AG: It’s super important.
As I mentioned, we don’t see ourselves as a debt collection agency. We see ourselves as a software and AI company.
One of our co-founders, Qingchen, is a professor at the University of Hong Kong, specialising in machine learning. When Paul and I started looking more deeply at the debt collection industry, we did some research within academic literature on how machine learning has been applied in this domain. To our surprise, there have only been two academically reviewed research papers published over the past 20 years on applying machine learning in debt collection — one of which was written by Qingchen.
Machine learning is key for us because a lot of today’s debt collection agencies don’t have the resources to be smart about who to contact, when to contact them, how to contact them, and most importantly, what message to contact them with. They basically take a one-size-fits-all approach: on day one, send the letter; on day three, send an SMS; on day five, make a phone call; on day seven, make another phone call… and so on.
The reality is that every person is different, and every person’s debt situation is different. While one person might simply be forgetful, another might have a dispute. One might be in severe financial difficulty, while another might have a disability. By using machine learning, we can build bespoke strategies for every single customer segment we detect and automatically find a strategy that works for them.
Not only that, we can test and trial. Our machine learning models use reinforcement learning. Over time, as we see more data, we A/B test, we test and trial, and as we notice certain outcomes performing better, we optimise our approach so that over time, our engagement improves as well. That’s something most companies just don’t do in this space.