as well as stolen payment information and credit cards. Machine learning is revolutionary in its ability to digest considerably more data than a human ever could, and synthesize meaningful predictions from all of that information. This technology can be used to look at any number of different fraud signals to uncover patterns and learn what fraud looks like for each user’s unique business case. As it sees more and more instances of the specific types of fraud a business encounters, it gets even better at spotting it. This gives merchants an extremely powerful tool for proactively detecting fraud, and blocking it before it happens.

What is the degree of damage merchants can suffer if they don’t take into consideration chargebacks prevention?

Losses through online fraud are only going to increase, with experts predicting card-not-present fraud could cost businesses USD 7.2 billion by 2020. We worked with EatStreet – an on-demand food delivery service – to battle credit card fraud and prevent scammers from gaming the site’s rewards program. EatStreet’s chargeback rate decreased by 70% within the first month of their integration with Sift Science’s machine learning-based solution, meaning fewer instances of coupon and order fraud where EatStreet was stuck with the bill. By decreasing their chargeback loss by 85%, EatStreet saved countless dollars and – perhaps more importantly – hours that the team now dedicates to improving the experience for legitimate customers.

How can merchants prevent headaches like fake accounts, spam and malicious behaviour without turning away good customers?

A good, safe payment experience is a constant balance between preventing fraud and not turning away quality customers. For example, while registration or account creation help merchants to verify the identity of a user, research shows 30% of users abandoned their cart when asked to register upfront. So how do merchants protect themselves without alienating potentially good customers? It’s about building a strong experience for the user that doesn’t make them feel like you assume they are a criminal. No one wants to be followed around the store by a security guard because the business assumes that visitor is going to steal something. Too many online experiences feel like that – perceived guilty until proven innocent.

With machine learning, however, businesses can quickly spot good users and reward them while flagging suspicious users. A clear example of a merchant turning their fraud tool into a revenue driver is OpenTable’s use of Sift Science to dynamically adapt the checkout flow of their e-gift card product in response to a user’s potential fraudulence. The power of real-time and large-scale machine learning solutions lies in their ability to always be running in the background, updating the moment new information is received. Preventing bad behaviour on sites can be as simple as plugging into a global network of data so merchants can prevent and proactively identify bad users, rather than simply responding once the damage is done.

From your point of view, which are the key takeaways from the MRC Seville event?

In addition to meeting customers in person and getting a chance to connect with 30% of our customer base outside the US, it was great to hear more about how fighting fraud differs in different parts of the world. The sessions on topics like 3D Secure were particularly valuable – it’s an area that is very important to our European base, but not very prevalent in the US. We couldn’t have gotten that same experience at a conference in the US. 

Additionally, it sounds like the European audience is more than ready for machine learning-based solutions to take center stage. The fraud managers that we spoke with are tired of being coddled and want to dive into the data analytics and powerful insights that real-time machine learning can offer. Finally, the fraud management space is thick with buzzwords – everyone is talking machine learning. Businesses need to really understand how solutions differ in order to pick the vendor that works best for their needs. Not all machine learning solutions are created equal.

About Jason Tan

Jason Tan (@jasontan) is the Co-Founder and CEO of Sift Science, a San Francisco technology company that fights online fraud with large-scale machine learning. He previously served as CTO of BuzzLabs, a machine learning startup acquired by InterActiveCorp. Prior to that, he was an early engineer at two Seattle startups, Zillow and Optify. Jason graduated magna cum laude from the University of Washington in 2006 with a Computer Engineering degree.

 About Sift Science

Sift Science is a US-based technology company dedicated to making world-class fraud detection accessible to everyone. Sift designed its automated, real-time, large-scale machine learning solution to make finding and stopping online fraud as quick, easy, and accurate as possible.