What’s holding up widespread data science adoption in banks?


Banks have always had large reservoirs of customer data. The growing number of customer touchpoints is exponentially driving up the volume of data too. Given that, developing the capability to collect, interpret, and leverage it fast enough to drive service and product improvements and improve operational efficiencies is crucial. 

Data science can help banks drive efficiencies and improve effectiveness, whether it is to personalize experiences, optimize cross-sell offers, or make compliance easier and more robust. Given the potential, it shouldn’t come as a surprise to see banks and financial institutions lean in heavily towards data science to improve interactions and engage with their customers in meaningful ways. So, what is the reality of data science adoption?

It’s true that technology-led innovations such as blockchain, open banking, mobile banking, and AI and machine learning are now finding more use cases in this sector. AI chatbots, for example, have evolved and are now capably handling a large chunk of interactions that were once handled by humans. But, it’s fair to say, the sector lags many others in adopting data science.

This then begs the question, if data science delivers such great value then what is holding up its widespread adoption in the banking industry?

Skills outstrip supply

One of the greatest barriers to adopting data science is the growing skills gap in this line of work. Creating data science use cases and a data science roadmap for the bank, developing the right algorithms, and exploring the data to apply the right data and data sets become essential skills to drive these initiatives.

Data science is booming and is all set to create the circumstances to replace many legacy functions. However, while the number of roles needing data scientists is increasing, finding qualified data scientists to develop and run data science initiatives is challenging for banks across the globe. It’s hard for banks to hire, engage, challenge, and retain data scientists who are often keen to seek roles that allow them wider as well as deeper exposure in this specialized area. This is why many banks choose to partner with expert service providers like Ellicium Technologies to address their data science needs through innovative services and solutions models.

Infrastructure challenges

Organizations need access to the right data science platforms and technologies to ensure that data science initiatives drive business outcomes. Data science platforms have specific infrastructure needs. Exploring vast data sets demand high computational capabilities as well…something that the infrastructure of the organization has to support.

The banking sector, however, is still largely working with legacy systems that make it hard to cope with growing data workloads or cater to the demands of modern analytics tools. Collecting, storing, securing, and analyzing huge data volumes with an outdated infrastructure can not only be counter-productive but can put the stability of the entire system at risk. 

Enhancing processing capabilities and re-building their systems to manage the growing volumes of data and enable new-age technology applications and algorithms become essential for developing data science capabilities. Of course, this isn’t an easy task.

However, given the high-stakes involved in the banking sector, it becomes essential to build robust on-premise or cloud infrastructures that fulfills the computational and data needs and allow banks to harness the power of data science.

Navigating the algorithm complexities

Data scientists spend most of their time in high-level languages such as Python/R/SQL. While a number of data science platforms offer ready-to-use, heavily optimized libraries, developing the right algorithms for data science according to use-case can need specialized customization. From putting together the right datasets all the way up to creating algorithms and integrating the analytics findings into business processes is a complex process that most banks seem to find hard to address meaningfully.

Navigating the algorithmic complexity and understanding how well code scales with data, identifying all the elements that go into algorithmic performance, and clearly understanding and managing algorithmic complexities are essential data science elements. Assessing which would be the right language to fulfill algorithmic needs while fulfilling the needs of the domain and business case also become important consideration points that, by themselves, are challenging for most banks.

That apart, rapid production deployment of algorithms, and ensuring regular tuning and maintenance of these algorithms also involve some heavy lifting. Provisioning the right skillsets, building up experience, and training resources for the same remain pain points for banks. 

The “Proof of concepts” pudding

The banking sector has been more cautious in its technology appetite given the strict regulatory and governance landscape along with the high-security needs of this sector. As such, before embarking on the data science journey establishing the right proof of concepts employing the right technologies for data science solutions becomes an inevitable need. 

The absence or the inability to deliver these proof of concepts can damage or impede any data science initiative as it then becomes difficult to identify the RoI and the tangible benefits of using this technology. The proof of the pudding, after all, is in the eating. An associated challenge is the inability to scale beyond the POC because of poor planning. Many banks stay stuck in “POC Purgatory” because they are unable to define the success criteria for the pilot projects or chart roadmaps for the way forward should the pilot succeed. Data science pilots at such banks remain nothing more than costly curiosities or expensive experiments.

While developing in-house data science capabilities is ideal, given the accelerating pace of change, banks have to ensure that they develop these capabilities fast. As the value of data keeps depleting and competition increases, only those who can leverage data science fast can stay ahead of the curve by creating compelling competitive differentiation. If an in-house data science team isn’t making the cut, identifying the right technology partners who can assist the data science journey becomes crucial enablers of organizational and data science success. May we suggest talking to us as a first step on your data science journey?