Dave Excell, Founding father of Featurespace – Interview Collection


Dave Excell is the Founding father of Featurespace, with a background within the playing business, Dave based Featurespace following his invention of Adaptive Behavioral Analytics, which makes use of  explainable AI to assist banks acknowledge and flag suspicious shopper conduct. Even in latest instances as shopper conduct shifts, this superior AI has been capable of curb fraud and assist authorities deal with cash laundering and different organized monetary crime whereas bringing belief again to fintech.

Might you share with us the story of how in collaboration with Professor Invoice Fitzgerald you got here up with the idea of Adaptive Behavioral Analytics?

Whereas pursuing my PhD, I labored with Professor Invoice Fitzgerald on the College of Cambridge to use machine studying and statistical methods to know human conduct. Throughout my time there, organizations would come to us searching for novel options to numerous challenges that they had automating efficient choice making from information they captured or to enhance effectivity in guide processes. I started to note a sample: organizations throughout industries struggled with understanding the underlying conduct or ‘intent’ behind the information they captured, particularly when making an attempt to establish unhealthy actors. For instance, with one group we modeled the choice making of gamers inside a pc recreation to know in the event that they have been real gamers or robots dishonest the system. The extra initiatives we did, the extra I noticed this want for machine studying that will adapt because the conduct (and information) behind the end result (e.g. dishonest or fraudulent exercise) would change to keep away from detection. That is really how I first got here up with the idea of Adaptive Behavioral Analytics, which later turned the primary elementary know-how inside Featurespace.

Might you share the genesis story of how this idea then led to the launch of Featurespace?

Though I do get pleasure from researching and discovering options, I don’t get pleasure from analysis only for analysis’s sake. I’m motivated by making use of know-how to sensible issues, then discovering methods to ship industrial worth and deploying the know-how to make a optimistic affect on the world that we stay in. That’s how I ended up founding Featurespace and we’ve been on a mission since to make the world a safer place to transact.

Might you talk about the prevailing methods which can be utilized in the direction of fraud and monetary crime prevention, and why these methods fall brief?

There have been varied tech purposes within the area for some time–in truth, the primary makes use of of AI to combat monetary fraud date again to the early Nineteen Nineties. Nonetheless, that primitive model of AI assumed that fraud behaviors would keep the identical. The algorithms have been constructed to acknowledge the identical fraudulent conduct again and again. This similar idea is broadly utilized in anti-fraud know-how to today. However fraud isn’t static. Fraudsters are always adapting their strategies to remain forward of anti-fraud know-how. That’s why at Featurespace, we created the world’s first adaptive AI mannequin to combat fraud. We keep three steps forward of fraudsters with out requiring any human intervention.

Why is Adaptive Behavioral Analytics so impactful in comparison with these legacy fraud prevention methods?

Our proprietary Adaptive Behavioral Analytics are so impactful in comparison with legacy fraud prevention methods as a result of legacy gamers depend on static fraud patterns–however fraud isn’t static. Legacy gamers study what several types of identified unhealthy conduct appears like, then got down to detect these unhealthy behaviors amongst hundreds of thousands of transactions. The issue is that these fashions can solely consider unhealthy behaviors which have been seen earlier than, and fraudsters are always adapting their strategies to remain forward of fraud prevention. As a substitute, our Adaptive Conduct Analytics mannequin learns what good conduct appears like, then detects modifications towards these good behaviors. There’s rather more good conduct being finished on the planet than unhealthy, giving us extra to study from good conduct. There’s a a lot smaller set of fraudulent behaviors, and they’re always altering. To attempt to detect solely identified fraudulent behaviors is a shedding recreation.

What are the several types of machine studying algorithms which can be used?

Featurespace’s Adaptive Behavioral Analytics makes use of a mixture of unsupervised and supervised machine studying methods. Unsupervised methods are used to establish modifications in conduct to point possible danger. Supervised methods are subsequently used to optimize the accuracy of our fashions to stop and detect fraud and monetary crime. Final 12 months Featurespace launched Automated Deep Behavioral Community fashions that make the most of a novel Recurrent Neural Community structure. Featurespace Analysis developed Automated Deep Behavioral Networks to automate function discovery and introduce reminiscence cells with native understanding of the importance of time in transaction flows, enhancing upon the market-leading efficiency of our present Adaptive Behavioral Analytics.

How adaptive are the fashions to studying new shopper conduct and optimizing buyer profiles?

Our Adaptive Behavioral Analytics fashions are precisely as adaptable as they must be–even within the face of unprecedented change. For instance, throughout the preliminary COVID-19 lockdowns in 2020, shopper buying conduct modified actually in a single day. By April 29, 2020, Mastercard noticed a 40% enhance in contactless funds. Non-adaptive fraud prevention AI fashions have been thrown for a loop, blocking legit funds being made by folks ordered to remain at residence. Our fashions tailored routinely, with out human intervention. That is most evident by way of the TSYS Foresight Rating, a fraud- and risk-management decisioning scoring instrument for funds issuers, constructed by TSYS and Featurespace. From January-June 2020, the TSYS Foresight Rating with Featurespace constantly delivered steady rating distributions on a weekly foundation, enabling shoppers ordered to remain residence to proceed buying groceries and different necessities with no interruption.

What are the largest use circumstances for this know-how?

This know-how is particularly geared towards banks, monetary establishments and funds processors. For instance, funds processing firm Worldpay was lately acknowledged for its FraudSight product powered by Featurespace for its capacity to mitigate fraud whereas growing retailers’ approval charges and defending shoppers.

Is there anything that you simply wish to share about Featurespace?

Scams are one of many quickest rising fraud classes on the planet. Regulators are recognizing this and trying to place protections in place. For instance, the UK authorities launched a reform of the On-line Security Invoice in March 2022 in an effort to stop scams and enhance shopper confidence in on-line transactions. Equally within the US, the Client Monetary Safety Bureau (CFPB) is contemplating taking motion to guard shoppers towards scams by placing extra accountability on banks and credit score unions. By stopping scams earlier than they occur, Featurespace can save banks cash and maintain their clients secure, routinely with out human intervention.

An instance of that is NatWest, the fourth largest UK financial institution when it comes to whole property, with round 19 million clients. NatWest noticed a rise within the worth of fraud and scams detected, together with a right away lower in false optimistic charges (real buyer exercise declined), inside simply 24 hours of deploying Featurespace’s ARIC Threat Hub. Because of our partnership, they’ve cited Featurespace as a “sturdy associate” to their buyers.

Thanks for the good interview, readers who want to study extra ought to go to Featurespace.


Please enter your comment!
Please enter your name here