Home »

Opportunities within Machine Learning & Payments Data

Jonathan Orritt, Data Scientist, Santander UK

Jonathan Orritt, Data Scientist, Santander UK

Many of us experience the impact of Machine Learning every day, ranging from the recommended posts we receive in our social media feeds to the emails that drop into our inbox. Every one of us is being recommended and classified into certain buckets by potentially hundreds of different machines around the world, most commonly within our Social Media (SM) Profiles. Much of what happens within social media and the successes these systems are experiencing can be applied to Payments Data. In a way, Payments Data is a form of interaction like those interactions we experience on SM; Ultimately, Customers spend money at the merchants they enjoy, much like spending time with the friends you enjoy, for example you can redevelop the recommended friends algorithm you may find on Facebook to recommend different Merchants to you. The same can be applied within Payment Interactions, from here we can build significant Analytical Insights into our Customers, to greatly enhance our business simply by reusing our Payments Data differently.

Payments Data harnesses the best possible opportunity within modern day Payment Service Providers (PSPs) to deliver personalised opportunities to our customers. Most of these opportunities can be split into two primary strategies and some examples are shown below.

1) Acquisition Strategies

Sell more of your products or services to those that need it.

i.e. Classify customers most likely to own a Mortgage based off their Payments. This can come from the obvious such as Direct Debits to other Mortgage Providers, Demographic Data etc. or the less noticeable such as Life Insurance Direct Debits, High DIY Spend etc.

2) Protective Strategies

Managing any form of risks within Operations & Customer Transactions

i.e. Greater support to credit consumers on managing their finances automatically and setting realistic spending controls to aid with credit repayments.

"Providing a more personalised service or product to our customers through recommendations requires a level of trust between consumer and PSP to ensure those recommendations are relevant and achievable"

Analysing Consumer Payments Data is something we do all the time to protect our customers, be it in the fraud or financial crime, or in helping the vulnerable, but it can also be used to drive analytical insights for our customers. The difficulty here is ensuring the promotion of that data product is beneficial to the consumer and is available. Providing a more personalised service or product to our customers through recommendations requires a level of trust between consumer and PSP to ensure those recommendations are relevant and achievable. Achieve this and PSPs can find themselves being able to predict what a consumer needs before they tell you (or more importantly before they tell anyone else) Transactions are the first indication of a change in situation or behaviour, sustained change can lead to acquisition opportunities and thus the chance to offer the consumer the product before any other. An example would be Parental Changes, having a child as many of us will know can be quite a dramatic financial change. From a Data Perspective, we can also observe this change, increased food spends, increased spending at Baby Clothing Retailers etc. All these features can be built and tested to greatly enhance the service and products we offer to our customers at the most important of times. However, Machine learning in Payments can only really be successful if certain conditions are met beforehand:

1) Data Consent – do you have the approvals from your customers to analyse their data? Things don’t get much more sensitive than Payments Data. Run focused pilots on different consumer groups to test their responses to these payment insights.

2) Data Availability – Do you have the data available in a standard format? Payments Data is managed in many different standards and systems, so getting that single unified view can be difficult, but ultimately there is great value in a common layer.

3) Data Scalability – Payments Data is probably the biggest repository PSPs manage, so running

predictive modelling on these datasets can take quite some time. Having access to Data Warehouses and Data Science Solutions sitting on the edges of those warehouses are crucial, ultimately allowing you to analyse Payments Data & release models at scale.

For UK Retail Banks, we must look to enhance the quality of service and products we provide to our consumers with data, whether we give them greater insights into their spending, deploy machines to track or recommend new products to consumers or even help them manage changes in their own personal life. Consumers are increasingly becoming more data & digital savvy, our ability to give them greater insights into their spending and others around them is greatly increased. This increase has driven consumers resistance to traditional engagement channels, our ability to influence the consumer is determined by the digital channel we can advise them through. Ultimately, data has become the most valuable asset to PSPs and is currently fuelling competitive advantage for many who chose to use it. The value of Payments Data within UK Retail Banks will only increase further as consumers move towards the data-rich channels of Cards & Faster Payments, giving us greater opportunity to personalise the product being offered to our customers. This increased opportunity to predict what a consumer will do next will help us protect the customers better and, ultimately determine PSPs successes within the field and test their true ability to process vastly more payments data in substantially less time.