While big data has been a popular buzzword among industry sectors over the past few years, adequately collecting and analysing customer information to gain key insights through the use of Artificial Intelligence (AI) is not something all organisations have managed to get right.
However, for Frankfurt-listed fiintech MyBucks, investing in AI has been a priority from the get-go. “AI has allowed us to not only position our products and services to respond to our customers’ greatest needs, but are increasingly able to accurately predict whether a customer is likely to pay back a loan.”
This is according to Dr Richard van der Wath who heads up the MyBucks in-house AI team as chief data officer (CDO) and who – alongside several other AI experts – continues to implement pioneering approaches, one of which is MyBucks’ implementation of a Neo4J graphing database.
“Graphing databases are essentially databases that are designed to not only store data but also classify and store related data. This relational functionality means that loan data is not stored in one single table, rather allowing us to connect clients and loans to one another across various platforms.
“This type of database brings with it two significant organisational benefits, namely speed and insight. Not only does it allow insight into client behaviour for marketing purposes but also a means to guard against potentially fraudulent transactions in real time through exploiting leading-edge technology found in the Neo4J architecture,” he comments.
The MyBucks anti-fraud service, known affectionately as Dexter, has already collected over 100-million unique relationships, each of which van der Wath says bind events and actions together in a web that characterises their clients and markets in fine detail. “This web of course keeps expanding by the minute,” he adds.
By conducting checks in the background, van der Wath explains that Dexter groups customers according to their MyBucks e-wallet activity as well as potential association with someone who is a known fraudster.
“From a business point of view there is a level where we wish to deal with unconnected clients only – when connectedness points to fraudulent behaviour and is inextricably built into Dexter’s reality,” he says.
Here van der Wath explains Dexter’s functionality is similar to that of the concept of ‘six degrees of separation’. The theory based on the idea that anyone on the planet can be connected to any other person through a chain of acquaintances which has no more than five intermediaries.
“Similarly, Dexter uses data relating to location, loan states, ID numbers, email accounts, IP addresses and bank account interaction, to group people into various categories. As technology continues to expand this list will continue to grow.
“This grouping functionality lies at the core of our business and based on this we are able to make informed decisions about whether to approve loans and disburse the necessary funds,” he adds.
In addition to conducting background checks and categorising clients, van der Wath says that by implementing AI solutions to monitor registration behaviour, they are closing the loop on fraudulent loan applications.
“While fraud will always be present in some form or another, fraudulent behaviour by its very nature leaves behind a unique trail in its wake. And it is this trail that will for ever persist between the nodes that makes up Dexter’s conscience.
“Essentially, the fraud functionality attaches a quantitative value to each registration on the basis of similarity with every other client. Here it analyses registration detail and logon patterns in real time to attach a suspicion number to a registration.
“Any application breaching this level of suspicion is passed on to a bridged process where the application is held for further review. Moving forward, however, biometric information such as online face recognition will add even more value to this kind of fraud service,” he says.
For van der Wath, AI ultimately presents the means to manage a business in a very quantitative and scientific manner, arriving at the sweet spot where risk, return and shareholder expectations are optimally met.
“Our client base is demographically diverse and so, AI allows us to arrive at a pure representation of our target market in the far reaches of online media. Repeatedly modelling client behaviour enables greater insight and understanding of our clients. We believe this will prove a key differentiator when it comes to product development, service delivery and risk mitigation,” he says.