How Data in Banking and Fintech is Bringing in New Risks

17 January 2017

by Joe Biancalana

Data is playing an increasingly central role within banking. It is key to developing intelligent omnichannel customer interactions, tailored to suit the needs of individuals and households. Data is also powering new technologies, such as AI and bots, which are in turn helping to improve operational efficiency and reduce risks. Data is even enabling new banking models, such as peer-to-peer lending, crowdfunding and the sharing economy. 

The impact of data is best highlighted by looking at the advances in consumer credit. Traditionally, banks primarily rely on credit scores, which are based on a narrow range of slow-moving data points. This modelling approach brings about two major constraints. Firstly, decision-making is slow due to banks having an incomplete view of a consumer’s financial health. Secondly, ‘thin files’ are created, especially on millennial consumers who lack a financial history and have an aversion to debt. In fact, in many countries, the use of consumer credit has been solely negative, whereby banks use the model to essentially blacklist people who have made late payments.

Today, banks are basing their lending and risk management decisions on integrated data. Debt repayment information is being combined with near real-time transactional and account balance data to build thorough risk assessment models. Instead of relying on the timeliness of payments or the percentage of available credit used, they can assess risk patterns from past behaviour to sense future changes.

With ‘thin file’ consumers, banks and credit reporting agencies are leveraging new data sources, such as bill payment history and mobile phone usage. In some cases, particularly with non-bank lenders, the nature of a consumer’s social media network can also contribute to assessing credit worthiness. Data can also be used to tailor sales and marketing interactions. In the same way that it helps banks form a detailed view of a consumer’s credit worthiness, it can also be used to customise sales messages and products for the benefit of increasingly service-savvy customers.

Acting on data

Often, the volume, velocity and range of data types can technically exceed the capabilities of traditional technologies (e.g. relational databases). Unstructured data, such as video, voice and text, are particularly unsuited to former IT approaches and first generation Big Data technologies.

To combat this, banks are adopting machine learning, where predictive models continually train themselves based on streams of data. Machine learning can help identify nuanced details for improved results. For instance, traditional regression or decision tree approaches may predict which customers are likely to churn based on relevant variables. Machine learning goes beyond linear relationships to recognise interactions across much broader sets of data.

Success will rely on cultural change. In light of fast-moving data and the increased pace of change in expectations, a much more iterative approach to planning is required. In particular, the move to agile product development requires a significant shift in product management style.

The challenge to banking technology

Inevitably, the central role of data brings about new risks. People need to understand how to govern and organise for an analytically-driven business. For example, many banks currently only keep data on people who successfully apply for credit. By definition, including only this subset, rather than all who applied, means banks are at risk of reducing their marketing opportunities with new prospects.

However, the most acute risks may be external. Cyber threats are damaging more than just reputation these days and are actually leading to the removal of CEOs, as is the case with Target and Sony. Similarly, senior government and academic leaders are also losing their jobs in response to data breaches. The nature of threats has changed as well, with hackers now seeking physical effects or attempting to discredit an organisation by subtly corrupting, rather than stealing its data.

Solutions for the future

In order to remain competitive, banks need to ensure they have the best security technology to hand. For end-user security, Unisys has worked with Behaviosec on an award-winning biometric authorisation solution that can confirm a consumer’s identity. This is achieved by understanding usage habits, for example, a person’s unique way of holding a phone, their typing speed and the angles they swipe at.

We can also apply machine learning to detect threats. A cyber analytics model can continually ingest large streams of network activity data to define activity baselines and detect anomalies. These models can be applied within an organisation’s cyber security software, as well as integrated with threat intelligence.

The author, Joe Biancalana is Practice Director EMEA Data Analytics, Unisys European Services Limited.

Leave a Reply

Your email address will not be published. Required fields are marked *