Tell me about Featurespace, and how you came to start it.
Our innovation began in the Engineering Department at the University of Cambridge, U.K., where I was interested in the cross-section of applying online statistical analysis to the interpretation of human behaviors. During this time, the amount of data being captured on the Internet exploded and I explored different applications of the research.
Fraud became an exciting use case, as existing systems were centered around rules or static models that didn’t adapt or evolve with the change in human and ‘fraudster’ behaviors. These systems were developed against known historical types of fraud and couldn’t adapt to the changes required to stop those perpetrators who changed their tactics. Over time, this resulted in a poor experience for genuine customers, as their transactions began to be blocked.
With that in mind, we developed the technology, adaptive behavioral analytics, for commercial use as a tool to improve manual processes that were being used to catch fraud. By ingesting a person’s individual behaviors to establish a statistical profile for what was “normal” behavior. This provided a benchmark against which the system could determine if certain behaviors were anomalies or irrelevant deviations.
This caught the attention of Betfair, an online gaming company, who were looking for a way to detect and prevent fraudulent transactions. After the initial success with Betfair and eventually other gaming companies, we knew we were onto something special – not just in gaming, but in any sector that suffers from fraud issues. And financial services was one of the most obvious.
What made Featurespace expand into the U.S.?
There are a couple of factors that contributed to our decision to grow our footprint. Our client base in the U.K. expanded quickly and we received a lot of recognition in advance of our move to the U.S. late last year, including Deloitte’s Fast 50 2017, The FinTech 50 2017, the European Business Awards‘ Ones to Watch 2017, 50 Smartest Companies of the Year 2017 by the Silicon Review and many others.
We felt we were truly having a positive impact by helping companies reduce fraud with the most advanced adaptive behavioral analytics and maximize revenue opportunities by accepting more genuine transactions. As our reputation grew, we recognized that many of the fraud issues we had been successfully solving in the U.K. were becoming more prevalent in the U.S. and felt that it was an ideal time to test our abilities in an entirely new market. For example, card-not-present fraud, which followed the migration to chip cards, was something we were helping our U.K. clients combat. By the time it came to America, we already had several years of experience in addressing it.
We had the opportunity to meet with some major financial services companies in the U.S. (Vantiv and Worldpay (prior to their merger), TSYS and Ally Bank) and participated in more than a few competitive rounds of evaluation. And one by one, we outperformed the others.
These successes were so incredibly rewarding, and we knew delivering on our core beliefs – which include taking a proactive, service-driven approach and surprising and delighting our clients – wouldn’t be possible without establishing a presence here. And as part of that, we needed bright, talented individuals on the frontlines who buy into our culture and share our desire to exceed our clients’ expectations.
What issues are financial institutions and payment processors facing today, and how does machine learning help address them?
One of the biggest issues is false positives, which occur when a customer’s genuine activity is misinterpreted as fraudulent and the transaction is subsequently declined. That has a proven, negative effect on the customer experience and loyalty to his or her financial services provider.
This happens largely because of outdated fraud detection systems, which rely on static machine learning models or rules that flag any transaction that falls outside of very rigid, predetermined parameters. In contrast, adaptive machine learning allows us to create sophisticated user profiles based on hundreds of data points that sculpt a more accurate picture of an individual customer’s habits. Within milliseconds, our platform can determine if a transaction is genuine or fraudulent, stopping fraudulent attempts before they eventuate – a real-life representation of Precognition from Minority Report. This creates a significant opportunity to prevent fraud, preserve relationships and produce additional revenue that would have otherwise been lost due to inadequate or outdated systems.
What’s a common misconception about machine learning that you’d like to clear up?
Many people see machine learning as a threat to jobs. This isn’t a new concern, but the World Economic Forum (WEF) estimates that machines will create 133 million new jobs globally by 2022, compared to the 75 million it may displace. The WEF also said that it will “vastly improve” the productivity of existing jobs, which is a welcome sight for businesses in the financial space.
Artificial intelligence and machine learning can automate many menial tasks and increase the accuracy with which those tasks are completed. This allows human workers to be more efficient and thereby important, while also freeing their cognitive capacity for critical and creative thinking and applying personal knowledge in a given situation.