The barriers to unlocking unstructured data in financial services are falling, thanks to machine learning and data science
Disruptive challenger banks like Monzo and Starling Bank have brought data to the fore in the finance sector – putting pressure on the largest firms in the industry to rethink their digital approach and adopt new data-driven strategies that can improve their offering to customers.
While the sector has struggled with the mountainous task of upgrading legacy systems and embracing digital transformation, there are signs that the tides are changing. In particular, the collection, analysis and subsequent use of data is increasingly becoming a core focus for IT decision makers in the industry.
So far, so structured
Our recent report into machine learning in the sector found that 75 per cent of senior IT decision makers in financial services are using machine learning technology as part of their data strategy. But, it is still largely isolated to the use of structured data. Only three per cent of the same group of decision makers currently use machine learning to analyse unstructured data. To help demonstrate the size of this missed opportunity – as much as 80 per cent of data currently held by the industry as a whole is unstructured.
Getting under the skin of these data sets could help firms understand customers more effectively and make better decisions. It represents a huge opportunity for businesses in the sector to gain a competitive advantage.
Understanding the unstructured
Unstructured data, by its definition, is data that doesn’t exist in a pre-defined model. So, this could include video, audio or entire documents or emails. Since files themselves contain dates, times, facts and opinions they hold a wealth of information. Yet, because this isn’t in a format that can be processed through traditional data analytics, banks have struggled to analyse it.
Other challenges hindering analysis of unstructured data through machine learning involve having the right skill set as well as compliance and an understanding of how secure the analysis of this data can be. However, these barriers are decreasing. Banks are hiring data scientists at pace with the aim of increasing their understanding of digital information. At the same time, the security of cloud platforms that host machine learning are becoming increasingly robust.
Crucially, the use of machine learning is also becoming more straightforward. Google in particular has made the technology underpinning machine learning surprisingly quick to implement and swift results can be achieved once organisations have identified their objectives.
A personal touch
While there may be challenges around unstructured data the rewards are proportional. The rise of data-driven challenger banks has coincided with customers demanding an increasingly tailored service.
We see this trend exemplified in research from Epsilon which reveals that 80 per cent of consumers are now more likely to use a company that offers personalised experiences. In the finance sector this number rises to 89 per cent for online financial institutions and 77 per cent for banks with a physical high street presence. Understanding more about what customers want is key to remaining competitive.
The ability to analyse audio files, documents and audit reports, for example, means that firms can identify patterns and trends amongst their customer base that they were previously unaware of.
This information can then be used to engage with customers, offering them an improved experience. For example, providing chatbots that learn from every interaction to automate and streamline customer service could ensure that customers’ queries are resolved more efficiently and conveniently.
This personalised approach also allows banks to address any issues before they become acute – for instance assessing whether a customer is likely to default, alerting them to it and providing them with advice tailored to their situation. In tandem, switching account trends can be analysed, allowing banks to make predictions about when customers may decide to close their accounts and react accordingly before it happens.
Meanwhile, the detection of fraud exists as its own discrete benefit to analysing data. The benefits here can be demonstrated by London FinTech start-up Ravelin, which has stopped over £100m of fraudulent transactions in three years using a machine learning model that analyses terabytes of historic and real-time data. Use cases like this give customers crucial peace of mind that not only are their finances secure, but that any issue will be resolved quickly with minimal financial damage.
Planning for the future
Encouragingly, there is evidence that financial institutions are beginning to understand that unstructured data may hold the key to future success.
Our teams surveyed senior IT decision makers in financial services and found that 65 per cent of them plan to use machine learning to analyse unstructured data in the next one to two years. In addition to this, a further 15 per cent said that they planned to do the same within the next three to four years.
This is a welcome development. The market is swiftly changing, and an agile approach to innovation and new technology will be a key selling point for many consumers in the future. It will also help improve trust in financial services institutions among customers.
Ultimately, the major players in financial services should look to encourage this activity by maintaining a creative approach to implementing these new insights. By hiring specialist data scientists, they can build data lakes of unstructured data to be analysed through machine learning. This will enable them to leverage the maximum possible benefit for both themselves and customers from the information that they have access to.