Why We Invested in Featurespace

Nearly two decades ago, the late Bill Fitzgerald, a professor of applied statistics and signal processing at Cambridge University in the UK, was mentoring Ph.D. student Dave Excell as he worked on applying an understanding of human behavior to statistical processes using machine learning. The models were predictive; data on a person’s individual behavior would be used to establish their “normal,” and once the statistical profile was created, the system could use it to determine whether that person’s other behaviors were true anomalies or irrelevant deviations. This was the invention of what is now known as adaptive behavioral analytics.

Even with years of fintech investing under our belts, we had never seen anything like adaptive behavioral analytics in 2015. Machine learning was relatively new, and Featurespace was in the early stages of applying its technology to payments transactions.

But after meeting with the team in Cambridge, we felt like they were on to something. As it turns out, Visa would arrive at a similar conclusion earlier this year, when they announced their intent to acquire the company: Featurespace’s groundbreaking technology would be a game-changer for fraud detection. 

Interestingly, Featurespace’s technology wasn’t initially built to prevent payments fraud and mitigate financial crime risks. Bill and Dave applied adaptive behavioral analytics to a range of test data including for surveillance, customer churn, and crop yield predictions. The first notable commercial use case was in 2008, when they partnered with Betfair, one of the largest online gaming companies in the UK, to identify fraudulent bets. Featurespace trained its models to develop an understanding of each user’s normal activity – what types of bets they placed, how often, what time of day, and how much money. Once the models learned which activities were normal and which were anomalous, Featurespace was able to identify genuine player behaviors leading to significant reductions in false positives while simultaneously identifying new types of fraud. Betfair could then deliver the appropriate intervention to the customer. 

Featurespace’s work with Betfair was incredibly successful, and they believed that adaptive behavioral analytics could be applied to other industries as well. That’s why, in 2015, the company completed a test exercise to identify fraudulent transactions for TSYS, one of the largest payment processors in North America and one of our strategic partners here at TTV Capital. TSYS wanted to understand whether adaptive behavioral analytics could reduce the rate of false positives, which at the time was around 13:1. This meant for every one transaction that was correctly identified as fraudulent, there were 12 times when customers’ cards were incorrectly denied when attempting a purchase. Naturally, having a card declined is embarrassing, stressful, and erodes trust in the financial institution, so TSYS was actively looking for more sophisticated solutions to lower the ratio of error. 

The results were unprecedented. Featurespace generated false positives at a rate of 3:1, far better than any technology that was available. Since Featurespace was looking for investors at the same time, TSYS connected them with us and shared their use case. 

TSYS’ ability to stress-test the technology and independently vet how the product would work in real-time financial fraud scenarios was invaluable. Because TSYS validated the core functionality before it was even on the market for the financial services industry, we were excited about the opportunity and made our initial investment in Featurespace in 2016. 

When we look back, it’s incredible to see how Featurespace has evolved – both from a technological and business perspective. Less than a year ago, Featurespace developed the financial services industry’s first generative large transaction model (LTM), aiming to do for payments what LLMs have done for language. But there’s even more to the story than groundbreaking technological innovation. As people who have come to know both Dave and Martina King, Featurespace’s CEO, we are just as impressed by their company culture and the values they’ve fostered. 

Featurespace is consistently at the forefront of innovation, and we’re eager to see how they will accelerate their mission to make the world a safer place to transact as part of Visa. All of us at TTV are proud to have played a small role in supporting Featurespace’s groundbreaking work to fight fraud and financial crime, and we’ll continue to cheer them on throughout this next chapter.