A data gateway to insights: Current accounts and the case for ML

Once a mainstay of the banking industry, current accounts are increasingly perceived as a utility product. Despite their maturity, they remain the second most complained about banking products in the UK, right after PPI. In recent years, costs have increased for everything from overdrafts to transaction fees, with average customer accounts unwittingly racking up £152 per year in hidden bank fees each year. At the same time, as banks proactively de-risk, customers have been making unfavourable headlines with their stories of having accounts closed with no explanation.

Compounding these issues, switch rates remain low – about 3% annually – and not for the right reasons. It’s notoriously difficult and time-consuming to switch in the UK, and customers don’t foresee any financial benefits after making the effort. In fact, over the past few years, only Nationwide has made good headway on this front.

Regulators are pushing for change – or certainly attempting

Regulators have clamped down with challenges on fees, minimum service level requirements, and mandates for cost transparency and reporting.This gives customers a degree of transparency into their current account service, tools to objectively compare offerings across banks, and the ability to switch banks quickly and easily. And with the passing of open banking legislation, customers can now choose to have their account data shared with other banks and third parties. Regulators are of course hoping, open banking will drive competition, which will necessitate real innovation that benefits all parties.

But barriers to digitisation continue to stall creative efforts, chief among them being poor collaboration between business and IT and the lack of integration between new and existing technology and data. As a result, we’ve only witnessed baby steps in the name of innovation – for example:

• Attractive switching schemes – such as sizeable cash bonuses for customers opening new accounts
• Alternative incentives – such as courses, holiday vouchers, and high-end gadgets
• 'Open Banking' innovations – such as the development of smarter banking apps that help people manage their money more effectively (like HSBC’s initiative to create an app that shows customers their bank accounts, credit cards, mortgages and loans from 21 different banks – all in one place).

However, given the pace of technological change today, these baby steps will quickly become table stakes and much of it needs to be proven with numbers in the P&L.

Fintechs such as Coconut, offering niche current accounts optimised for the self-employed and freelancers – a rapidly growing market segment in the UK – are innovating data driven, cloud-based account services that could dramatically impact people’s lives and solve real business problems. At the same time, new banking propositions are already harnessing the power of big data to transform traditional banking services with digital solutions. Examples include Monese’s banking option that lets people bank without a local address, Starling, Fidor and Monzo to name a few.

So what does all this mean for established banks?

Clearly it’s time to shake up and harness the extraordinary current account data to understand their customers and innovate to stay ahead of nimble, digital competitors. Machine learning (ML) is the key to making this possible.

Current accounts are the gateways to deeper customer insights

Machine learning is creating new possibilities for banks in terms of what they can do with their terabytes of customer data. They should already be mining their current account data for deep customer insights and delivering those insights in a personalised way at the right time, and through the right channel, to bring new value to customers.

At the core of their relationship with banks, customers leave behind a stamp of their activities (transactional and channel interactions), behaviours, and preferences. The result is massive amounts of data ready to be mined.

But are customers getting proportionate value back from the data that they willingly give to their bank? Isn’t it time for banks to deploy advanced machine learning algorithms to predict patterns (supervised ML) or give new, unknown insights (unsupervised ML) to make the relationship real-time, pre-emptive and proactive?

Within the current account product line, examples of ML use cases could include:

• Customer financial health management tools – to keep customers abreast of their financial health, as well as pre-empt future risks and exposures to fees to the individual. For example, ML could be used to help customers continuously monitor cash flow, determine optimal account balances given standing orders and transaction patterns, predict balances, foresee difficulties, and more. 
• Current account product performance monitoring – to give customers real-time visibility into a product’s service levels relative to targets and customer agreements. Machine learning can be used to monitor performance in real time, when services are compromised, the availability of alternative options, etc. This is key given new regulatory expectations governing how long it takes to open an account, resolve customer issues, and more.
• Customer retention and value-added services – to identify customers who may be at risk of leaving the bank. Machine learning can help banks identify and address root causes of churn and the best ways to improve retention, proactively offer loyalty offers, and gain deeper insights into customers and their behaviors. These insights can be used to recommend and proactively market value-added partner services (such as receipts management, accounting, and concierge services) to the right customers.
• Complaints management – to pre-empt complaints and transform customer complaints management into a business-enabling capability. Machine learning can analyse customer behaviour to help proactively head off complaints, identify patterns of complaints, monitor customer sentiment and behaviour after issue resolutions, and identify and pre-empt the next potential mis-selling scandal.
• Fraud and financial crime management – to detect transaction anomalies and keep customers safe. Machine learning can help by detecting customer behaviour anomalies, differentiating high-risk behaviors while minimising false positives. The HSBC AI implementation on this agenda has already caught the market’s attention.

But ML ruccess requires a right implementation approach and mindset

As noted in my earlier blog , success with AI and machine learning-driven innovation is not a given. The use case and business justification must be robust and have strategic business relevance. There must be business readiness to implement and adopt ML technologies and preparation to deal with the operational and operating model impacts.

The ML opportunities are huge, and starting with current accounts, as the primary relationship product could deliver clear monetary benefits. Innovative technology vendors and change organisations are on stand-by to partner with banks, customers are on stand-by hoping banks will grab the opportunity, and regulators have called out the need for change. How long can your bank wait?

ML could be the key to unlocking the gateway to insight and innovation. Isn’t it time to call the “digital locksmiths”?