Machine learning and data are powering Monzo's fintech disruption
App-only fintech startup Monzo is scaling up its machine learning capabilities and plugging employees into the rich stream of data generated by 1.7m global users as it forges ahead of the high-street banks in the digital space.
How machine learning and data are powering Monzo's fintech disruption
Neal Lathia, data scientist and machine learning lead at the UK-based bank, told The Drum how his team is delivering “incremental gains” in the app, in customer and the product using smarter automation since he took the role on one year ago.
Lathia, previously lead data science at SkyScanner, commands a team of more than 20 people. When he joined there were only two. Now he is looking to harness the wealth of data at the company’s disposal to get Monzo “working faster”.
It contrasts with the role a data scientist at a high street bank may have, weighed down by legacy systems, corporate baggage and sheer scale. Lathia likens what his contemporaries there do to to “steering the Titanic”.
The Guardian reports that Monzo’s value is set to double to £2bn when it puts the ink on a 100m investment from an unnamed American entity. This comes just months after bringing in £20m in a customer crowdfunding campaign, leapfrogging the valuation of rival Revolut that has been weighed down by numerous PR crises of late.
In pursuit of profit, Monzo, like many startups, needs to aggressively pursue scale and deliver overdrafts and other products to increase its margin. Monzo's attraction is that setup is easy, it currently has no usage fees, and carries no charges for overseas spend. Monzo doesn’t process mortgages or savings accounts – many of its users still have bank accounts and use it to manage daily spend instead.
It also lacks a high street presence. As a result, the app is both the first and last point of contact customers have and there is, therefore, a greater emphasis on user experience.
This is where the startup has to be smarter than its competitors. To succeed, it has to be easier to use the app than the other financial solutions available on the market.
Monzo provides a comprehensive spend report on its members' smartphones. Handy push notifications, bill splitting and immediate cash transfers further sweeten the deal.
Machine learning techniques are being more widely implemented to make these operations more seamless. This is particularly important with customer support, where agents are aided by machine learning scripts that improve the more they process queries, rather than automating everything through chatbots.
Lathia says: “When a customer writes to us 'why was this transaction declined' or 'I lost my card', we use machine learning to check what the customer has sent and generate possible responses for our customer service agent.”
The language of the written query (most complaints are text-based and are sent in the app), is scanned by the system and responses are generated. This helps speed up how requests are processed – users cannot be denied access to their funds or be sitting with the chat box open forever. It's not a good look for a disruptor brand that has to build trust and signal reliability.
Lathia believes "chatbots were very over-hyped a few years ago and still haven't found their rightful place in the world". Instead, he says, Monzo prefers to use machine learning "to help both sides talk to each other better, rather than just talking to a machine".
The programme measures the volume of queries submitted to the team, and agents tag their nature. Once a week, the team analyses the complaints and uses the patterns to fast track improvements to the service.
An example of customer feedback improving the app is one of the latest shipped features: customers now do not need to get in touch to change their default email addresses.
“They had to go through a security flow so we shipped a new version of the app that allowed customers to do it themselves in the app," explains Lathia. "They didn’t need to get in touch with us and so that feels like a win-win, it frees us up to work with other queries."
The wider applications
However, Lathia's data team is now branching out to new areas, helping internal analysts increase the efficiency of financial crime investigations and also creating tools that speed up engineers' ability to ship new code.
“Nearly every team is using data to help them achieve their goals to access data and see what is going own. We are setting a very early precedent in terms of people being able to access data and see what is going on. Our mindset is about enabling how to get our teams their data quickly.”
It is his intent to plug as many members of staff as possible into the relevant data, even if they may need some help interpreting it. "You could lock away all your data and then it would be super secure but there would be no value in collecting it in the first place. It is finding that right balance," he says.
The high street banks, meanwhile, face "the very different problem" of centralising data across sprawling ecosystems.
"They were founded many decades ago when cloud computing didn't exist. We are lucky that we are only four years old and were built from the ground up with things like BigQuery, AWS, and Looker.”
Amid all of the possibilities the data team can deliver, its implementation into Monzo’s marketing is still at an early stage according to Lathia. “We are using data to quantify the impact of campaigns rather than using it to try and alter them.”
In the app, there are efforts to create ways for users to tell their friends about the service and incentivise them to join. However, it does not a massive leap of the imagination to assume the brand will use the customer service tickets to identify popular features and inform its marketing from that. The bill-splitting feature, for example, is massively convenient and how Lathia attracted his friends into the ecosystem.
New players like Monzo and Revolut, both founded in 2015, have ambitions to “turn the financial banking sector on its head”.
To disrupt the industry they must be careful not to adopt the worst traits of the multinational corporations they are looking to displace. Joel Biswas, a planning partner at Coley Porter Bell, recently wrote in The Drum that their path represents a marathon, not a sprint, indicating that reliability and trust are just as important factors as the speed of disruption.