London | Tech geek Dave Excell left Canberra in 2004 for Cambridge University intending to do a PhD and enjoy life in the ivy-clad cloisters. But things didn’t go quite according to plan.
He arrived in England with a BA in engineering and IT from the Australian National University, three suitcases and a burning interest in artificial intelligence – which back then was more of a wild-west stretch of the tech landscape than it is today.
Barely had he unpacked the bags and opened the books when he found himself helming a fraud-fighting AI start-up, one that now operates across five continents and is said to be closing in on unicorn status.
It’s a classic start-up tale: a bright idea that takes wing on several gusts of serendipity. But it is told against a backdrop of the gothic spires of Cambridge, where his company, Featurespace, is still headquartered.
Excell’s doctorate was going to look at how computers might be trained to look at the world the same way as humans.
Sitting in the café of a shared-workspace building in central London, he explains his erstwhile scholarly passion.
“As individuals, we could be sitting in an environment like this and very quickly we know who’s on a Zoom call, who’s a tourist, who’s here for a business meeting, who is being interviewed by a journalist. There are these visual clues that we as humans can pick up really quickly,” he says.
“But how would you translate that into a computer-based system for it to be able to have that same intuition about the environment around itself?”
His PhD supervisor was Bill Fitzgerald, whose speciality was applied statistics. That was his first bit of luck. The professor was able to round out Excell’s programming aptitude with expertise in data and how to crunch it.
They were looking at computer games and how to tell whether a player was genuine or a cheat, someone trying to game the system. Then came the really lucky break.
The big bet
Cambridge has a reputation as a hothouse for scientific and tech innovation, which lures companies to regularly jump on the train up from London to scout for ideas that are still in bud.
The tech team from online gambling company Betfair turned up, and one of the questions on their minds was how they might fight the increasing prevalence of fraud on their rapidly growing platform.
They saw potential for their business in what Excell and Fitzgerald were doing and gave them a sandbox in which to do it.
“We didn’t come from a background of payments and fraud. We came from an academic background, so Betfair helped educate us around the subject,” Excell says.
“That’s where we really started to learn the craft of preventing fraud. And we gained an understanding that market opportunity and how the technology could be applied there.”
One thing they learnt was that fraudsters move quickly, adapting to detection systems and bypassing them. This forces companies to rejig their products, leaving cat and mouse trapped in a vicious circle.
But Excell and Fitzgerald came up with a circuit breaker.
“We flip the focus around and say, well, 99.9 per cent of our customers are genuine good customers – why don’t we learn what their activity and behaviour is and, from that, identify the 0.1 per cent?” he says.
A genuine customer will always back their favourite team, or almost always take a roughly similar level of risk. But a person using someone else’s money will act differently, often more recklessly.
It seems obvious now, but AI was still in its infancy, and most models were pretty static. At Betfair, Featurespace developed the early versions of a more dynamic and self-improving software. And Excell started to think the market might be pretty big.
“We started thinking, where’s the home for the technology?” he says – and it wasn’t hard to spot that the financial services sector is the front line of fraud.
In Britain, the losses from fraud in the financial services sector hit £784 million in 2020. Banks prevented another £1.6 billion but, even so, more than £3 in every £10 is still slipping through.
Back in 2005, banks were using fraud detection systems geared to the offline world. These were cumbersome, slow-moving and generated too many false positives – which led to many legitimate transactions being blocked as potentially fraudulent.
Featurespace – named for the AI term that describes the set of elements that a program or algorithm is taking into account – began marketing to British banks and financial firms, and also raising seed money.
The American dream
The global financial crisis of 2008-09 could have been dangerous for a bank-oriented start-up at Featurespace’s phase of development, but Excell’s niche happens to be pretty impervious to cyclical forces.
“One of the interesting things about fraud is that it is relatively recession-proof and industry-proof. Even during COVID, we still saw fraud was prevalent,” he says.
“Especially during a recession, if you have more people out of work, it can drive people to do things that they wouldn’t normally do.”
As with any start-up, achieving scale means cracking the United States. It had always been Excell’s dream to follow up his Cambridge experience with a move to Silicon Valley.
His dad worked for Microsoft in the 1990s, and Excell says he’s always imagined following in his footsteps by getting involved in a West Coast tech company as it put pedal to the metal.
Again, things didn’t go quite to plan. In 2017, he did move to the US, but not to California. He went to Atlanta.
He thought about San Francisco and Seattle, among others, but he was drawn to Atlanta’s universities and its high concentration of payment processing firms.
Excell no longer runs the company, having relinquished the CEO role in 2012 (his founding partner, Fitzgerald, has died). He solves problems, strategises and pitches to potential clients – he’s the “product evangelist”, as one of his colleagues called him.
The company earned $US26.5 million in revenue in 2020, and has more than 400 employees and 50 customers. It has raised more than $US100 million in capital and is understood to be nearing a valuation of $US1 billion.
Because some of its customers are payment companies such as TSYS and Worldpay, which serve a wide range of banks and finance firms, Featurespace claims more than 97,000 financial institutions now use Featurespace tech.
Excell is a long-term expat, but his company at least has come home. Featurespace took on eftpos as a client last year, and is also working with financial services fintech Shaype.
Fraud is not going away and is probably getting even worse. This is unfortunate, but also does mean there’s a perennial appetite for Featurespace’s product.
Although the number of companies offering machine-learning weaponry to fight fraud is growing, Excell says he is confident the pie is big enough for Featurespace to keep scaling up.
If the past is anything to go by, however, there’s still the potential for things not to go quite according to plan. For Excell, hopefully again in a good way.
Speaking of which, he never did finish that PhD. “I often just say, ‘I studied at Cambridge University,’ ” he says with a smile – but no regrets.
Fetching latest articles