Londoner Martin Sweeney started off at taxi app Hailo as a founding engineer. While there, he worked on a project to predict which orders were likely to be fraudulent.
Finding nothing suitable on the market to solve this problem, or at least with the speed or accuracy he and his team wanted, Sweeney and some colleagues decided to build their platform. Say hello to Ravelin.
Since the startup was founded in 2014, Ravelin raised £30m from investors and grew into a multi-million pound business, working with e-commerce titans like Deliveroo, Just Eat, Fiserv and Booking.com.
Now, with the pandemic driving the growth of e-commerce and therefore placing more pressure on fraud detection, Sweeney is convinced Ravelin’s growth is likely to continue at pace as the company is “perfectly placed” to help large businesses protect themselves against fraudsters, as he puts it.
City A.M. sat down with the young entrepreneur to find out where he wants to take Ravelin next.
Since March of last year fraud detectors have mainly worked from home, there is less oversight. Has fraud gone up significantly? Is it harder to detect?
It’s true that over the last year, fraud teams have had to work from home, which does make managing fraud that little bit more challenging. That said, the real reason fraud has become harder to detect is because of the huge rise in e-commerce, thanks to the pandemic.
Merchants around the world are now processing significantly higher volumes of transactions online at a much faster rate than ever before, and fraudsters are looking to take advantage.
As a result, since the pandemic all types of fraud have grown. Our own research found that online payment fraud has increased for 39 per cent of merchants over the past 12 months, while friendly fraud has increased for 41 per cent of merchants and account takeover increased for 48 per cent.
Your clients include Just Eat and e-Shopworld and you use machine learning to help online merchants and payment service providers reduce fraud and improve acceptance rates of orders. How automated is this process? Will bots ultimately take over? The reason why I ask this is, what I keep hearing from any cybercrime specialist, fraud detector or investor in your field, is the rapidly changing role of technology.
We work with lots of large merchants all around the world, to build a tailored machine learning-driven fraud detection system for each and every customer. That’s what sets us apart from the crowd. We don’t just provide a solution off the shelf — we build models from scratch and make sure it works exactly how you need it to. But for all our clients, the detection and authentication process is largely automated — and it needs to be.
Automation is the only way to accurately and cost-effectively detect fraud among the sheer velocity and volume of online payments and account logins, and to detect abuse with refund requests and promotion use.
The machine learning aspect of our technology is vital here — as it ensures our systems become even stronger and more accurate over time. But automation will never take over the whole process of fraud detection. We can’t overstate the vital role that humans play in fraud detection. The most effective fraud detection ecosystems comprise a combination of data and human insight and oversight because technology is only as good as the instructions it is given.
Humans are still behind the increase in and changing face of fraud attacks, so it still has to be humans that figure out how these fraudsters are operating. Technology simply aids this process and gives humans the information they need to make effective decisions to mitigate risks.
How are technological developments impacting your work and the way you serve your clients?
The nature of machine learning means it constantly evolves. Therefore, developments in the quantity and quality of data we can access allows our technology to work faster and even more accurately as time goes on. However, the way we serve our clients will always remain the same. We work with each client to understand its unique business model, operations and pain points before building a machine learning-based fraud detection system tailored for them, rather than providing an off-the-shelf solution.
One very obvious change in this relationship is there’s much less air travel than before.
Building a partnership and relationship with another company is very different over Zoom, whereas before we’d always like to see people in person. Our saving grace is that everyone is in the same boat.
The pandemic has impacted every sector, every industry. How have the coronavirus outbreak and subsequent lockdowns affected your company?
We’re very lucky that as a young, tech-native company, everything we did already worked remotely. In fact, pre-pandemic we had ‘work-from-home Wednesdays’, which inadvertently prepared us well for the more mechanical parts like data access and tools. But like most businesses, we’ve had to completely overhaul the way we work — setting up employees to service our clients from home, embracing a whole new way of communicating within teams, and finding new ways of keeping everyone’s spirits up while keeping the team culture alive and well. While we’ve changed the way we work, we’ve also never been busier.
There are now more dedicated professional fraudsters out there than ever before and the vectors of attack are becoming broader and more complicated.
Consequently, over the last year, demand for our services has increased dramatically in line with this rise in fraud. We’ve seen 100 per cent growth in revenue year on year, similar expansion in customer numbers and have almost doubled our employee numbers since last year.”
Let’s look ahead, life after lockdown, the market post-Covid. What are your thoughts on the recovery process?
The recovery process is likely to be rocky and uneven. We’ve seen a dramatic shift in the retail market during the pandemic, with lockdowns moving merchants and customers alike online. This rise in e-commerce won’t simply drop to pre-pandemic levels after Covid — it’s very much here to stay — and online fraud is highly likely to stick around as well.
It’s also worth noting that amid a global recession and with the furlough scheme coming to an end, more people are, unfortunately, going to turn to fraud and abuse to augment their spending power.
There’s also a higher risk of opportunistic fraud from genuine customers under increased financial strain. So fraud is likely to get worse before it gets better. The pandemic also taught us some important lessons about the way we work and the future workplace. Many businesses including Ravelin will be taking these into consideration going forward. That said, there’s no direct substitute for face-to-face collaboration and creativity.