How banks can use data analytics to identify early signs of financial trouble

At the start of August new rules come into force in the UK setting out tougher consumer protection standards for more than 50,000 financial services firms.

These rules, imposed by the UK's Financial Conduct Authority (FCA) under its new Consumer Duty policy, establish an obligation on firms to deliver 'good outcomes' for retail customers while taking into account the characteristics of those customers, including their 'vulnerabilities'.

The new obligations represent a challenge for financial services firms at a time when a cost of living crisis is stretching personal finances. Data from consumer organisation Which? for March 2023 revealed that some 2.5 million households, eight per cent of all UK homes, missed or defaulted on at least one mortgage, rent, loan, credit card or other bill payment that month.

Yet our research (conducted with Edika Quispe-Torreblanca, of the University of Leeds, John Gathergood, of the University of Nottingham, and David Leake, of Lloyds Banking Group) suggests that many financial services firms have access to an under-exploited trove of transactional data that could be used to better understand customers and help discharge the firms' regulatory duties.

How data analytics can reveal our spending habits

In the developed world people spend a significant proportion of their lives online. From doing the grocery shopping to booking a holiday, work and leisure, banking to healthcare, our lives are mapped out in a trail of digital interactions.

Potentially these details are a source of valuable information that many organisations fail to make the most of. For most companies collecting and storing this information is part of the administrative routine of their business. The key to unlocking its value is understanding the right questions to ask and giving careful consideration to what patterns and associations might be hidden in the data.

In our case we wanted to see if people's spending patterns might tell us something useful about their likelihood of defaulting on payments and getting into financial distress.

To investigate this we decided to look at one specific aspect of a person's spending – what their spending habits are. The idea being that people who have very chaotic or complicated lives may find it difficult to stay on top of that complexity and are therefore more likely to get into financial trouble.

The more widely a person's spending is distributed across different categories in a given period the more difficult it is to predict their spending from one period to the next. We call this unpredictable buying pattern 'spending entropy', as the term entropy is often referred to as a measure of disorder. High entropy – more unpredictable spending – is where the spending spread across different categories is greater than average. While low entropy spending is where spending is more focused on a few categories and more predictable.

Our reasoning was that if a person's spending behaviour is more chaotic – and so less predictable – then that could be indicative of a disorderly lifestyle or personality that might adversely affect their personal finances.

Our study draws on two main sources of data: the Argus Credit Card Payments Study and customer data from  Lloyds Banking Group, which includes not only Lloyds Bank, but Halifax Bank and the Bank of Scotland too. The Argus study collects data from all the credit card providers in the UK. Our sample covered a two-year period and included 190,429 individuals along with 197,179 credit cards. While Lloyds Bank Group’s data covered personal current account and credit card details for 100,963 people in 2018.

Using the data we were able to look at the distribution of people's spending on a month-by-month basis across a range of spending categories including retail, travel, household bills, financial services, and hobbies and interests.

It was then possible to see if there was an association between unpredictable spending behaviour and problems with managing personal finances. An individual was considered to be in financial trouble if the data showed they missed payments on credit cards, personal loans and mortgages, or exceeded overdraft limits.

The results showed a clear association between the unpredictability of spending behaviour and the likelihood of getting into financial distress. High spending entropy in one month is a good predictor of missed payments and financial distress in the following month, as well as three months down the line. And the greater the deviation from the average towards higher spending entropy, the greater the effect.

To give a rough idea of the size of the effect, in monetary terms the impact of moving from average to high spending entropy on the chances of missing a credit card payment would be similar to a reduction of £200 in monthly income. Or, for the chances of missing a loan payment, it would be similar to a reduction of £3,300 in monthly income. 

The exact mechanism for the relationship between spending predictability and short-term and long-term financial distress is uncertain. However, our results suggest that it is more likely to be a personal trait that persists over time, than a state that changes from month to month depending on the circumstances.

Although, it is also important to note that this is an association revealed from the patterns in the data – not everyone with high spending entropy will fall into financial distress, just as not everyone with low spending entropy will avoid it. It is just that the data suggests those individuals are more likely to miss important financial payments than people with average or low entropy scores.

How banks can protect vulnerable customers in new FCA regulation

The  rules being introduced in 2023 by the FCA include an obligation for financial services firms to consider the characteristics and objectives of their customers and how those customers behave, including 'vulnerable customers', at every step of the customer journey.

A vulnerable customer, according to the FCA, is “someone who, due to their personal circumstances, is especially susceptible to harm, particularly when a firm is not acting with appropriate levels of care”. It's a fairly broad definition that covers a lot of consumers, with the FCA identifying four factors likely to be linked to vulnerability - health, life events, resilience and capability.

The provisional results of the FCA's 2022 Financial Lives Survey suggest that almost 30 million UK adults (46 per cent) exhibited one or more characteristics of vulnerability as of May 2022.

The FCA's new Consumer Duty means that financial services firms must not only aim to have good customer outcomes but also understand and evidence whether those outcomes are being met. That duty extends specifically to vulnerable customers. It is linked to a longer term strategy that reflects the FCA's aspiration to be more data-led and agile, allowing it to intervene earlier to reduce and prevent harm.

Our work fits neatly with these objectives as it shows how routinely collected administrative data can be used by firms to try to gain a better insight into the lives of their customers in order to deliver better outcomes.

Using spending entropy offers an additional tool instead of intrusive surveys of customers that rely on self-reported responses and more opaque black box machine learning models.

Financial firms can use transactional data that they already possess to analyse their customer's spending behaviour and help identify customers that are likely to get into financial distress and that may be vulnerable. That's a use case that should interest managers at both the financial firms and their regulators.

A version of this article was published in The Banker.

Further reading: 

Muggleton, N. K., Quispe-Torreblanca, E. G., Leake, D., Gathergood, J., & Stewart, N. (2020). Evidence from mass-transactional data that chaotic spending behaviour precedes consumer financial distress. PsyArXiv. 

Muggleton, N., Parpart, P., Newall, P., Leake, D., Gathergood, J. and Stewart, N. (2021). The association between gambling and financial, social, and health outcomes in big financial data. Nature Human Behaviour, 5, 319-326.

Naomi Muggleton is Assistant Professor of Behavioural Science and lectures on The Economics of Wellbeing on the Undegraduate programme.

Neil Stewart is Professor of Behavioural Sciecen and teaches Behavioural Finance and Big Data on the Global Central Banking and Financial Regulation Qualifications. He also lectures on Big Data Analytics on MSc Management, MSc Finance, MSc Finance & Economics and the suite of MSc Business courses.

Nick Lee is Professor of Marketing and teaches Marketing on the Executive MBA and Distance Learning MBA plus Marketing and Strategy Analytics on MSc Marketing & Strategy.

For more articles on Decision-Making and Analytics sign up to the Core Insights newsletter here.