CaixaBank is a leader in Spain’s retail banking sector. With more than 5,000 branches and 9,000 ATMs, CaixaBank has one of the largest networks to deliver accessibility and flexibility for its customers. With wire banking prominent between countries throughout the EU and beyond, a strong infrastructure across digital channels is necessary to maintain CaixaBank’s leadership in the marketplace.
Market leadership also means looking forward and providing services faster, as digital access and the emergence of all-digital competitors have transformed the finance industry.
Combining the logistics of a digital infrastructure with the power of machine learning (ML) and artificial intelligence holds the potential to transform the way CaixaBank operates. On the customer side, it was important for CaixaBank to offer services and solutions that stayed ahead of the competition and even reduced cost and effort for customers. Internally, these tools could create a positive impact in many different areas, from customer engagement to fraud detection and streamlining processes.
In 2013, CaixaBank recognized the changing landscape and sought a scalable solution as it looked to the future.
Caixa used a combination of Oracle Machine Learning, Oracle Advanced Analytics, and Oracle Data Mining, in addition to exploratory work with Oracle Analytics Cloud. Oracle's products covered the analytics framework from all angles, from data ingestion and information to data governance to business analytics. This also enabled the integration of ML models with Caixa's big data coming from various resources.
Because the banking industry as a whole was moving toward this digital transformation (especially with startups entering the marketplace with 100% digital offerings), regulatory impacts had to be considered, both for now and the future. Given the inherent computing power required to handle ML and AI and the need to scale up as the volume of big data increases, Caixa recognized several things:
First, any solution had to be scalable for future volumes and growth. Second, current and future regulations regarding security and other guidelines had to be taken into account. And third, data from multiple sources (commercial, retail, internal, research) had to be handled in a way that could offer unified and consolidated access.
Caixa examined many possibilities, but only one solution delivered the unified performance and accessibility the institution desired: Oracle.
There were few people in Spain who knew how to handle a big data project (of this scope). We analyzed the market to learn which solution could be a partner for us. We decided that Oracle was the best option because it offered an integrated solution to build a solid analytics environment.
Director of Big Data Analytical Tools, CaixaBank
With so many different tools and features available under a single platform, working with Oracle meant streamlining internal processes and external access for great accuracy and efficiency. This would take time to properly assemble and build, which is why Caixa performed thorough due diligence before moving forward.
Using a combination of Oracle Machine Learning, Oracle Advanced Analytics, and Oracle Data Mining, plus exploratory work with Oracle Analytics, CaixaBank covered the analytics framework from all angles, from data ingestion and information to data governance and business analytics. This solution also enabled the integration of ML models with CaixaBank’s big data.
Specifically, regarding Oracle Machine Learning, the ability to train and execute many models in the same environment housing the data proved to be invaluable when it came to saving time. From the elimination of data transfers to streamlined deployment, the overall integration maximized efficiency due to the simple fact that models lived in the same environment that housed the data. Machine learning algorithms could also easily be embedded with SQL processes while also being available in R this was particularly important as many Spanish universities teach R to statistics students. Flexibility between languages allowed for faster spreading of ML, which hit one of Caixa's primary goals.
By integrating Oracle's family of products, Caixa Bank built infrastructure and processes to take advantage of the latest emerging technology. And while these steps have created significant improvements for both internal efficiency and customer service, they are just the foundation for the next decade of change. As the Spanish and European financial markets move further into the digital age, everything from security to storage to transaction regulations will evolve. Because Caixa invested in products today, its future is in good hands thanks to Oracle.
With Oracle Machine Learning, CaixaBank saw results quickly across the board in a number of processes. For example, the standard banking process of risk analysis for loan grants was transformed; previously, textbook derivation models were used, but Oracle Machine Learning was able to power sophisticated algorithms. The result was a 7% improvement in the accuracy of models, which translated to a 12% increase in profits on loans.
More-complex situations benefited from simplifying and streamlining processes as well. One example of this is the direct debits process. In Spain, utilities such as electricity, water, and essential home supplies are paid through the bank. This required thousands of human hours of oversight at individual branches. CaixaBank developed and deployed an algorithm trained from thousands of historical decisions, and the result was a 99% accurate match to human decision-making. Once that level of accuracy was achieved, the algorithm took over the task. CaixaBank projected that 60,000 hours of human effort were saved across all branches, allowing employees to spend more time on value-added tasks such as financial advisory, selling products, and services.
Oracle’s solutions have enabled this kind of improvement in many areas of CaixaBank’s operations, both at headquarters and at individual branches. As the ML initiative evolved over six years, perhaps the best showcase of the results is how people view emerging technology. “Six years ago, it was mostly impossible that anyone wondered if machine learning could solve a certain problem,” says Gonzalez. “And now, many people have this question in their mind (when solving problems).”